Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Essays in health economics
(USC Thesis Other)
Essays in health economics
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
ESSAYS IN HEALTH ECONOMICS
by
Jonathan Salcedo
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the Requirements for the Degree
DOCTOR OF PHILOSOPHY
(HEALTH ECONOMICS)
August 2020
Copyright 2020 Jonathan Salcedo
ii
Dedication
For my family.
iii
Acknowledgements
I have many to thank for shaping my graduate school journey into the great learning
experience it was. To my professors and classmates at the California State University,
Fresno, particularly Dr. David Vera, Dr. Qin Fan, and Dr. Antonio Avalos: Thank you for
piquing my curiosity in economics and encouraging me to go further in my studies. Your
courses, mentorship, and guidance prepared me well for graduate school. To the faculty
and participants of the American Economic Association summer training program, thank
you for providing me with a taste of graduate school in a controlled environment. To my
classmates and professors at Duke University, particularly Dr. Charlie Becker and Dr.
Henry Adams, thank you for your mentorship and support. Your help and expertise were
invaluable in navigating the PhD application process.
To my professors and classmates at the University of Southern California, thank
you for helping make the past four years of my life into a bearable process of consistent
improvement. To my dissertation committee members, Dr. Darius Lakdawalla, Dr. Sze-
chuan Suen, and Dr. Bill Padula: thank you for all your guidance and support. Your
research mentorship has helped shape this project from an idea during fellowship to a
work I hope can inform policy. I am extremely grateful. To my coauthors at the MIT
Center for Biomedical Innovation, particularly Dr. Colin Young and Dr. Jenniffer Bulovic:
working with you was a pleasure. I am happy to say your expertise has greatly
expanded my understanding of the financing challenges surrounding cell and gene
therapy in the United States.
Lastly, I would like to acknowledge the USC School of Pharmacy, USC Graduate
School, and MIT Center for Biomedical Innovation for providing fellowship, travel, and
iv
other support. Not surprisingly, it is significantly easier to study financing challenges
when they are absent in your own academic life.
v
Contents
DEDICATION ................................................................................................................... ii
ACKNOWLEDGEMENTS ............................................................................................... iii
LIST OF TABLES ........................................................................................................... vii
LIST OF FIGURES .......................................................................................................... ix
ABSTRACT ..................................................................................................................... xi
LIST OF ABBREVIATIONS ........................................................................................... xii
CHAPTER 1: INTRODUCTION ....................................................................................... 1
Background ................................................................................................................. 1
Aims ............................................................................................................................ 3
Hypotheses ................................................................................................................. 3
References .................................................................................................................. 6
CHAPTER 2: THE LIFETIME COSTS OF SICKLE CELL DISEASE IN THE UNITED
STATES .......................................................................................................................... 8
Abstract ....................................................................................................................... 8
Introduction ................................................................................................................. 9
Literature Review ...................................................................................................... 10
Literature Review Results ......................................................................................... 11
Other Methods for Estimating Lifetime Costs ............................................................ 14
Discussion and Limitations ........................................................................................ 23
Conclusions ............................................................................................................... 25
References ................................................................................................................ 26
Appendix ................................................................................................................... 30
CHAPTER 3: PEDIATRIC INPATIENT COSTS OF SELECT ORPHAN
CONDITIONS WITH POTENTIAL GENE THERAPY LAUNCHES: A CROSS-
SECTIONAL ANALYSIS OF 2016 ................................................................................ 37
Abstract ..................................................................................................................... 38
Introduction ............................................................................................................... 39
Background ............................................................................................................... 40
Methods .................................................................................................................... 50
Results ...................................................................................................................... 51
Discussion ................................................................................................................. 54
Conclusions ............................................................................................................... 55
References ................................................................................................................ 57
Appendix ................................................................................................................... 61
vi
CHAPTER 4: THE TOTAL DIRECT COST OF HEALTHCARE IN THE UNITED
STATES FOR COMMERCIALLY INSURED PATIENTS WITH SICKLE CELL
DISEASE: AN AGE-SPECIFIC ANALYSIS ................................................................... 63
Abstract ..................................................................................................................... 63
Introduction ............................................................................................................... 64
Methods .................................................................................................................... 66
Results ...................................................................................................................... 74
Discussion and Limitations ........................................................................................ 88
Conclusions ............................................................................................................... 91
References ................................................................................................................ 92
Appendix ................................................................................................................... 99
CHAPTER 5: COST-EFFECTIVENESS OF A HYPOTHETICAL DURABLE
TREATMENT FOR SICKLE CELL DISEASE .............................................................. 114
Abstract ................................................................................................................... 114
Introduction ............................................................................................................. 116
Methods .................................................................................................................. 117
Results .................................................................................................................... 134
Discussion and Limitations ...................................................................................... 144
Conclusions ............................................................................................................. 147
References .............................................................................................................. 149
Appendix ................................................................................................................. 154
CHAPTER 6: CONCLUSIONS AND POLICY IMPLICATIONS ................................... 167
Conclusions ............................................................................................................. 167
Aims and Hypotheses Revisited .............................................................................. 168
Policy Implications ................................................................................................... 172
References .............................................................................................................. 177
vii
List of Tables
TABLE PAGE
1.1 Select cell and gene therapy trials applicable to sickle cell disease .......................... 2
2.1 Patients, prognostic factors, and outcomes (PPO) table ......................................... 11
2.2 Included studies and detailed characteristics .......................................................... 14
A.2.1 All studies identified in PubMed (PMC and MEDLINE) ........................................ 31
3.1 Epidemiology and notable costs for orphan diseases of interest ............................. 43
3.2 Costs and inpatient resource utilization by orphan disease and payer, 2016 .......... 53
A.3.1 Select gene therapy clinical trials for orphan conditions of interest ...................... 61
4.1 Baseline characteristics and annualized healthcare costs by specification and
study arm ...................................................................................................................... 77
4.2 Annualized and yearly average incremental cost of SCD by age group and time
post-index (unadjusted) ................................................................................................. 80
4.3.a Generalized linear model of annualized total costs, gamma family with log-link,
females .......................................................................................................................... 82
4.3.b Generalized linear model of annualized total costs, gamma family with log-link,
males ............................................................................................................................. 83
4.4 Pooled severity lifetime cost estimates of patients with SCD and controls .............. 87
A.4.1 Patient level variables, values, and definitions ..................................................... 99
A.4.2 Diagnosis codes of sickle cell disease, ICD-9-CM and ICD-10-CM ................... 101
A.4.3 Goodness of fit comparisons for generalized linear models ............................... 102
A.4.4 Standardized mean differences for matching variables by group....................... 103
A.4.5 Regression predicted annualized cost by exposure and subgroup (2018 US
dollars) ........................................................................................................................ 104
5.1 Model input parameters and probabilistic sensitivity analysis distributions ........... 122
viii
TABLE (CONT.) PAGE
5.2.a Ordered logistic regression of second year severity, females ............................ 130
5.2.b Ordered logistic regression of second year severity, males ............................... 130
5.3 Discounted and undiscounted base case and scenario analysis results,
per-patient ................................................................................................................... 136
A.5.1 Univariate deterministic sensitivity analysis results on ICER, ordered by
base case model sensitivity ......................................................................................... 154
ix
List of Figures
FIGURE PAGE
2.1 CONSORT diagram of included and excluded studies ............................................ 12
2.2 Conceptual Markov model of disease progression .................................................. 18
3.1 Mean inpatient stay costs (95% CI), by disease area (2016) .................................. 52
A.3.1 Nationally representative aggregated inpatient stay costs for select diseases,
by payer (2016) ............................................................................................................. 62
4.1 Study observation period for patients with SCD and matched controls ................... 68
4.2 Inclusion and exclusion flow chart ........................................................................... 75
4.3 Composition of annualized average incremental cost by age at first
disease-related claim .................................................................................................... 79
4.4.a GLM (gamma, log-link) predictive margins of total annualized healthcare cost
with 95% confidence intervals by severity, females ...................................................... 85
4.4.b GLM (gamma, log-link) predictive margins of total annualized healthcare cost
with 95% confidence intervals by severity, males ......................................................... 86
A.4.1.a Empirical distributions of matching variables, in patients with sickle cell
disease and controls ................................................................................................... 107
A.4.1.b Empirical distributions of matching variables in patients with sickle cell
disease and controls ................................................................................................... 108
A.4.1.c Empirical distributions of matching variables in patients with sickle cell
disease and controls ................................................................................................... 109
A.4.1.d Empirical distributions of matching variables in patients with sickle cell
disease and controls ................................................................................................... 110
A.4.1.e Empirical distributions of matching variables in patients with sickle cell
disease and controls ................................................................................................... 111
A.4.2.a Linear predictive margins with 95% confidence intervals by severity,
females ........................................................................................................................ 112
x
FIGURE (CONT.) PAGE
A.4.2.b Linear predictive margins with 95% confidence intervals by severity,
males ........................................................................................................................... 113
5.1 Markov process for lifetime management of SCD ................................................. 121
5.2 Markov trace for SOC and DT arms (N=10,000 theoretical patients) .................... 137
5.3 Tornado diagram of deterministic sensitivity analysis results (top ten) .................. 139
5.4 Two-way deterministic sensitivity analyses of select model parameters ............... 141
5.5 Probabilistic sensitivity analysis results (A) Cost-effectiveness acceptability
curves (CEAC) and frontier (CEAF); (B) Expected value of perfect information
(EVPI) .......................................................................................................................... 143
A.5.1.a Ordered logit predicted probabilities of transitioning to mild SCD by gender,
age, and state ............................................................................................................. 156
A.5.1.b Ordered logit predicted probabilities of transitioning to moderate SCD by
gender, age, and state ................................................................................................ 157
A.5.1.c Ordered logit predicted probabilities of transitioning to severe SCD by
gender, age, and state ................................................................................................ 158
A.5.2.a Markov trace for SOC and DT arms, females ................................................. 159
A.5.2.b Markov trace for SOC and DT arms, males .................................................... 160
A.5.3.a Monte Carlo draws (N=10,000), Cost parameters and coefficients for GLM
log-link gamma family annualized total cost regressions............................................. 161
A.5.3.b Monte Carlo draws (N=10,000), Utility parameters ......................................... 162
A.5.3.c Monte Carlo draws (N=10,000), Initial condition and transition probability
parameters: coefficients for ordered logit transition probability regressions ................ 163
A.5.4 Tornado diagram of deterministic sensitivity analysis results ............................. 164
A.5.5.a PSA results among females (A) Cost-effectiveness acceptability curves
(CEAC) and frontier (CEAF); (B) Expected value of perfect information (EVPI) .......... 165
A.5.5.b PSA results among males (A) Cost-effectiveness acceptability curves
(CEAC) and frontier (CEAF); (B) Expected value of perfect information (EVPI) .......... 166
xi
Abstract
Sickle cell disease (SCD) is an inherited blood disorder with significant healthcare
resource utilization. It is commonly associated with frequent complications including
pain crises, infections, acute chest syndrome, and stroke, all of which lead to reduced
life expectancy. Treatment for SCD has evolved significantly over past decades in the
United States (US), with gene therapy treatment currently in trials and showing promise.
i
Lifetime costs for managing SCD are uncertain.
This dissertation encompasses three aims. These include determining: (1) the
magnitude and composition of direct healthcare costs faced by patients with SCD in the
US, (2) a hypothetical lifetime economic burden for managing SCD, and (3) the potential
cost-effectiveness of a cure provided at birth, and value of information (VOI) of obtaining
additional data on the burden of SCD. We propose addressing these aims using a large
database of private insurance claims data (Optum’s de-identified Clinformatics® Data
Mart Database) supplemented by other data sources.
We estimate age-specific healthcare costs using real-world evidence (RWE) of
patients with SCD relative to propensity-score matched control patients. We then use
these estimates to simulate the incremental economic burden associated with the
disease over a lifetime. The cost-effectiveness of a durable treatment can be evaluated
based on these lifetime costs and quality-adjusted life years, assuming the cure is
provided at birth. VOI analysis provides additional inference regarding uncertainty
around these figures and may inform future RWE generation efforts in SCD.
i
Rubin R. Gene Therapy for Sickle Cell Disease Shows Promise. JAMA. 2019;321(4):334-334.
xii
List of Abbreviations
Abbreviation
ABM Agent-based model
ACS Acute chest syndrome
ADA Adenosine deaminase deficiency
AIC Akaike information criterion
AIC Average incremental cost
AK Alaska
AL Alabama
ALL National ancillaries contracted for all products
AN Ancillary
AR Arkansas
ASO Administrative services only
AZ Arizona
BIC Bayesian information criterion
BMT Bone marrow transplant
BSC Best supportive care
BT Beta thalassemia
CA California
CBA Cost-benefit analysis
CCR Cost-to-charge ratio
CDC The Centers for Disease Control and Prevention
CDHP Consumer driven health plan
CDM Clinformatics® Data Mart
CEA Cost-effectiveness analysis
CEAC Cost-effectiveness acceptability curve
CEAF Cost-effectiveness acceptability frontier
CF Cystic fibrosis
CFTR Cystic fibrosis transmembrane conductance regulator
CHEERS Consolidated Health Economic Evaluation Reporting Standards
CHIP Children’s Health Insurance Program
CI Confidence interval
CMS Centers for Medicare and Medicaid Services
CO Colorado
COM Commercial plan
CONSORT Consolidated Standards of Reporting Trials
CONT Continued
CPAP Continuous positive airway pressure
CPI Consumer price index
CT Connecticut
DC District of Columbia
DE Delaware
DIV Geographic division
DNA Deoxyribonucleic acid
xiii
DSA Deterministic sensitivity analysis
DT Durable therapy
EHR Electronic health record
EPO Exclusive provider organization
EQ-5D EuroQol 5 Dimensions
ERT Enzyme replacement therapy
ESRD End-stage renal disease
EVPI Expected value of perfect information
FDA Food and Drug Administration
FL Florida
GA Georgia
GD Gaucher disease
GDP Gross domestic product
GLM Generalized linear model
GPO Group purchasing organization
HAA Hemophilia A
HAB Hemophilia B
HCC Hepatocellular carcinoma
HCT Hematopoietic cell transplantation
HCUP Healthcare Cost and Utilization Project
HCV Hepatitis C virus
HI Hawaii
HIV Human immunodeficiency virus
HMO Health maintenance organization
HRA Health reimbursement account
HSA Health savings account
HSCT Hematopoietic stem cell transplant
IA Iowa
ICD-10-CM International Classification of Diseases, Tenth Revision, Clinical
Modification
ICD-9-CM International Classification of Diseases, Ninth Revision, Clinical
Modification
ICER Incremental cost-effectiveness ratio
ICER Institute for Clinical and Economic Review
ID Identification document
IL Illinois
IN Indiana
IND Indemnity
IP Inpatient
IPP Individual program plan
IPW Inverse probability weighting
IRB Institutional Review Board
ITT Intention-to-treat
IV Intravenous therapy
IVIG Intravenous immunoglobulin
KID Kids' Inpatient Database
xiv
KS Kansas
KY Kentucky
LA Louisiana
LB Lower bound
LEB Life expectancy at birth
LOS Length of stay
LR Likelihood ratio
LY Life year
MA Massachusetts
MCPI Medical component of consumer price index
MCR Medicare Advantage plan
MD Maryland
ME Maine
MEDLINE Medical Literature Analysis and Retrieval System Online
MEPS Medical Expenditure Panel Survey
MI Michigan
MIL Million
MN Minnesota
MO Missouri
MPS Mucopolysaccharidosis
MRI Magnetic resonance imaging
MS Mississippi
MT Montana
NBS Newborn screening
NC North Carolina
NCHS National Center for Health Statistics
NCT National Clinical Trial
ND North Dakota
NE Nebraska
NH New Hampshire
NHIS National Health Interview Survey
NJ New Jersey
NM New Mexico
NMB Net monetary benefit
NONE No industry product code
NV Nevada
NY New York
OH Ohio
OK Oklahoma
OP Outpatient
OR Oregon
OTCD Ornithine transcarbamylase deficiency
OTH Other
PA Pennsylvania
PD Pompe disease
PF Professional
xv
PMC PubMed Central
PMID PubMed identifier
POS Point of service
PPO Preferred provider organization
PPV Positive predictive value
PSA Probabilistic sensitivity analysis
PSM Propensity score matching
PV Present value
QALY Quality-adjusted life year
RCT Randomized controlled trial
RI Rhode Island
RWE Real-world evidence
RX Pharmacy
SC South Carolina
SCA Sickle cell anemia
SCD Sickle cell disease
SCHIP State Children’s Health Insurance Program
SCID Severe combined immunodeficiency
SD Standard deviation
SE Standard error
SEER Surveillance, Epidemiology, and End Results
SF-12 12-item Short Form
SMA Spinal muscular atrophy
SMD Standard mean difference
SOC Standard of care
SPN State policy network
TN Tennessee
TX Texas
UB Upper bound
UNK Unknown
UPIRB University Park Institutional Review Board
US United States of America
USA United States of America
USD United States dollar
UT Utah
VA Virginia
VAS Visual analog scale
VBA Visual Basic for Applications
VG Von Gierke disease
VOC Vaso-occlusive crisis
VOI Value of information
VSL Value of a statistical life
VT Vermont
WA Washington
WHO World Health Organization
WI Wisconsin
xvi
WTP Willingness-to-pay
WV West Virginia
WY Wyoming
1
Chapter 1
Introduction
Background
Incurable and ongoing chronic diseases affect approximately 133 million persons living
in the United States.
1
The Centers for Disease Control and Prevention (CDC) report that
90% of the annual health care expenditures in the US are for people with chronic
conditions.
2-4
With improvements in medical technology, it is likely several conditions
that are chronically managed over a lifetime will become curable through acute or even
single administration therapies. Hepatitis C virus (HCV) is a modern example of this.
5
Understanding the economic impact of therapies that cure patients and restore current
and future health involves accounting for many factors.
6
While the feasibility, access, and costs of single-dose or short-term
administration therapies will vary greatly across disease areas, sickle cell disease
(SCD) provides a compelling case study. There are an estimated 100,000 prevalent
cases in the US alone for this heritable disorder, a relatively large number for an
orphan
ii
disease.
7
It manifests in chronic and debilitating effects and has few disease-
modifying therapies currently available.
8
The only available cure, hematopoietic stem
cell transplant (HSCT) is typically available to fewer than 20% of patients and carries
significant risks.
9
Looking forward, there are several cell and gene therapies currently in
ii
The US defines an orphan disease as one affecting fewer than 200,000 persons (Orphan Drug
Act of 1983).
2
the US Food and Drug Administration (FDA) pipeline aimed towards curing SCD (see
Table 1). In this chapter we briefly outline the aims and hypotheses of this dissertation.
Table 1.1 Select cell and gene therapy trials applicable to sickle cell disease
Drug name Originator Phase Status* NCT
ARU-1801 Aruvant Sciences
GmbH
Phase 1/2 Recruiting NCT02186418
BIVV003 Bioverativ
Therapeutics Inc.
Phase 1/2 Recruiting NCT03653247
LentiGlobin™
BB305 (Zynteglo)
bluebird bio, Inc. Phase 1/2 Active, not
recruiting
NCT02140554
CTX001 Vertex
Pharmaceuticals Inc.
Phase 1/2 Recruiting NCT03745287
*Status on ClinicalTrials.gov as of February 24, 2020. Abbreviations: NCT, National
Clinical Trial.
A cure for SCD has the potential to generate tremendous societal welfare gains
through improved quality of life, life expectancies, and productivity. Despite likely health
gains, the overall cost impact of therapy is uncertain and must be evaluated with offsets
in chronic disease management and other spillover effects considered. In this
dissertation we study (i) the age-specific economic burden of SCD in the US, (ii) the
lifetime economic burden relative to a comparable control population, and (iii) the
hypothetical cost-effectiveness of a cure. This chapter serves to outline the high-level
aims and hypotheses of the dissertation, which we present as a series of self-contained
essays in the economics of sickle cell disease.
3
Aims
Aim 1. Determine the magnitude and composition of direct healthcare costs faced by
patients with sickle cell disease in the US. This includes an age-specific analysis to
inform how disease management costs change by stage of life.
Aim 2. Determine the hypothetical lifetime burden for a patient born with sickle cell
disease in the US. Use matched control patients to conduct counterfactual inference.
Aim 3. Estimate the cost-effectiveness of a hypothetical cure for sickle cell disease
provided to patients at birth. Determine uncertainty associated in cost-effectiveness
estimates and analyze the value in obtaining additional data on sickle cell disease
burden.
Hypotheses
Hypothesis 1. There is a substantial direct economic burden associated with chronic
management of sickle cell disease over a lifetime for patients born and treated in the
US. The most commonly cited previous estimate
iii
in the literature likely underestimates
true current burden due to population, cost measure, and technological advances.
10
Hypothesis 2. A hypothetical cure provided at birth is likely to result in increased life
years, increased quality-adjusted life years, and decreased direct healthcare
expenditures over a lifetime.
iii
Kauf et al. (2009) estimates lifetime cost of care (discounted at 3%) of $460,151 (2005 USD)
for patients in a Florida Medicaid program.
4
Hypothesis 3. A cure for sickle cell disease will be cost-effective to the healthcare
sector at willingness to pay of $150,000/QALY, given consideration to variability in
incremental costs, quality of life, and mortality.
In the following chapters, we address these aims and test our hypotheses. Our first aim
is important for understanding current costs faced by the healthcare system in
managing patients with chronic SCD. Age has been shown to be associated with
inpatient lengths of stay and episodic severity in SCD; both drivers for healthcare
utilitization.
10-12
For this reason, cost estimation in SCD should account for age in
addition to disease severity. We hypothesize that age-specific estimates available in the
literature may not be applicable to privately insured individuals. A reason for this is the
substantially different patient populations; however, even studies which control for
observable characteristics have found higher costs for those covered under private
plans.
13,14
Our second aim is to take age-specific costs and estimate a hypothetical lifetime
burden of a patient with SCD as compared to a similar patient unaffected by the
disease. This counterfactual inference is important for understanding economic burden
that can be directly attributed to managing SCD. In the scenario in which a cure
becomes available, we must understand the costs that are likely to be offset by curing a
patient. It is straightforward for us to hypothesize a cure would improve life expectancy,
quality of life, and (non-cure related) healthcare costs over a lifetime.
For our third and final aim, we intend to estimate the potential cost-effectiveness of a
cure for SCD from the US healthcare sector perspective. Given the large and persistent
5
economic and humanistic burdens faced by many patients with SCD, we hypothesize
that a cure is cost-effective at a willingness to pay threshold of $150,000 per QALY.
Despite the high upfront costs associated with single administration curative therapies,
e.g. 2.1 million USD most recently for a spinal muscular atrophy cure, downstream
benefits are likely to be large.
15
Patients cured in the first decade of life may avoid costly
health care utilization, prevent irreparable organ damage, and experience substantially
improved quality-adjusted life expectancies.
6
References
1. Centers for Disease Control and Prevention. The Power of Prevention. 2009;
http://www.cdc.gov/chronicdisease/pdf/2009-Power-of-Prevention.pdf.
2. Buttorff C, Ruder T, Bauman M. Multiple chronic conditions in the United States.
2017.
3. Centers for Medicare and Medicaid Services. National Health Expenditure Data
for 2016—Highlights. 2017.
4. Centers for Disease Control and Prevention. Health and Economic Costs of
Chronic Diseases. 2019;
https://www.cdc.gov/chronicdisease/about/costs/index.htm.
5. Voelker R. The 8-Week Cure for Hepatitis C. JAMA. 2017;318(11):996-996.
6. Drummond MF, Neumann PJ, Sullivan SD, et al. Analytic Considerations in
Applying a General Economic Evaluation Reference Case to Gene Therapy.
Value Health. 2019;22(6):661-668.
7. Hassell KL. Population estimates of sickle cell disease in the U.S. Am J Prev
Med. 2010;38(4 Suppl):S512-521.
8. Kapoor S, Little JA, Pecker LH. Advances in the Treatment of Sickle Cell
Disease. Mayo Clin Proc. 2018;93(12):1810-1824.
9. Ikawa Y, Miccio A, Magrin E, Kwiatkowski JL, Rivella S, Cavazzana M. Gene
therapy of hemoglobinopathies: progress and future challenges. Hum Mol Genet.
2019;28(R1):R24-R30.
7
10. Kauf TL, Coates TD, Huazhi L, Mody-Patel N, Hartzema AG. The cost of health
care for children and adults with sickle cell disease. Am J Hematol.
2009;84(6):323-327.
11. Bou-Maroun LM, Meta F, Hanba CJ, Campbell AD, Yanik GA. An analysis of
inpatient pediatric sickle cell disease: Incidence, costs, and outcomes. Pediatr
Blood Cancer. 2018;65(1).
12. Panepinto JA, Brousseau DC, Hillery CA, Scott JP. Variation in hospitalizations
and hospital length of stay in children with vaso-occlusive crises in sickle cell
disease. Pediatr Blood Cancer. 2005;44(2):182-186.
13. Hadley J, Holahan J. Is health care spending higher under Medicaid or private
insurance? Inquiry. 2003;40(4):323-342.
14. Ku L, Broaddus M. Public and private health insurance: stacking up the costs.
Health Aff (Millwood). 2008;27(4):w318-327.
15. Rosenmayr-Templeton L. Industry update for May 2019. Ther Deliv.
2019;10(9):555-561.
8
Chapter 2
The Lifetime Costs of Sickle Cell Disease in the
United States
Abstract
Understanding costs to patients, providers, and society for managing chronic conditions
over a lifetime is an important topic for understanding value of curative therapy. Sickle
cell disease (SCD) is a chronic orphan disease with an estimated 100,000 prevalent
cases in the United States. There are multiple durable therapies for SCD in clinical trials
seeking Food and Drug Administration (FDA) approval for sale in the US. Despite this,
the extent to which current disease management costs vary by age and overall lifetime
cost of treatment in SCD are not well understood.
In the first part of this chapter, we conduct a literature review of clinical studies
published since 2009. Here we identified three papers with lifetime estimates. Direct
cost estimates varied widely across studies, from $460,151
iv
per patient over 45 years
to $8,747,908
v
per patient over 50 years. One study reported a lifetime indirect cost
estimate of $695,000
vi
in lost earnings per patient over a simulated life expectancy of 54
years. There does not appear to be a standardized methodology or measure of cost
used in the lifetime costing literature for SCD. We find a clear need for additional and
iv
2005 USD discounted at 3% per year, Kauf et al. (2009)
v
Presumably 2009 USD and unadjusted, Ballas (2009)
vi
2014 USD unadjusted, Lubeck et al. (2019)
9
directed research on the direct and indirect economic burden of SCD in the United
States.
In the second part of this chapter we summarize methods utilized in the health
economic evaluation literature to estimate healthcare costs in chronic diseases over a
lifetime. We describe strengths and weaknesses of each approach and explore potential
applications in SCD. We find that the lack of detailed longitudinal data through sources
like a disease registry necessitates the use of modeling and simulation approaches to
estimate lifetime costs in SCD. The application of these standardized costing
methodologies may provide more reliable estimates of SCD lifetime economic burden.
Introduction
Sickle cell disease (SCD) is an orphan disease
vii
with an estimated prevalence of up to
100,000 affected patients in the United States (US).
1
Often used interchangeably, sickle
cell anemia (SCA) is the most common and severe form of the disease and is
characterized by two hemoglobin S genes (hemoglobin SS).
2
The clinical management
throughout the course of the disease is variable. Children born with SCD in the US
receive routine comprehensive care that significantly reduces their morbidity and
mortality. Upon transition to adulthood, there are marked increases in disease-related
complications and morbidity. The age-specific and lifetime costs of chronic management
in this population are not well understood. These estimates are often of interest to
understand the costs that can be offset by curing a patient. They may also serve as
vii
The US defines an orphan disease as one affecting fewer than 200,000 persons (Orphan
Drug Act of 1983).
10
intermediate steps within studies such as cost-benefit analyses (CBA), cost-
effectiveness analyses (CEA), or other forms of economic evaluation.
In this chapter we first conducted a literature review of PubMed (MEDLINE and
PMC databases) to identify studies which provide age-specific and lifetime cost
estimates for patients with SCD treated in the US. In the second part of this chapter, we
describe methods applied to generate lifetime cost estimates for the chronic
management of other diseases. We discuss their strengths and weaknesses and
describe their potential in determining lifetime costs in SCD. Lastly, we briefly describe
how lifetime estimates of treatment costs may translate to value of curative therapies.
Literature review
We searched the PubMed database, which includes MEDLINE and PubMed Central
(PMC), for published studies between January 1, 2009 and November 30, 2019. Our
search terms included the terms sickle cell disease, sickle cell anemia, costs, economic
burden, income, lifetime, age-related, and age-specific, and select variations. A
complete PubMed search string including Medical Subject Headings (MeSH) is
available in the Appendix.
We included only published peer-reviewed studies in our literature review.
Studies of interest included prospective cohort studies, retrospective database studies,
decision-analytic modeling studies, meta-analyses, original reviews, and expert
commentaries. Excluded categories of works included unpublished studies, gray
literature, and conference abstracts and/or posters. We also excluded studies that:
studied animal subjects, were not in English, had a non-US setting, did not report costs
11
in USD, only studied patients with sickle cell trait (symptomatic or asymptomatic), or did
not report a lifetime estimate. See Table 2.1 for our patients, prognostic factors, and
outcomes (PPO) table.
Table 2.1 Patients, prognostic factors, and outcomes (PPO) table
Included Excluded
Patients
Patients with any form of sickle
cell disease (SCD) treated in a
US setting
Patients solely with sickle
cell trait (symptomatic or
asymptomatic); all other
diagnoses
Prognostic factors Any N/A
Outcome
Age-specific and lifetime cost of
healthcare resources and
disease treatment
All other outcomes
Abbreviations: SCD, sickle cell disease; N/A, not applicable.
Literature review results
Of 158
viii
published peer-reviewed studies from the medical literature, 6 satisfied title
and abstract screening and were chosen for full-text review. A list of all studies identified
is available in Appendix Table 1. Our final sample included 3 papers which satisfied all
inclusion and exclusion criteria (see Figure 1 for our CONSORT diagram).
viii
157 studies were identified via search in PubMed. We identified one additional study for
inclusion as a published commentary of an included study.
12
Figure 2.1 CONSORT diagram of included and excluded studies
Included studies included a retrospective database analysis, a literature-based decision
modeling study, and an expert commentary. Two studies reported direct costs. Kauf et
al. (2009) studied patients with SCD in a Florida Medicaid program between 2001 and
2005.
3
The authors estimated SCD-related and other healthcare costs, finding that
51.8% of care was disease related. By taking mean age-specific total costs and
constructing a theoretical cohort, they find a present value of lifetime cost of care
(discounted at 3%) of $460,151 per patient (2005 USD)
ix
. Ballas (2009) assumed a 50-
year life expectancy at birth (LEB) and used fee tables to estimate total fees of
$8,747,908 per patient (presumably 2009 USD)
x
over a lifetime with SCD.
4
Ballas
ix
This figure is $690,115 per patient when inflated to 2018 USD using the medical services
component of the CPI (MCPI).
x
Ballas’ estimate is $11,291,023 in 2018 USD when inflated using MCPI.
Excluded (N=152)
Did not meet all study criteria
Title and abstract review
(NPUBMED=157; NEXTERNAL=1)
Full text review
(N=6)
Final studies
(N=3)
Excluded (N=3)
No lifetime cost estimates
13
acknowledges Kauf et al.’s figure and attributes the part of the discrepancy to
differences in charges (what Ballas measured), Medicaid reimbursement (what Kauf
measured), and true costs. Both Kauf et al. and Ballas lacked appropriate comparison
groups of patients unaffected by the disease.
The only study reporting indirect costs, Lubeck et al. (2019) utilized a cohort
simulation modeling approach parametrized by publicly available information and data
from the medical literature to estimate lifetime incomes for SCD and matched non-SCD
cohorts.
5
Their approach involved estimating LEBs for each arm and aggregating
incomes for each (starting at age 15) using sex- and race-adjusted data from the US
Bureau of the Census. Lifetime income estimates include $1,227,000 for the SCD
cohort and $1,922,000 for the matched controls – resulting in a $695,000 difference
attributable to early mortality (2014 USD)
xi
. The authors claim an undiscounted figure is
appropriate to approximate real income over a lifetime, given they also do not adjust for
inflation. See Table 2.2 for included studies and their detailed characteristics.
xi
This difference is $779,642 when inflated to 2018 USD using the Social Security Wage Index.
14
Table 2.2 Included studies and detailed characteristics
Study Population
Method Lifetime
Estimate
Type
Lifetime
Estimate
Value
Estimate
Year
Lubeck et al.
2019
5
SCD patient
projected to
live to 54
years
PV using US
Bureau of
Census data
Unadjusted
lost
earnings
(indirect)
$695,000 2014
Kauf et al. 2009
3
SCD patient
assumed to
live to 45
years
PV using age-
specific FL
Medicaid data
Discounted
cost of
healthcare
(direct)
$460,151 2005
Ballas 2009
4
SCA patient
assumed to
live to 50
years
Simple
aggregation of
assumed fees
for life care plan
Unadjusted
total fees
(direct)
$8,747,908 2009
xii
Abbreviations: SCD, sickle cell disease; PV, present value; FL, Florida.
Other methods for estimating lifetime costs
Developing lifetime cost estimates involves understanding the natural history of a
disease and its treatment phases. In addition, proper analytical and numerical methods
are necessary to accurately estimate lifetime costs under chronic disease. In this
section we describe how information from clinical trials, patient registries, longitudinal
cohorts, and other data involving human subjects contribute to the understanding of
survival and healthcare resource utilization in chronic disease. We describe analytical
and simulation methods commonly employed to estimate costs in the presence of
censored or limited data. Lastly, we briefly describe how lifetime estimates of treatment
costs translate to value of curative therapies.
xii
Not explicitly stated in Ballas 2009.
15
Statistical methods and survival analysis
Randomized clinical trials (RCTs) often use overall survival as a clinical endpoint when
evaluating treatments in chronic disease, for instance in oncology.
6
For this reason,
average or median outcomes for a standard of care (SOC) or placebo arm serve as an
appropriate approximation for patients living with the disease. However, using this
approach for a lifetime estimate is not straightforward. A main challenge with using
survival from RCTs is horizon; patients are not always followed for multiple years. In
addition, RCTs suffer from attrition and data may not reflect all desired subgroups.
Survival analysis addresses these points in RCT data with estimators such as the
Kaplan-Meier which can estimate the survival function in the presence of right-censoring
and the Cox proportional hazards model which can provide covariate-adjusted
estimates.
7,8
However, another limitation is that resource utilization and costs are often
not tracked or reported alongside clinical trials.
9
This means that while RCTs often serve
as a basis for information on treatment and associated survival, costs must be imputed
from other sources.
In addition to trial data, statistical methods are often applied to other health and
cost data sources. In conditions where patients are tracked in registries, e.g.
Surveillance, Epidemiology, and End Results (SEER) for cancer, it is possible to
estimate costs using available survival and treatment information.
10
Many estimators
have been proposed to estimate lifetime costs from longitudinal data such as in claims
databases or registries.
11-14
These estimators deal with censored at random cost data
which are often skewed with varying periods of observation. Many of these methods are
based on inverse probability weighting (IPW). Application to disease areas which do not
16
have population-based surveillance systems or registries in place is difficult due to lack
of outcomes data linked to death records.
Life table methods
In the absence of patient-level mortality data, it is not possible to conduct survival
analysis where the failure event is death. A simple approach to estimate a lifetime cost
in this scenario is to assume an average LEB and sum costs at each age up to LEB,
discounting when necessary. This approach was applied in SCD by Kauf et al. (2009)
using costs from a Florida Medicaid program and Ballas (2009) using institution-specific
fee tables.
3,4
A primary limitation of assuming a mean LEB and calculating the
discounted present value of future costs at birth is patient heterogeneity in longevity.
This method assumes all patients experience the mean LEB; if healthcare costs are
correlated with longevity, estimates will be biased. Another scenario applies when life
tables or age-dependent mortality rates are available. A recent analysis by Lubeck et al.
(2019) applied a variation of this method by estimating smoothed mortality rates by age
using Poisson regression on all-cause death data for patients with SCD. The authors
then used the smoothed estimates to transition a cohort from birth to death, tracking
income at each age. This method addresses the heterogeneity limitation of assuming a
LEB.
Markov models
Markov models are commonly used tools for modeling recurring events. They are often
applied in health and medicine to simulate outcomes over lifetime horizons. Markov
17
models involve mutually exclusive and exhaustive states. Key properties include the
memoryless property which states that transitions must only depend on the current
state, and homogeneity meaning all individuals in the same state experience the same
rewards and transition probabilities. Despite the memoryless property, there exist
methods for incorporating forms of “memory” into a Markov model. Tunnel states are
temporary states which capture patients who have experienced a certain event and
funnel them into a different set of states.
15
Second and higher-order Markov models
xiii
also allow transitions to depend on previous states. For an 𝑛 𝑡 ℎ
order Markov model, the
transition probabilities depend on the current and 𝑛 − 1 previous states visited.
The general steps in building a Markov model to estimate lifetime costs involve (1)
defining the health states and possible transitions, (2) defining the horizon and cycle
length, (3) attaching rewards (costs) to each state, (4) defining all possible transition
probabilities, i.e. 𝑃 (𝑆𝑡𝑎𝑡 𝑒 𝑖 ,𝑡 +1
|𝑆𝑡𝑎𝑡 𝑒 𝑗 ,𝑡 ) for all i and j, and (5) evaluating outcomes by
multiplying a vector of initial state membership (1 𝑥 𝐾 for 𝐾 health states) by the
transition matrix (𝐾 𝑥 𝐾 ) until the horizon is satisfied. In chronic disease, this may
involve a model like Figure 2.2. In this Markov model (𝐾 = 4), patients are born with a
mild condition which is incurable. Throughout life, their disease may stay mild or
progress to moderate or severe disease. At any point there is a nonzero probability of
death, which is dependent on the state of their disease.
xiii
Second and higher order Markov models are uncommon in economic evaluation due to the
rich data required to inform them.
18
Figure 2.2 Conceptual Markov model of disease progression. Theoretical “patients” can
remain in their current health state, experience disease progression up to severe
disease, or die.
In this toy example if all patients (say N=10,000) are born with mild disease, then the
initial conditions vector is:
𝑿 𝟎 = [𝑋 0,𝑚𝑖𝑙𝑑 𝑋 0,𝑚𝑜𝑑
𝑋 0,𝑠𝑒𝑣 𝑋 0,𝑑𝑒𝑎𝑑 ]
= [10,000 0 0 0]
Where membership for the mild, moderate, severe, and death states are each
respective entry from left to right.
The transition or stochastic matrix is:
𝑷 = [
Mild → Mild Mild → Mod Mild → Sev Mild → Dead
Mod → Mild Mod → Mod Mod → Sev Mod → Dead
Sev → Mild Sev → Mod Sev → Sev Sev → Dead
Dead → Mild Dead → Mod Dead → Sev Dead → Dead
]
= [
0.75 0.20 0 0.05
0 0.60 0.30 0.10
0 0 0 0.15
0 0 0 1
]
We can then multiply these components to find the state membership at the end of the
first cycle:
19
𝑿 𝟏 = 𝑿 𝟎 𝑷 = [10,000 0 0 0] [
0.75 0.20 0 0.05
0 0.60 0.30 0.10
0 0 0 0.15
0 0 0 1
]
= [7,500 2,000 0 500]
This process can be repeated and rewards by state membership, such as costs, are
added up and discounted. The chain stops after a predetermined horizon, or often once
a large proportion (~99%) of patients are in the “dead” state.
Note that transition probability matrix entries will be limited by model structure.
For example, since our specification does not allow for a jump from mild to severe
disease within one cycle, the model probability is zero. This emphasizes the need for
appropriate specification of the health states and transitions. Model structure, among
other inputs including transition probabilities and rewards should be informed by the
highest-quality evidence available. This is often data from systematic literature reviews,
meta-analysis studies, RCTs, and increasingly real-world evidence (RWE).
16
Markov
models have been used within SCD, particularly for decision analytic models of disease
screening and treatment comparisons.
17-23
Microsimulation and agent-based modeling
Our previous approaches are primarily deterministic, albeit with some probabilistic
elements (e.g. probabilistic sensitivity analysis in a decision tree or compartmental
model). Microsimulation and agent-based modeling (ABM) are methods which utilize
health states and movement among said states but allow for the tracking of a single
patient’s set of current and previous health states. In addition, ABM allows for the
20
interaction between patients which is necessary for the modeling of transmission such
as in infectious disease.
24
In models of health, microsimulation and ABM often use similar health states to
those in compartmental models. A microsimulation model with identical structure,
probabilities, and rewards will converge to the Markov model outcome with sufficient
iterations given a set tolerance.
25
For this reason, the added complexity of
microsimulation may only be preferable under certain conditions. One example is if
various subgroups (gender, body mass index, comorbidities, etc.) are of interest
xiv
. With
each additional subgroup, the number of health states needed to characterize the
problem in Markov model grows rapidly (also known as the state-explosion problem).
Other reasons to use microsimulation over a Markov model would be if previous state
membership matters for transitions (which would violate Markov property), or if tracking
individual outcomes is necessary. ABM is preferable to Markov models or
microsimulation if modeling transmission is necessary. Within sickle cell disease, few
subgroups are likely of interest and transmission is not possible. For these reasons, the
added complexity of microsimulation or ABM may be unnecessary.
Translating costs to value
Healthcare costs alone do not fully characterize societal gains and losses. In scenarios
where a cure is being evaluated, it is important to also value the potential gains in health
from new treatment. As many conditions associated with increased healthcare resource
xiv
This dissertation focuses primarily on understanding lifetime outcomes for patients with SCD
in aggregate. We do also consider outcomes by gender, which is still tractable in a cohort
Markov model.
21
utilization also affect survival, it is commonplace to attach a willingness to pay (WTP) to
a life year (LY) or quality-adjusted life year (QALY) saved. Interventions are then
evaluated and those with incremental cost-effectiveness ratios below the threshold are
deemed “cost-effective.” A challenge with this approach is choosing a WTP threshold
that is consistent with the preferences of the analyst’s chosen perspective. The Second
Panel on Cost-Effectiveness in Health and Medicine notes that WTP values for a QALY
in US studies vary dramatically between a commonplace $50,000 to roughly
$300,000.
16
This large variation has been described to be a result of the many different
assumptions, methods, and justifications used
in estimating the value of life, often
calculated in the form of the value of a statistical life (VSL).
Hirth et al. searched the literature to find studies which employed different
methods
xv
for estimating the value of life.
26
They then converted these values to a WTP
per QALY in 1997 US dollars using life expectancies, age-specific QALY weights, and
discounting. The authors found that the median QALY value varied dramatically based
on study method: human capital ($24,777), revealed preference-job risk ($428,286),
revealed preference-safety ($93,402), and contingent-valuation ($161,305). These
values are substantially higher
xvi
when converted to 2019 US dollars using the medical
component of the consumer price index (CPI). In another study, Braithwaite et al.
estimated lower and upper bounds for the societal value of a QALY in the US.
27
They
argue that the incremental cost-effectiveness of “modern” health care advances serves
xv
Hirth et al. identified 42 value of life estimates which they classified by method: human capital,
contingent valuation, revealed preference (job risk), or revealed preference (non-occupational
safety).
xvi
In 2019 USD: human capital ($52,628), revealed preference-job risk ($909,702), revealed
preference-safety ($198,391), and contingent-valuation ($342,620).
22
as a lower bound since most Americans favor expanding the share of GDP spent on
health care. Their upper bound is determined through revealed preference for health
insurance among nonelderly adults without employer- or government-subsidized
insurance. Given most of these adults did not elect to purchase the coverage, they use
the incremental cost-effectiveness for unsubsidized insurance as an upper bound. From
this, the authors find that the WTP per QALY can be reasonably inferred to be between
$109,000 and $297,000 in 2003 US dollars. This range is $182,814 to $498,126 when
inflated to 2019 US dollars using the medical CPI.
These studies, among others, demonstrate that while it is possible to estimate a
WTP/QALY using sophisticated methods, the results are highly contingent on the
methods and assumptions employed. In practical application, the WTP for a QALY in
studies from the US healthcare sector or societal perspectives often follow convention
or rules of thumb. The World Health Organization (WHO) has endorsed the use of three
times GDP per capita or roughly $150,000 in developed nations.
16,28
This can be
multiplied by the quantity of quality-adjusted years saved and added to the incremental
lifetime cost of chronic management (less any costs of the new therapy) to obtain the
total financial impact, usually as a present value, of curing a patient of their condition. In
the healthcare economic evaluation literature, this is commonly expressed as a net
monetary benefit (NMB):
𝑁𝑀𝐵 = 𝑊𝑇𝑃 ∗ 𝑄𝐴𝐿𝑌 − 𝐶 (1)
23
In equation (1) C is total incremental monetary cost, inclusive of any cost savings from a
new therapy. Any formal evaluation of a healthcare intervention for the treatment or cure
of a condition should factor in both net costs of treatment and the value of potential
health gains to society. In the case of a cure for sickle cell disease, potential health
gains are estimated to be large.
5
This is due to the significantly decreased life-
expectancy and quality-adjusted life expectancy for individuals born with the disease
relative to comparable individuals without the disease. It is not yet commonplace to
factor in equity concerns when calculating QALYs.
29
However, if a healthcare system or
society values equity, evaluating a treatment which is expected to improve distribution
of health against a higher cost-effectiveness threshold carries a similar effect. Given a
cure for SCD is likely to benefit a disproportionately underserved community in the US,
we feel this justifies a WTP threshold of $150,000 per QALY when evaluating a cure for
the condition.
Discussion and Limitations
We find a lack of published studies which study direct or indirect healthcare costs for
patients with SCD in the United States. Three studies provide two estimates for direct
costs and one for indirect costs (lost wages) over a lifetime. Methodologies including
data sources, discounting procedures, and populations of interest vary substantially. No
studies identified direct lifetime costs for US patients in a commercially insured plan.
We also discussed the state of the literature on health research inputs, current
methods, and example data resources commonly employed in generating lifetime
economic burden estimates. We find that methods vary by disease area; however,
24
some, like Markov models, are practical in sickle cell disease. Conditions with less
centralized data and relatively small sample sizes, such as sickle cell disease, require
creative modeling and synthesis approaches using treatment and survival information
from various sources. These data can include published clinical trials, disease-specific
life tables, and commercial health insurance claims data. Lastly, we described the
various approaches currently used in determining the value to society that curative
therapies bring.
Our approach to review of the literature is not without limitations. We only
searched PubMed database which does not include the universe of all published
medical studies. We did not search other databases including the economics literature.
We only included peer-reviewed published studies. This is because we intended to
provide an overview of credible estimates that have undergone a peer-review process.
Lastly, we take a treatment-agnostic approach in identifying lifetime cost studies. There
are likely several studies which estimate costs from very specific interventions, specific
patient populations, or both over limited horizons. However, we intend to understand
costs for the population of patients with sickle cell disease in the United States
managed with current real-world clinical practice over the entire course of the disease.
Our overview of lifetime costing methods does not include all methods as we did not
intend to be comprehensive. Rather, we summarized methods commonly used in health
economic evaluation and briefly described their applicability to our disease area of
interest.
25
Conclusions
There is limited information in the published medical literature on the age-specific and
lifetime costs of chronic management of SCD. Available studies focus primarily on
Medicaid covered populations for direct costs. While Medicaid and the State Children’s
Health Insurance Program (SCHIP) insure a majority of patients with SCD in the United
States (56.2% by one estimate in Boulet et al.), these reported costs are likely not
representative of those borne by patients covered under commercial plans.
30
Additional
studies are needed to fully understand the direct and indirect economic burdens faced
by patients and providers in managing SCD by age and over a lifetime. Updating current
figures and incorporating costs for patients covered by commercial plans are necessary
to demonstrate the potential cost savings from the development of additional disease-
modifying treatments. We hope additional research which contributes to the
understanding of direct and indirect economic burden faced by patients with chronic
disease over a lifetime is forthcoming.
26
References
1. Hassell KL. Population estimates of sickle cell disease in the U.S. Am J Prev
Med. 2010;38(4 Suppl):S512-521.
2. Hall JB, Kress J, Schmidt GA. Principles of Critical Care, 4th edition. McGraw-Hill
Education; 2015.
3. Kauf TL, Coates TD, Huazhi L, Mody-Patel N, Hartzema AG. The cost of health
care for children and adults with sickle cell disease. Am J Hematol.
2009;84(6):323-327.
4. Ballas SK. The cost of health care for patients with sickle cell disease. Am J
Hematol. 2009;84(6):320-322.
5. Lubeck D, Agodoa I, Bhakta N, et al. Estimated Life Expectancy and Income of
Patients With Sickle Cell Disease Compared With Those Without Sickle Cell
Disease. JAMA Netw Open. 2019;2(11):e1915374.
6. Kemp R, Prasad V. Surrogate endpoints in oncology: when are they acceptable
for regulatory and clinical decisions, and are they currently overused? BMC Med.
2017;15(1):134-134.
7. Kaplan EL, Meier P. Nonparametric Estimation from Incomplete Observations.
Journal of the American Statistical Association. 1958;53(282):457-481.
8. Cox DR. Regression Models and Life-Tables. Journal of the Royal Statistical
Society Series B (Methodological). 1972;34(2):187-220.
9. Ramsey SD, Willke RJ, Glick H, et al. Cost-effectiveness analysis alongside
clinical trials II-An ISPOR Good Research Practices Task Force report. Value
Health. 2015;18(2):161-172.
27
10. Mariotto AB, Yabroff KR, Shao Y, Feuer EJ, Brown ML. Projections of the cost of
cancer care in the United States: 2010-2020. J Natl Cancer Inst.
2011;103(2):117-128.
11. Lin DY, Feuer EJ, Etzioni R, Wax Y. Estimating medical costs from incomplete
follow-up data. Biometrics. 1997;53(2):419-434.
12. Bang H, Tsiatis AA. Median regression with censored cost data. Biometrics.
2002;58(3):643-649.
13. Basu A, Manning WG. Estimating lifetime or episode-of-illness costs under
censoring. Health Econ. 2010;19(9):1010-1028.
14. Hwang JS, Hu TH, Lee LJ, Wang JD. Estimating lifetime medical costs from
censored claims data. Health Econ. 2017;26(12):e332-e344.
15. Sonnenberg FA, Beck JR. Markov models in medical decision making: a practical
guide. Med Decis Making. 1993;13(4):322-338.
16. Neumann PJ, Sanders GD, Russell LB, Siegel JE, Ganiats TG. Cost
effectiveness in health and medicine. Second edition. ed. Oxford ; New York:
Oxford University Press; 2017.
17. Kacker S, Ness PM, Savage WJ, et al. Economic evaluation of a hypothetical
screening assay for alloimmunization risk among transfused patients with sickle
cell disease. Transfusion. 2014;54(8):2034-2044.
18. Kacker S, Ness PM, Savage WJ, et al. Cost-effectiveness of prospective red
blood cell antigen matching to prevent alloimmunization among sickle cell
patients. Transfusion. 2014;54(1):86-97.
28
19. Kuznik A, Habib AG, Munube D, Lamorde M. Newborn screening and
prophylactic interventions for sickle cell disease in 47 countries in sub-Saharan
Africa: a cost-effectiveness analysis. BMC Health Serv Res. 2016;16:304.
20. Mbulaiteye SM, Biggar RJ, Bakaki PM, et al. Human herpesvirus 8 infection and
transfusion history in children with sickle-cell disease in Uganda. J Natl Cancer
Inst. 2003;95(17):1330-1335.
21. Cherry MG, Greenhalgh J, Osipenko L, et al. The clinical effectiveness and cost-
effectiveness of primary stroke prevention in children with sickle cell disease: a
systematic review and economic evaluation. Health Technol Assess.
2012;16(43):1-129.
22. Nietert PJ, Abboud MR, Silverstein MD, Jackson SM. Bone marrow
transplantation versus periodic prophylactic blood transfusion in sickle cell
patients at high risk of ischemic stroke: a decision analysis. Blood.
2000;95(10):3057-3064.
23. Panepinto JA, Magid D, Rewers MJ, Lane PA. Universal versus targeted
screening of infants for sickle cell disease: a cost-effectiveness analysis. J
Pediatr. 2000;136(2):201-208.
24. Hamill L, Gilbert GN. Agent-based modelling in economics. Chichester, UK ;
Hoboken, NJ: Wiley; 2016.
25. Drummond M. Methods for the economic evaluation of health care programmes.
Fourth edition. ed. Oxford, United Kingdom ; New York, NY, USA: Oxford
University Press; 2015.
29
26. Hirth RA, Chernew ME, Miller E, Fendrick AM, Weissert WG. Willingness to pay
for a quality-adjusted life year: in search of a standard. Med Decis Making.
2000;20(3):332-342.
27. Braithwaite RS, Meltzer DO, King JT, Leslie D, Roberts MS. What Does the
Value of Modern Medicine Say about the $50,000 per Quality-Adjusted Life-Year
Decision Rule? Med Care. 2008;46(4):349-356.
28. Neumann PJ, Cohen JT, Weinstein MC. Updating cost-effectiveness--the curious
resilience of the $50,000-per-QALY threshold. N Engl J Med. 2014;371(9):796-
797.
29. Bobinac A, van Exel NJ, Rutten FF, Brouwer WB. Inquiry into the relationship
between equity weights and the value of the QALY. Value Health.
2012;15(8):1119-1126.
30. Boulet SL, Yanni EA, Creary MS, Olney RS. Health status and healthcare use in
a national sample of children with sickle cell disease. Am J Prev Med. 2010;38(4
Suppl):S528-535.
30
Appendix
Search string (PubMed database)
("anemia, sickle cell"[MeSH Terms] OR "sickle cell anemia"[All Fields] OR "sickle cell
disease"[All Fields] OR "sickle cell anaemia"[All Fields] OR "sickle cell disorders"[All
Fields] OR ("sickle"[All Fields] AND "cell"[All Fields] AND "anemia"[All Fields]) OR
("sickle"[All Fields] AND "cell"[All Fields] AND "disease"[All Fields]) OR ("sickle"[All
Fields] AND "cell"[All Fields] AND "anaemia"[All Fields]) OR ("sickle"[All Fields] AND
"cell"[All Fields] AND "disorders"[All Fields]))
AND ("Health Care Economics and Organizations"[MeSH Terms] OR
"economics"[MeSH Terms] OR "costs and cost analysis"[MeSH Terms] OR
"economics"[All Fields] OR "income"[All Fields] OR "cost"[All Fields] OR "costs"[All
Fields] OR "economic burden"[All Fields])
AND ("lifetime"[All Fields] OR "life"[All Fields] OR "lifespan"[All Fields] OR "course of
life"[All Fields] OR "by age"[All Fields] OR "age-specific"[All Fields] OR "age-related"[All
Fields] OR "age-associated"[All Fields])
AND ("2009"[PDat] : "2019"[PDat])
Date searched: December 4, 2019
Results: 157
31
Appendix Table 2.1 All studies identified in PubMed (PMC and MEDLINE)
First Author Journal Year PMID Title and
Abstract
Full Text Final
Inclusion
Lubeck, D JAMA Netw Open 2019 31730182 Y Y Y
Kauf, TL Am J Hematol 2009 19358302 Y Y Y
Ballas, SK Am J Hematol 2009 19415728 Y Y Y
Minniti, CP Am J Hematol 2017 28211097 Y Y N
Blinder, MA J Emerg Med 2015 25910824 Y Y N
Saenz, C Biol Blood Marrow Transplant 2015 25783634 Y Y N
Shah, N Health Qual Life Outcomes 2019 31619251 Y N N
Johnston, EE J Palliat Med 2019 31390292 Y N N
Onyeaka, HK Medicina (Kaunas) 2019 31319584 Y N N
Hankins, JS Pediatr Blood Cancer 2018 29797644 Y N N
Adam, SS Blood Adv 2017 29296845 Y N N
Tsitsikas, DA Transfus Apher Sci 2017 28602485 Y N N
Arnold, SD Biol Blood Marrow Transplant 2015 25615608 Y N N
Spackman, E Eur J Haematol 2014 24329965 Y N N
Kacker, S Transfusion 2014 23692415 Y N N
Blinder, MA Pediatr Blood Cancer 2013 23335275 Y N N
Cherry, MG Health Technol Assess 2012 23140544 Y N N
Cronin, RM Hematology 2019 30479187 N N N
Ojelabi, AO PLoS One 2019 31738762 N N N
Stallings, AM Br J Haematol 2019 31713857 N N N
Motta, I Mediterr J Hematol Infect Dis 2019 31700592 N N N
Oliveira, CDL Pediatr Hematol Oncol 2019 31490092 N N N
Ilyas, S Clin Chim Acta 2019 31678569 N N N
Smaldone, A BMC Pediatr 2019 31615480 N N N
Lippi, G Acta Haematol 2019 31550707 N N N
Gallo, AM J Adv Nurs 2019 31287187 N N N
Maier, AG Trends Parasitol 2019 30595467 N N N
Hildenbrand, AK Health Psychol 2019 31045421 N N N
Cooper, O Health Qual Life Outcomes 2019 31036017 N N N
Hammoudi, N Eur Heart J 2019 31005979 N N N
32
Pai, M BMC Med Res Methodol 2019 30922227 N N N
Salek, MS Qual Life Res 2019 30456713 N N N
Steele, C Am J Hematol 2019 30290004 N N N
Lee, L Public Health Rep 2019 31600481 N N N
Hockham, C Sci Rep 2018 30523337 N N N
Lobitz, S Br J Haematol 2018 30334577 N N N
Palermo, TM Contemp Clin Trials 2018 30316999 N N N
Connor, JL Medicine (Baltimore) 2018 30431560 N N N
Flynn, KE Health Qual Life Outcomes 2018 30340500 N N N
Cavazzana, M Hum Gene Ther 2018 30200783 N N N
Inusa, BPD F1000Res 2018 30228870 N N N
Delicou, S Mediterr J Hematol Infect Dis 2018 30210742 N N N
Sii-Felice, K Exp Hematol 2018 29807062 N N N
Anderson, LM Pediatr Blood Cancer 2018 29693797 N N N
Badawy, SM Biomed Res Int 2018 30105252 N N N
Streetly, A Arch Dis Child 2018 29104181 N N N
Aroke, D BMC Pregnancy Childbirth 2018 29929479 N N N
Sarri, G Health Qual Life Outcomes 2018 29784054 N N N
Foster, N Nurs Child Young People 2018 29697214 N N N
Chattoo, S Anthropol Med 2018 29533092 N N N
Cox, SE Lancet Haematol 2018 29548623 N N N
Kato, GJ Nat Rev Dis Primers 2018 29542687 N N N
Smart, LR Ann Hematol 2018 29147848 N N N
Creary, MS Soc Sci Med 2018 28477966 N N N
Davis, JM Transfusion 2018 29076161 N N N
Figueiredo, SV Rev Bras Enferm 2018 30517401 N N N
Hulihan, M MMWR Morb Mortal Wkly Rep 2017 29166365 N N N
Chevret, S Contemp Clin Trials 2017 28821470 N N N
Hill, LD PLoS One 2017 29088256 N N N
Torabian, K Biosensors (Basel) 2017 28930183 N N N
McGann, PT Blood Cells Mol Dis 2017 28844459 N N N
Young, JR BMJ Open 2017 28698351 N N N
Tluway, F Br J Haematol 2017 28295224 N N N
33
Compas, BE Am Psychol 2017 28481580 N N N
Alapan, Y Expert Rev Med Devices 2016 27785945 N N N
Gardner, K Blood 2016 27439910 N N N
Williams, LA South Med J 2016 27598370 N N N
Jung, J Sci Rep 2016 27546097 N N N
Yang, X Anal Chem 2016 27442043 N N N
Madderom, MJ BMC Hematol 2016 27493757 N N N
Evensen, CT Medicine (Baltimore) 2016 27583862 N N N
Kuznik, A BMC Health Serv Res 2016 27461265 N N N
Moore, A Nurs Stand 2016 27440342 N N N
Anders, DG Am J Prev Med 2016 27320463 N N N
Neunert, CE Am J Prev Med 2016 27320461 N N N
Raji, SO Int J Psychiatry Med 2016 28629283 N N N
Crighton, G Intern Med J 2016 27040044 N N N
Adams-Graves,
P
Expert Rev Hematol 2016 27098013 N N N
Maximo, C Hemoglobin 2016 27117565 N N N
Iughetti, L World J Clin Pediatr 2016 26862499 N N N
McGann, PT Am J Hematol 2016 26537622 N N N
Kunz, JB Ann Hematol 2016 26658910 N N N
Ajayi, TA J Palliat Med 2016 26840856 N N N
Piety, NZ PLoS One 2016 26735691 N N N
Chaturvedi, S Am J Hematol 2016 26547630 N N N
Master, S J La State Med Soc 2016 27797344 N N N
Arogundade, FA Clin Nephrol 2016 27509585 N N N
McGann, PT J Pediatr 2015 26477868 N N N
Stollon, NB J Pediatr Hematol Oncol 2015 26492583 N N N
Myint, KT Cochrane Database Syst Rev 2015 26451693 N N N
Williams, TN BMC Med 2015 26399886 N N N
Tewari, S Haematologica 2015 26341524 N N N
Key, NS Br J Haematol 2015 25754217 N N N
Fernandes, ML BMC Oral Health 2015 26085066 N N N
Bianchi, N Eur J Pharmacol 2015 25701725 N N N
34
Crosby, LE Adv Ther 2015 25832469 N N N
Duru, KC Evol Bioinform Online 2015 25788827 N N N
Nath, KA Nat Rev Nephrol 2015 25668001 N N N
Naghavi, M et al. Lancet 2015 25530442 N N N
Chakravorty, S Arch Dis Child 2015 25239949 N N N
Issom, DZ Stud Health Technol Inform 2015 26262200 N N N
Smith, WR J Opioid Manag 2015 25985809 N N N
O'Connor, S Clin J Oncol Nurs 2014 25427701 N N N
Wonkam, A J Genet Couns 2014 24557660 N N N
Serjeant, G Nature 2014 25390144 N N N
Parker, CS Am J Prev Med 2014 25245797 N N N
Yawn, BP JAMA 2014 25203083 N N N
Khoriaty, E PLoS One 2014 25180595 N N N
Wonkam, A OMICS 2014 24754796 N N N
Badaki-Makun, O Pediatr Blood Cancer 2014 24443249 N N N
Thakur, TJ Gene Regul Syst Bio 2014 24932102 N N N
Diallo, DA Curr Opin Hematol 2014 24613937 N N N
Imber-Black, E J Fam Nurs 2014 24707033 N N N
Anderson, N Popul Health Manag 2014 23965046 N N N
Nwenyi, E Int J Health Care Qual Assur 2014 25115050 N N N
Robinson, MR J Pediatr Hematol Oncol 2014 24136028 N N N
Claster, S Pediatr Blood Cancer 2013 24039121 N N N
Bayramoğlu, T Turk J Haematol 2013 24385828 N N N
Xie, LH J Health Care Poor Underserved 2013 24241268 N N N
Adams, RM Proc Natl Acad Sci U S A 2013 24019482 N N N
Zempsky, WT J Pain 2013 23701707 N N N
Ukachi, N Pediatrics 2013 23958767 N N N
Sheth, S Br J Haematol 2013 23772687 N N N
Amendah, DD PLoS One 2013 23593408 N N N
Ladis, V Eur J Haematol 2013 23331260 N N N
Morris, ZS BMC Health Serv Res 2013 23497301 N N N
Barakat, LP Contemp Clin Trials 2013 23149214 N N N
35
Cheng, C Conf Proc IEEE Eng Med Biol
Soc
2013 24110179 N N N
Pereira, SA Rev Bras Hematol Hemoter 2013 24255615 N N N
Dos Santos, JP Rev Bras Hematol Hemoter 2013 24106440 N N N
Raphael, JL Acad Pediatr 2013 24011745 N N N
Roth, M Biol Blood Marrow Transplant 2012 22659467 N N N
Voskaridou, E Ann Hematol 2012 22526366 N N N
Anie, KA BMJ Open 2012 22761289 N N N
Kutlar, A Am J Hematol 2012 22488107 N N N
Dickerson, AK Pediatr Blood Cancer 2012 21796763 N N N
Akodu, SO Pediatr Hematol Oncol 2012 22568799 N N N
Perrin, JM Future Child 2012 22550689 N N N
Koura, KG PLoS One 2012 23185556 N N N
Ohara, DG Rev Bras Fisioter 2012 22886311 N N N
Grosse, SD Am J Prev Med 2011 22099364 N N N
Porter, JB Int J Hematol 2011 21993873 N N N
Ddungu, H Br J Haematol 2011 21707576 N N N
Feliu, MH Hemoglobin 2011 22035043 N N N
Kyerewaa Edwin,
A
Pan Afr Med J 2011 22187596 N N N
Amr, MA Pan Afr Med J 2011 22121419 N N N
Ebakisse-
Badassou, E
Med Trop (Mars) 2010 21520648 N N N
Hijmans, CT Health Qual Life Outcomes 2010 20977722 N N N
Grigorescu, V Am J Prev Med 2010 20331953 N N N
Buchanan, G Biol Blood Marrow Transplant 2010 19819341 N N N
Greco, N Methods Mol Biol 2010 20680811 N N N
Krediet, CT Ned Tijdschr Geneeskd 2010 20178659 N N N
Treadwell, MJ J Dev Behav Pediatr 2010 20585265 N N N
Asnani, MR Psychol Health Med 2009 19844839 N N N
Imran, F Expert Rev Pharmacoecon
Outcomes Res
2009 19670988 N N N
Panepinto, JA Qual Life Res 2009 18989755 N N N
36
McLeod, C Health Technol Assess 2009 19068191 N N N
Wright, K Br J Nurs 2009 19223799 N N N
Abbreviations: PMID, PubMed Identifier. Y, yes; N, no
37
Chapter 3
Pediatric Inpatient Costs of Select Orphan
Conditions with Potential Gene Therapy
Launches: A Cross-Sectional Analysis of 2016
xvii
Salcedo J
a,b
, Bulovic J
b
, Trieu S
c
, Shinkle R
d
a
Department of Pharmaceutical and Health Economics, School of
Pharmacy, University of Southern California, Los Angeles, CA, USA
b
Center for Biomedical Innovation, Massachusetts Institute of Technology,
Cambridge, MA, USA
c
MedImpact Healthcare Systems, Inc., San Diego, CA, USA
d
REGENXBIO Inc., Rockville, MD, USA
xvii
This chapter is based on co-authored work by Salcedo J, Bulovic J, Trieu S, and Shinkle R.
The conceptualization, data analysis, modeling, and writing of this chapter were primarily my
work.
An abstract of a preliminary version of this work was presented at the Academy of Managed
Care Pharmacy (AMCP) Nexus 2019 meeting. Full citation: Salcedo J, Bulovic J, Shinkle R,
Trieu S. Pediatric Inpatient Costs of Select Orphan Conditions with Potential Gene Therapy
Launches: A Cross-Sectional Analysis of 2016. Journal of managed care pharmacy: JMCP.
2019;25:S89.
An abridged version of this work was shared as a research brief by the MIT Center for
Biomedical Innovation’s New Drug Development Paradigms initiative. Link:
https://newdigs.mit.edu/focus-research-brief-2019f205v038-pediatric-gene-therapy-launches
38
Abstract
Importance: Recent advances have led to cell and gene therapies that offer durable
treatment for previously incurable conditions. Understanding the costs of inpatient
hospitalizations in target disease areas for cures is necessary for estimating financial
impact by durable therapies to the United States healthcare system.
Objective: To estimate the direct costs of pediatric inpatient hospitalizations within
disease areas with potential gene therapy cures.
Design: Cross-sectional analysis between January 1, 2016 and December 31, 2016.
Setting: Healthcare and Cost Utilization Project (HCUP) Kids’ Inpatient Database (KID)
of 2016.
Participants: Pediatric patients 20 years or younger discharged from KID participating
states.
Exposure(s): Discharges with at least one International Classification of Diseases,
Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis of: Adenosine deaminase
deficiency with severe combined immunodeficiency (ADA-SCID), Beta thalassemia
(BT), Cystic fibrosis (CF), Gaucher disease (GD), Hemophilia A or B (HAA/HAB),
Mucopolysaccharidosis III (MPS-III), Ornithine transcarbamylase deficiency (OTCD),
Pompe disease (PD), Sickle cell disease (SCD), or Von Gierke disease (VG).
Main Outcome(s) and Measure(s): Our primary outcomes of interest are mean costs
for inpatient stays within and across all disease areas of interest. We also calculated
nationally representative healthcare expenditures within and across all disease areas of
interest.
39
Results: We identified an estimated 55,188 stays nationally in 2016. These estimates
ranged from 34 stays for ADA-SCID to 37,000 for SCD. Average hospitalization costs
were lowest for SCD: $9,141, 95% confidence interval [CI] ($8,538, 9,744) and highest
for ADA-SCID: $91,636, 95% CI ($26,072, 157,199). In aggregate, CF-related stays
accounted for 41.5% of total costs: ($414.8M, 95% CI [$333.5M, 496.0M]). Total
healthcare expenditures for inpatient stays in all selected disease areas were $998.8
million in 2016.
Conclusions and Relevance: Inpatient costs vary significantly across orphan
diseases. Overall financial impact to the US healthcare system in 2016 was substantial
($998.8M). Curative therapies provided to pediatric populations have the potential to
avoid significant resource utilization across disease areas. Additional research is
necessary to characterize complete direct cost burdens (outpatient, pharmacy,
professional, etc.) and indirect cost burdens to patients and caregivers.
Introduction
Gene therapy cures for hereditary genetic conditions are likely to reach the United
States market in the coming decades. Several disease areas with potential cures are
orphan conditions, conditions that affect fewer than 200,000 patients in the US. To
understand the current cost burden associated with managing orphan conditions, we
analyzed 55,188 pediatric inpatient stays across disease areas including select
hemoglobinopathies, bleeding disorders, lysosomal storage diseases, cystic fibrosis,
and other enzymatic deficiencies. We find national inpatient cost estimates of nearly $1
40
billion in 2016, attributable to our eleven orphan diseases of interest –despite
accounting for less than one percent of pediatric inpatient stays that year.
Background
Durable treatments for genetic diseases
Human gene therapy has advanced tremendously since the first human gene trial was
initiated in the 1990’s. Even in its infancy the therapy showed great potential with the
successful treatment of a patient with an inherited immunodeficiency. Although
promising and revolutionary, gene therapy is not without risk, as it was evidenced early
on with the often-referenced death of Mr. Jesse Gelsinger.
1
While medicine and
treatment regimens have improved life expectancy for numerous conditions, in the case
of hereditary genetic diseases there are only two curative treatments: hematopoietic cell
transplantation (HCT) or gene therapy. The prior, if an adequate donor is found, carries
complications before (conditioning regimen), during (rejection) and after (graft-vs-host
disease), all of which make gene therapy a superior option.
Other conditions for which curative therapy is not available, are managed with
frequent enzyme replacement therapy (ERT), or organ transplantation, which slow down
the disease progression or is associated with significant morbidity and mortality and
require immune suppressants, respectively.
2,3
Potential cures target a heterogeneous group of diseases
There are several gene therapy trials seeking United States (US) Food and Drug
Administration (FDA) approval for hereditary conditions, from which we have selected
41
representative diseases to illustrate the impact that the approval of these drugs will
have on the US healthcare system (Box 1). Our population of interest includes a diverse
group of diseases, from rapidly fatal without early intervention (i.e. gene therapy, HCT
or ERT), such as adenosine deaminase deficiency with severe combined
immunodeficiency (ADA-SCID), to sickle cell disease (SCD) in which the current
standard of care has increased life expectancy enough to become a chronic condition.
See Appendix Table 1 for a detailed list of select gene therapy trials for our conditions of
interest.
Box 3.1 Our orphan diseases of interest and etiology, in brief
• Sickle cell disease and beta thalassemia, caused by an abnormality in the beta
gene and beta protein ratio, respectively. This leads to early destruction of red
blood cells, vessel obstruction, among other symptoms.
• Hemophilia A and B, caused by a deficiency of a clotting factor. As a result, these
patients experience life-threatening bleeding.
• Pompe, Von Gierke, Gaucher, and Mucopolyssacharidosis-III (MPS-III), caused by
an enzyme deficiency that leads to the accumulation of glycogen (Pompe and Von
Gierke), glucocerebroside (Gaucher) and glucosaminoglycans (MPS-III) that results
in cellular dysfunction and clinical abnormalities.
• Adenosine Deaminase Deficiency, in most cases due to an enzyme deficiency that
results in the inhibition of DNA synthesis and inadequate/absent immune response
to infections.
• Ornithine Transcarbomylase Deficiency, caused by an enzymatic deficiency that
results in the accumulation of ammonia.
• Cystic fibrosis, caused by a mutation that leads to an abnormal cellular
transmembrane protein. Clinically, this leads to thicker mucous in the lungs,
pancreatic insufficiency, and frequent infections.
42
For the purposes of this study we have classified them into hemoglobinopathies (SCD
and beta thalassemia [BT]), bleeding disorders (hemophilia A [HAA] and B [HAB]),
lysosomal storage diseases (Pompe [PD], Von Gierke [VG], Gaucher [GD], and
Mucopolysaccharidosis Type III [MPS-III]), cystic fibrosis (CF) and other enzymatic
deficiency (ADA-SCID, ornithine transcarbamylase deficiency [OTCD]). In the US these
diseases are classified as orphan diseases, given that fewer than 200,000 people are
affected by each condition (Orphan Drug Act of 1983). See Table 3.1 for detailed
information on our disease areas of interest including current estimates for incidence
and age of onset.
43
Table 3.1 Epidemiology and notable costs for orphan diseases of interest
Condition ICD-10-CM Incidence Age of onset
*
Notable costs
ADA-SCID D81.3 1/200,000 births
4
Infancy,
neonatal
• Infections
• Bone Marrow Transplant (BMT)/ Hematopoietic
Stem Cell Transplant (HSCT)
• Enzyme replacement therapy (ERT)
• Other: Anti-infectives (antifungals, antibiotics,
antivirals), IVIG
BT D56.1 1/100,000 births
5
Infancy,
childhood
• BMT
• Blood transfusions
• Complications: Venous thrombosis, HIV, Hep. C
• Other: Chelation Therapy, osteoporosis
prevention, vitamin D
CF E84X 1/3,200 Caucasians,
1/10,000 Hispanics,
1/10,500 Native
Americans, 1/15,000
African Americans,
and 1/30,000 Asian
American births
6,7
All ages • Organ Transplant: Liver and Lung/Heart and Lung
• Upper respiratory and Pulmonary complications
• ERT
• Other: CFTR modulators, Pulmozyme
GD E75.22 1/63,000 births
8,9
All ages
• Surgical: joint replacement, splenectomy
• HSCT
• Blood transfusions
• ERT and substrate reduction
• Other: analgesics, calcium, vitamin D
HAA D66 1/5,000 male births
*
Infancy,
neonatal
• Severe bleeding episodes
• Surgical: total joint replacement
• Factor administration (acute and/or prophylactic)
• Other: desmopressin, inhibitor therapy and
screening
HAB D67 1/30,000 male births
*
Infancy,
neonatal
• See HAA
MPS-III E76.22 1.26/100,000 births
10
Childhood • Supportive care
• Other: hernia repair, orthopedic procedures
44
OTCD
†
E72.4
1/8,200 births
11
Neonatal, all
ages
• Hyperammonemic crisis and/or coma
• Organ Transplant: Liver transplant
• Hemodialysis
• Other: gastrostomy, ammonia scavenger therapy,
supplements, monitoring
PD E74.02 1/21,979 births
12
Antenatal,
neonatal,
childhood,
adolescent,
adult
• Ventilatory assistance
• Surgical: contractures, tracheostomy
• Infections
• ERT
• Complications: cardiac disease
• Other: monitoring, physical therapy, nutrition,
respiratory training (e.g. CPAP)
SCD D57.X sans
D57.3X
1/85,000 births
13
All ages • Vaso-occlusive crises
• Blood transfusions
• Surgical: splenectomy, HSCT
• Infections
• Complications: Acute Chest Syndrome, Stroke,
splenic sequestration
• Other: Hydroxyurea, antibiotics, analgesics,
chelation therapy
VG E74.01 1/100,000 births
14
Infancy,
neonatal
• Organ Transplant: Liver, kidney
• Surgical: percutaneous ethanol injections and
radiofrequency for hepatic adenomas
• Infections
• Other: nutritional therapy, monitoring
Abbreviations: ICD-10-CM, International Classification of Diseases, Tenth Revision, Clinical Modification; ADA-SCID,
adenosine deaminase deficiency with severe combined immunodeficiency; BT, beta thalassemia; CF, cystic fibrosis; GD,
Gaucher disease; HAA, hemophilia A; HAB, hemophilia B; MPS-III, mucopolysaccharidosis III; OTCD, ornithine
transcarbamylase deficiency; PD, Pompe disease; SCD, sickle cell disease; VG, Von Gierke disease.
*
Estimates obtained
from Orphanet.
†
Urea cycle disorders statistics.
45
We have selected these conditions because of our ability to identify the patient
population using International Classification of Diseases, 10th Revision, Clinical
Modification (ICD-10-CM) codes, the high number of inpatient admissions relative to
other orphan conditions, and the potential for approval of curative gene therapy
treatment in the US (see Appendix Table 3.1). These conditions have diverse
mechanisms of disease, age of onset, rate of progression, level of disability,
management, and complications – factors which contribute to cost-burden at different
stages of life and to different payers.
Healthcare costs associated with inpatient stays provide an estimate
for disease-related resource utilization in childhood
Inpatient stays are costly healthcare events that can be disproportionately expensive for
individuals with rare genetic diseases, given their complex needs. Such individuals may
be more likely to be hospitalized, or hospitalized for longer, for interventions that may
otherwise be carried out in an outpatient setting. Such conditions may in future be
treated by one-time curative therapies so that these costs may be avoided, along with
costs associated to the intensive comprehensive care upon diagnosis and the
subsequent outpatient visits and management.
Our selection of orphan diseases includes conditions managed differently in an
inpatient and outpatient basis, however, all these patients experience frequent
hospitalizations and increased morbidity and mortality. SCD and BT are two conditions
in which affected individuals are at higher risk of stroke, thromboembolism, and heart
failure – complications that are often managed in an inpatient basis, and contribute
46
substantially to the disease cost. Similarly, severe bleeds in HAA/HAB and infections in
children with SCD are common complications treated on an inpatient basis. Other
diseases may require more aggressive interventions such as HCT and organ transplant,
like in CF, OTCD, ADA-SCID, SCD, and BT. There is limited research on the inpatient
cost burdens of these orphan conditions in the United States. See Table 3.1 for more
detailed information on notable costs.
Relevance of inpatient stays to evaluate cost of treating disease
relative to other cost drivers, by disease
Adenosine Deaminase Deficiency with Severe Combined Immunodeficiency
15
ADA-SCID treatments address manifestations and underlying disease, both types of
which entail hospitalizations. Many children with ADA-SCID are hospitalized for
infections, with the initial hospitalization often due to pneumonitis. The preferred
definitive treatment to limit recurrent infections is BMT/HSCT which typically requires a
lengthy inpatient stay. Other cost drivers include pharmaceutical therapies [ERT,
antifungals, antibiotics, antivirals, IVIG]. Anti-infectives are required prior to initiation of
ERT or HSCT with reduced requirements afterwards. ERT is lifelong therapy for those
who cannot or choose not to undergo HSCT.
Beta Thalassemia
16
Hospitalizations may be required for individuals with thalassemia major who opt for BMT
to eliminate transfusions and chelation therapy. Thalassemia intermedia treatment is
symptomatic and typically doesn’t require hospitalizations. Hospitalizations may also
47
result other complications of disease or regular transfusions, such as venous
thrombosis, HIV infection, HCC. Non-inpatient cost drivers are numerous including
transfusions, regular monitoring for iron overload and other effects of treatment and
disease, chelation and osteoporosis therapy, and vitamin D.
Cystic Fibrosis
17
Liver and lung or heart lung transplant may be indicated. Frequent hospitalizations for
pulmonary disease, including surgical intervention for nasal/sinus symptoms. Additional
contributors to cost include various evaluations to assess and monitor overall disease
status, including respiratory, exocrine pancreatic insufficiency. In addition, individuals
with CF are treated with various respiratory medications, nutritional therapy, and insulin
if needed.
Gaucher Disease
18
Inpatient care may be provided to individuals with Gaucher disease for joint
replacement surgery, HSCT (rarely), splenectomy, and blood product transfusions. ERT
and substrate reduction therapy are most commonly used to treat disease. Other cost
drivers may include analgesics for bone pain and supplemental treatments such as
calcium and vitamin D, as well as regular surveillance of disease manifestations.
Hemophilia A
19
and B
20
Hemophilia A or B patients with severe, painful, bleeding that cannot be controlled tend
to present to the hospital. These patients are at higher risk of intracranial hemorrhage
48
and are often admitted for full evaluation and monitoring. Many, except minor, surgeries
can be performed with adequate factor replacement, and may include orthopedic
procedures for disease manifestations as well as typical elective surgeries. Significant
cost due to acute or prophylactic factor VIII/IX administration, with frequency depending
on severity of disease. Inhibitor therapy and regular screening for inhibitors and overall
disease status also contribute to cost of care.
Mucopolysaccharidosis Type III (Sanfilippo)
21
As there is no disease modifying treatment (ERT, HSCT), patients with Sanfilippo
mainly receive supportive care, typically multidisciplinary given range of disease
manifestations which are primarily cognitive defects and neurological dysfunction as
well as somatic symptoms. Some care may require hospitalization, such as for hernia
and orthopedic procedures.
Ornithine Transcarbamylase Deficiency
22
Absent newborn screening, OTCD is often diagnosed when the affected person
experiences a crisis. For example, infantile onset disease emerges shortly after birth
through onset of hyperammonemic coma, requiring hospitalization to receive
hemodialysis and other treatment. Severe cases are ultimately treated in many cases
with liver transplantation, also requiring hospitalization. Interventions to aid nutrition
include gastrostomy tube placement may require hospitalization. Other key cost drivers
include ammonia scavenger therapy, nutritional supplementation, and regular
monitoring, e.g. of liver function.
49
Pompe Disease
23
While anesthesia and intubation are generally limited to cardiac and respiratory
manifestations of disease, some patients may require surgery for contractures,
tracheostomy and procedures to establish venous access—all of which require
hospitalization. Pneumonia and cardiac disease may also contribution to
hospitalizations of individuals with Pompe disease. Outpatient care includes significant
individualized care and monitoring for cardiomyopathy, physical therapy, nutrition
feeding support, respiratory limitations (training, CPAP). Many patients take ERT.
Sickle Cell Disease
24
Severe pain episodes require hospitalization. Complications can include Acute Chest
Syndrome and stroke, which are managed in an inpatient basis. Individuals with SCD
often undergo splenectomy and less often undergo HSCT, both of which require
hospitalization. Lifelong, multimodal comprehensive care is required, resulting in non-
inpatient costs including medications including hydroxyurea, antibiotics, analgesics, and
transfusions; therapy to manage symptoms, and monitoring and treatment of
complications, etc.
Von Gierke Disease (Glycogen Storage Disease Type I)
25
Hospitalizations related to VG include: liver transplant in individuals refractory to
medical treatment; kidney transplantation for end-stage renal disease (ESRD); surgery
for percutaneous ethanol injections and radiofrequency ablation for hepatic adenomas.
Core therapy for VG is medical nutritional therapy. Other cost drivers include monitoring
50
of disease status, including blood tests, MRI and ultrasound exams, etc. Many
medications may be required to address disease, blood pressure, recurrent infections,
lipids, gout, etc. associated with VG.
Methods
We used the 2016 edition of the Kids’ Inpatient Database (KID), Healthcare Cost and
Utilization Project (HCUP), Agency for Healthcare Research and Quality to analyze
stays for select orphan diseases with potentially durable treatments or cures in the US
FDA pipeline.
26
The HCUP KID database is sampled from 4,200 U.S. community
hospitals in the US and contains information on patients admitted under the age of 21. It
is the largest publicly-available all-payer pediatric inpatient care database in the US.
HCUP provide discharge weights per stay which allow for construction of nationally
representative estimates. See Appendix Table 3.1 for select gene therapies and
corresponding clinical trials.
To classify inpatient stays as disease-related, we utilized ICD-10-CM diagnosis
codes. These are provided by the Centers for Medicare and Medicaid Services (CMS)
and the National Center for Health Statistics (NCHS) for classification of morbidities in
the United States. We classified stays into mutually exclusive categories by orphan
disease based on ICD-10-CM diagnosis present. In cases where a stay had multiple
orphan disease diagnoses (<1% of cases), we used the primary orphan diagnosis.
Cases with missing total charges (<3% of cases) were excluded from our analysis. We
converted charges to estimated costs using HCUP hospital-level cost-to-charge (CCR)
ratios. Costs reflect the actual expenses incurred by hospitals in providing care, not
51
necessarily the amount paid by payers. We performed all statistical procedures using
Stata/SE software version 15.1 (StataCorp, College Station, TX). Our study solely
utilized a limited data set (HCUP KID) and thus institutional review board (IRB) review
was not required.
Results
We find significant inpatient hospitalization costs associated with our orphan diseases of
interest in 2016. Across 55,188 representative stays, we estimated total financial cost to
US payers of $998.8M.
By disease area
Average hospitalization costs varied widely across disease areas. Costs were highest
for ADA-SCID $91,636, (95% CI = $26,072 to 157,199) and lowest for SCD $9,141,
(95% CI = $26,072 to 157,199). Average cost per hospitalization across all disease
areas was $18,098, 95% confidence interval (CI) = $16,394 to 19,801. This contrasted
with the average hospitalization cost for other pediatric stays: $7,700, 95% CI = $7,161
to 8,240 (p<0.05, for all comparisons). Despite being only 24.0% of hospitalizations,
stays for CF accounted for $414.8M (41.5%) of the total cost burden among these
orphan diseases.
By payer
Medicaid and private plans including HMO were the primary expected payer for over
92% of orphan hospitalizations. Among these, average hospitalization cost for Medicaid
52
covered stays was $16,034, 95% CI ($14,351 to 17,718) relative to the private payer
stay average of $22,021, 95% CI ($19,745 to 24,298) (p<0.0001). In aggregate,
Medicaid and private payers were primary payers for stays costing $909.0 million
(91.0%) in 2016.
Length of stay
In addition to costs, we studied length of inpatient stay (LOS) in days, across disease
areas and by payer. Mean LOS ranged from 4.12 days, 95% CI (3.99 to 4.25) for SCD
to 20.89 days, 95% CI (7.73 to 34.06) for ADA-SCID. Mean LOS was 3.99 days, 95% CI
(3.90 to 4.08) for other stays (p<0.05 for all comparisons except SCD, p=0.0943). As
expected, LOS was positively correlated with total inpatient cost within orphan disease
stays (r = 0.61, p<0.0001). See Table 3.2 for detailed breakdown of total and average
hospitalization costs by disease area and payer.
Figure 3.1 Mean inpatient stay costs (95% CI), by disease area (2016)
53
Table 3.2 Costs and inpatient resource utilization by orphan disease and payer, 2016
Category
National Stays
*
No. (%)
Total Cost (95% CI) Mean Cost (95% CI)
Mean LOS (95%
CI)
Orphan condition
ADA-SCID 34 (0.1%) $3.2M (0.5, 5.8) $91636 (26072, 157199) 20.9 (7.7, 34.1)
BT 614 (1.1%) $19.9M (9, 30.8) $32487 (21075, 43898) 7.5 (5.8, 9.3)
CF 13239 (24%) $414.8M (333.5, 496) $31329 (28466, 34193) 10.2 (9.8, 10.5)
GD 147 (0.3%) $8.8M (2.9, 14.7) $59638 (26046, 93230) 11.1 (8.1, 14.1)
HAA 2181 (4%) $120.9M (87.4, 154.5) $55462 (43087, 67836) 4.9 (4.5, 5.4)
HAB 442 (0.8%) $26.1M (13.9, 38.4) $59144 (38191, 80098) 6.7 (5, 8.3)
MPS-III 117 (0.2%) $3.7M (1.6, 5.9) $31771 (16602, 46941) 6.7 (4.9, 8.5)
OTCD 414 (0.8%) $34.8M (20.4, 49.2) $84036 (56669, 111403) 9.5 (7.4, 11.5)
PD 249 (0.5%) $13.4M (7, 19.7) $53697 (33936, 73459) 12.2 (8.6, 15.8)
SCD 37000 (67%) $338.2M (281.8, 394.6) $9141 (8538, 9744) 4.1 (4, 4.3)
VG 752 (1.4%) $15M (9.5, 20.5) $19928 (13208, 26647) 7.3 (5.6, 9)
Payer type
†
Medicare 485 (0.9%) $5.6M (3.9, 7.3) $11506 (9572, 13439) 5.6 (5, 6.3)
Medicaid 35358 (64.1%) $566.9M (464.8, 669.1) $16034 (14351, 17718) 5.6 (5.3, 5.8)
Private 15533 (28.1%) $342.1M (274.8, 409.3) $22021 (19745, 24298) 6.3 (6, 6.6)
Self-pay 1143 (2.1%) $19.8M (11.1, 28.5) $17311 (12135, 22487) 5.2 (4.5, 5.9)
Other/Unknown 2668 (4.8%) $64.4M (37.7, 91.1) $24134 (17666, 30602) 6.5 (5.9, 7)
Total 55188 $998.8M (828.8, 1168.7) $18098 (16394, 19801) 5.8 (5.6, 6)
Abbreviations: CI, confidence interval; LOS, length of stay (days); M, million; ADA-SCID, adenosine deaminase
deficiency; BT, beta thalassemia; CF, cystic fibrosis; GD, Gaucher disease; HAA, hemophilia A; HAB, hemophilia B; MPS-
III, mucopolysaccharidosis III; OTCD, ornithine transcarbamylase deficiency; PD, Pompe disease; SCD, sickle cell
disease; VG, Von Gierke disease.
*
Nationally representative estimates obtained using hospital-level discharge weights.
†
Primary expected payer.
54
Discussion
Our analysis provides an estimate of inpatient costs that may be avoidable if durable
therapies are developed and provided to patients within the first two decades of life. We
estimate that nearly $1 billion in healthcare costs are attributable to pediatric
hospitalizations for ADA-SCID, BT, CF, GD, HAA, HAB, MPS-III, OTCD, PD, SCD, or
VG.
We acknowledge limitations within our approach. Our unit of analysis is hospital
stay and hence it is not possible to measure resource utilization by patient. Additionally,
this limits the ability to distinguish disease-related admissions to those due to pre-
existing comorbidities. Our primary outcome of interest is healthcare cost which is a
measure of actual expenses incurred including wages, supplies, and utilities. However,
costs are estimated from charges and are not necessarily the amount each payer is
ultimately responsible for. Like other analyses which utilize ICD-10-CM diagnosis codes,
disease identification in administrative data is imperfect and validation studies are
necessary to confirm diagnostic accuracy across disease areas.
The focus of our study is to estimate overall burden to the US healthcare system
attributable to our orphan diseases of interest. By disease area, SCD accounted for
most stays (67.0%) and $338.2M (33.9%) of total hospitalization costs. Despite only
accounting for 24.0% of hospitalizations, CF-related stays cost $414.8M (41.5%) of the
total cost burden across all orphan disease areas of interest in 2016. This discrepancy
is likely attributable to hospitalizations for pulmonary disease and costly lung
transplants. Approximately 16.4% of lung transplants globally are performed in patients
with CF.
27
55
Breakdowns by primary expected payer showed significant burden to both
Medicaid and private plans including HMO. Despite being the primary expected payer in
64.1% of hospitalizations, Medicaid only bore $566.9M (56.8%) of the total cost burden
in 2016. This was in contrast to private payers which accounted for 28.1% of stays but
faced costs of $342.1M (34.3%). Other payers overall, including Medicare, were primary
expected payers for <10% of cases and <10% of total costs ($89.8M).
Conclusions
We have studied inpatient costs in orphan conditions with differing patient populations,
pathologies, and treatment pathways. Within the context of sickle cell disease, the
primary focus of this dissertation, pediatric inpatient costs provide an insight into total
cost of care. A recent study of health care spending in 154 conditions, Dieleman et al.
(2020), found that all-payer costs were $10.7 billion for hemoglobinopathies and
hemolytic anemias in 2016.
28
The authors reported that 18.4% of costs were associated
with inpatient care within this group. While the fraction of overall health care spending
that is associated with inpatient stays may vary by age group, using this value as
benchmark suggests total health care spending for pediatric patients with SCD may be
roughly $1.8 billion per year. While cures for SCD or other genetic orphan diseases are
likely to bring high upfront costs (e.g. Zolgensma for spinal muscular atrophy, priced at
$2.1 million), there is a clear potential within certain disease areas to offset the direct
costs associated with chronic management.
29
Curative therapies provided to pediatric populations have the potential to avoid
significant resource utilization across disease areas. Cell and gene therapy cures or
56
other durable treatments with one-time administrations may result in significantly
decreased disease-related hospitalizations among children with orphan conditions in the
US. Depending on treatment uptake and age of administration, avoidable inpatient
healthcare costs in these disease areas may be as high as $1 billion USD. Broadly,
additional research is necessary to characterize complete direct cost burdens
(outpatient, pharmacy, professional, etc.) and indirect cost burdens to patients and
caregivers. Further research at the specific disease-level is necessary to characterize
the complete economic net impacts of durable cures in various orphan diseases.
57
References
1. Wilson JM. Lessons learned from the gene therapy trial for ornithine
transcarbamylase deficiency. Mol Genet Metab. 2009;96(4):151-157.
2. Kanters TA, Hoogenboom-Plug I, Rutten-Van Molken MP, Redekop WK, van der
Ploeg AT, Hakkaart L. Cost-effectiveness of enzyme replacement therapy with
alglucosidase alfa in classic-infantile patients with Pompe disease. Orphanet J
Rare Dis. 2014;9:75.
3. Busuttil AA, Goss JA, Seu P, et al. The role of orthotopic liver transplantation in
the treatment of ornithine transcarbamylase deficiency. Liver Transpl Surg.
1998;4(5):350-354.
4. Sanchez JJ, Monaghan G, Borsting C, Norbury G, Morling N, Gaspar HB. Carrier
frequency of a nonsense mutation in the adenosine deaminase (ADA) gene
implies a high incidence of ADA-deficient severe combined immunodeficiency
(SCID) in Somalia and a single, common haplotype indicates common ancestry.
Ann Hum Genet. 2007;71(Pt 3):336-347.
5. Galanello R, Origa R. Beta-thalassemia. Orphanet J Rare Dis. 2010;5:11.
6. Hamosh A, FitzSimmons SC, Macek M, Jr., Knowles MR, Rosenstein BJ, Cutting
GR. Comparison of the clinical manifestations of cystic fibrosis in black and white
patients. J Pediatr. 1998;132(2):255-259.
7. O'Sullivan BP, Freedman SD. Cystic fibrosis. Lancet. 2009;373(9678):1891-
1904.
8. Poorthuis BJ, Wevers RA, Kleijer WJ, et al. The frequency of lysosomal storage
diseases in The Netherlands. Hum Genet. 1999;105(1-2):151-156.
58
9. Dionisi-Vici C, Rizzo C, Burlina AB, et al. Inborn errors of metabolism in the
Italian pediatric population: a national retrospective survey. J Pediatr.
2002;140(3):321-327.
10. Zelei T, Csetneki K, Voko Z, Siffel C. Epidemiology of Sanfilippo syndrome:
results of a systematic literature review. Orphanet J Rare Dis. 2018;13(1):53.
11. Brusilow SW, Maestri NE. Urea cycle disorders: diagnosis, pathophysiology, and
therapy. Adv Pediatr. 1996;43:127-170.
12. Burton BK, Charrow J, Hoganson GE, et al. Newborn Screening for Lysosomal
Storage Disorders in Illinois: The Initial 15-Month Experience. J Pediatr.
2017;190:130-135.
13. Hassell KL. Population estimates of sickle cell disease in the U.S. Am J Prev
Med. 2010;38(4 Suppl):S512-521.
14. Melis D, Fulceri R, Parenti G, et al. Genotype/phenotype correlation in glycogen
storage disease type 1b: a multicentre study and review of the literature. Eur J
Pediatr. 2005;164(8):501-508.
15. Hershfield M. Adenosine Deaminase Deficiency. In: Adam MP, Ardinger HH,
Pagon RA, et al., eds. GeneReviews((R)). Seattle (WA): University of
Washington, Seattle; 1993-2019.
16. Origa R. Beta-Thalassemia. In: Adam MP, Ardinger HH, Pagon RA, et al., eds.
GeneReviews((R)). Seattle (WA): University of Washington, Seattle; 1993-2019.
17. Ong T, Marshall SG, Karczeski BA, Sternen DL, Cheng E, Cutting GR. Cystic
Fibrosis and Congenital Absence of the Vas Deferens. In: Adam MP, Ardinger
59
HH, Pagon RA, et al., eds. GeneReviews((R)). Seattle (WA): University of
Washington, Seattle; 1993-2019.
18. Pastores GM, Hughes DA. Gaucher Disease. In: Adam MP, Ardinger HH, Pagon
RA, et al., eds. GeneReviews((R)). Seattle (WA): University of Washington,
Seattle; 1993-2019.
19. Konkle BA, Huston H, Nakaya Fletcher S. Hemophilia A. In: Adam MP, Ardinger
HH, Pagon RA, et al., eds. GeneReviews((R)). Seattle (WA): University of
Washington, Seattle; 1993-2019.
20. Konkle BA, Huston H, Nakaya Fletcher S. Hemophilia B. In: Adam MP, Ardinger
HH, Pagon RA, et al., eds. GeneReviews((R)). Seattle (WA): University of
Washington, Seattle; 1993-2019.
21. Tomatsu S. Mucopolysaccharidoses update (2 volume set). New York: Nova
Medicine & Health; 2018.
22. Lichter-Konecki U, Caldovic L, Morizono H, Simpson K. Ornithine
Transcarbamylase Deficiency. In: Adam MP, Ardinger HH, Pagon RA, et al., eds.
GeneReviews((R)). Seattle (WA): University of Washington, Seattle; 1993-2019.
23. Leslie N, Bailey L. Pompe Disease. In: Adam MP, Ardinger HH, Pagon RA, et al.,
eds. GeneReviews((R)). Seattle (WA): University of Washington, Seattle; 1993-
2019.
24. Bender MA. Sickle Cell Disease. In: Adam MP, Ardinger HH, Pagon RA, et al.,
eds. GeneReviews((R)). Seattle (WA): University of Washington, Seattle; 1993-
2019.
60
25. Bali DS, Chen YT, Austin S, Goldstein JL. Glycogen Storage Disease Type I. In:
Adam MP, Ardinger HH, Pagon RA, et al., eds. GeneReviews((R)). Seattle (WA):
University of Washington, Seattle; 1993-2019.
26. HCUP Kids’ Inpatient Database (KID). Healthcare Cost and Utilization Project
(HCUP). Agency for Healthcare Research and Quality 2016; www.hcup-
us.ahrq.gov/kidoverview.jsp.
27. Lynch JP, 3rd, Sayah DM, Belperio JA, Weigt SS. Lung transplantation for cystic
fibrosis: results, indications, complications, and controversies. Semin Respir Crit
Care Med. 2015;36(2):299-320.
28. Dieleman JL, Cao J, Chapin A, et al. US Health Care Spending by Payer and
Health Condition, 1996-2016. JAMA. 2020;323(9):863-884.
29. Rosenmayr-Templeton L. Industry update for May 2019. Ther Deliv.
2019;10(9):555-561.
61
Appendix
Appendix Table 3.1 Select gene therapy clinical trials for orphan conditions of interest
Condition Gene therapy Sponsor Latest trial
ADA-SCID OTL-101 Orchard Therapeutics NCT02999984
*,†
BT LentiGlobin BB305 Bluebird Bio NCT03207009
‡
ST-400 Sangamo Therapeutics
Bioverativ
NCT03432364
*,†
CF
tgAAVCF Targeted Genetics
Corporation
NCT00073463
§
GD AAV9-CMV-Gba Shanghai Medical
College, Fudan University
Preclinical
HAA Valoctocogene
roxaparvovec (BMN
270)
BioMarin NCT03370913
‡
SPK-8011 Spark NCT03003533
†,‡
HAB Fidanacogene
elaparvovec
Pfizer NCT03861273
‡
AAV5-hFIXco-Padua
(AMT-061)
UniQure Biopharma NCT03569891
‡
MPS-III LYS-SAF302 Lysogene NCT03612869
†,‡
scAAV9.U1a.hSGSH Abeona NCT02716246
*,†
OTCD DTX301
(scAAV8OTC)
Ultragenyx NCT02991144
*,†
PD AAV2/8-LSPhGAA Acutus NCT03533673
*,†
AAV9-GAA_IM University of Florida
Lacerta Therapeutics
Audentes Therapeutics
NCT02240407
*
SCD
LentiGlobin BB305 Bluebird Bio NCT02140554
*,†
RVT-1801 Roivant Sciences
Aruvant Sciences
NCT02186418
*,†
VG DTX401 (AAV8G6PC) Ultragenyx NCT03517085
*,†
Abbreviations: ADA-SCID, adenosine deaminase deficiency with severe combined
immunodeficiency; BT, beta thalassemia; CF, cystic fibrosis; GD, Gaucher disease;
HAA, hemophilia A; HAB, hemophilia B; MPS-III, mucopolysaccharidosis III; OTCD,
ornithine transcarbamylase deficiency; PD, Pompe disease; SCD, sickle cell disease;
VG, Von Gierke disease.
*
Phase 1.
†
Phase 2.
‡
Phase 3.
§
Terminated.
62
Appendix Figure 3.1 Nationally representative aggregated inpatient stay costs for
select diseases
1
, by payer (2016)
63
Chapter 4
The Total Direct Cost of Healthcare in the United
States for Commercially Insured Patients with
Sickle Cell Disease: An Age-Specific Analysis
xviii
Salcedo J
a,b
, Young CM
b
, Bulovic J
b
a
Department of Pharmaceutical and Health Economics, School of
Pharmacy, University of Southern California, Los Angeles, CA, USA
b
Center for Biomedical Innovation, Massachusetts Institute of Technology,
Cambridge, MA, USA
Abstract
Sickle cell disease (SCD) is a group of inherited blood disorders associated with
significant complications and morbidity. There are an estimated 100,000 individuals
xviii
This chapter is based on co-authored work by Salcedo J, Young CM, and Bulovic J. The
conceptualization, data analysis, modeling, and writing of this chapter were primarily my work.
An abstract of a preliminary version of this work was presented at the 61st American Society of
Hematology Annual Meeting. Full citation: Salcedo J, Young CM, Bulovic J. The Total Direct
Cost of Healthcare in the United States in Patients with Sickle Cell Disease: A Propensity Score
Matched Analysis. Blood. 2019;134(Supplement_1):4671-4671.
64
living with SCD in the United States. The age-specific and lifetime costs of healthcare in
commercially insured patients with this condition are not well understood. In this study
we estimated direct costs of chronic management for SCD through a retrospective
database analysis of Optum’s de-identified Clinformatics® Data Mart Database (CDM)
commercial insurance medical and pharmacy claims from January 1, 2007 to December
31, 2017. We identify patients with SCD using administrative diagnosis codes and use
propensity score matching to identify similar control patients unaffected by the disease.
Our primary outcome of interest is the annualized average incremental cost of treating
patients with the disease. We find persistent and statistically significant costs associated
with managing SCD across all age groups and use these estimates to calculate a
theoretical lifetime cost burden.
Introduction
Sickle cell disease (SCD) is an inherited blood disorder associated with significant
morbidity and mortality. It is caused by a point mutation in the beta hemoglobin gene
that forms the sickle hemoglobin (HbS) gene. The disorder is inherited from both
parents, or one HbS from one parent paired with another hemoglobin mutation such as
hemoglobin C, D, E (HbC, HbE, HbD) or beta-thalassemia from the other parent. Those
homozygous for the HbS gene or the compound heterozygosity of HbS and beta-
thalassemia null experience the most severe disease, and are the genotype of sickle
cell anemia (SCA).
1
At low oxygen levels, the abnormal hemoglobin compound
polymerizes, resulting in an damaged erythrocyte that is rigid and shaped like a
crescent moon.
2,3
These changes produce the characteristic hemolytic anemia and
65
small blood vessel obstruction, as well as initiate a cascade of events that are seen in
this disorder.
1,4
There are approximately 100,000 individuals in the United States (US) with SCD,
a number expected to rise with global migration.
5,6
In the US cases are identified early
through a mandatory universal newborn screening (NBS) of SCD performed in all 50
states since 2006.
7
In a 2014 report by the National Heart, Blood, and Lung Institute, an
expert panel supported the practice of universal NBS and recommended a confirmatory
test within 2 months, when necessary.
8
Once a case is identified, clinical management
throughout the course of the disease is variable. This variability of complications and
resulting medical needs, even among individuals within the same genotype, is due to
SCD’s phenotypic heterogeneity partially explained by genetic and non-genetic
factors.
9-11
Infants born with SCD in the US and other high-income countries receive routine
comprehensive care that significantly reduces their morbidity and mortality throughout
childhood. As a result of these interventions, children with SCA in high-income countries
have similar survival to healthy children, and adults can expect to live into the sixth
decade.
12,13
Some of these early interventions in infants are vaccination against
streptococcus pneumoniae, oral penicillin until age 5 for those with SCA, appropriate
screenings, and primary prevention with blood transfusion to reduce the risk of stroke
for those at risk.
8,14-19
Other common interventions include hydroxyurea, pain
medications, and iron chelators.
20
Care is often aimed toward alleviating pain. For this
reason, there is increased interest in the development of new drugs to reduce painful
episodes. These investigative efforts have led to two U.S. Food and Drug Administration
66
(FDA) approvals for interventions that inhibit HbS polymerization and block sickle
erythrocytes adhesion to the vessels surface by binding P-selectin, respectively.
21,22
As individuals live longer and transition into adulthood, the recurrent pain
episodes and multisystem organ damage become more evident, giving rise to chronic
complications.
1
The number of acute painful episodes that require medical care may
decrease as individuals learn to avoid triggers and manage their pain at home.
However, older age is associated with a higher resource utilization during a painful
episode requiring hospitalization.
23
In addition, access to comprehensive care may
decrease with advanced age. This is also reflected in the reduced life expectancy of
adults with SCD compared to the general population.
12
The extent to which disease
management costs vary by age and the overall lifetime cost of treatment in a
commercially insured population are unknown. The aim of this study is to estimate the
total direct healthcare cost among patients with SCD in the US across consecutive age
ranges and use this information to approximate a lifetime cost.
Methods
Data
We utilized real-world evidence (RWE) to estimate the direct cost burden faced by
patients with SCD in the US. Previous analyses have used state Medicaid data to
analyze the publicly insured.
24-27
However, private insurance claims data allow for larger
sample sizes and nationwide coverage which are useful qualities for analyzing orphan
diseases. To this end, as our primary dataset we employ Optum’s de-identified
Clinformatics® Data Mart Database (CDM) medical and pharmacy claims for
67
commercially insured patients. CDM provides information on 15 to 18 million annual
insured lives and roughly 57 million unique lives over 2007 to 2017. The population
spans 50 states and is similar to the US distribution with respect to sex, age, and
geography.
28
Our data is known the oversample southern US states; however, SCD is
highly linked to African ancestry and most cases are expected to exist in these states.
29
Variables
We use various data frames from Optum CDM. These include the member, medical,
pharmacy, diagnosis, and death tables. The member table provides variables including
patient ID number, administrative services only indicator, business product, consumer
driven health plan indicator, eligibility effective date, eligibility end date, gender code,
product, state, and year of birth. From the medical claims table, we utilize claim ID
number, confinement ID number, first date of service, International Classification of
Diseases (ICD) flag, standard cost value and year, and type of service code. From the
pharmacy claims table, we take information on fill date, standard cost value, and
standard cost year. We take the diagnosis code, date, and position variables from the
diagnosis table. Patient and plan ID numbers link patient-level information across most
frames. A summary of patient-level variables is available in Appendix Table 4.1.
Patient identification
We identified patients in Optum CDM using ICD-9-CM (282.41, 282.42, or 282.6x) and
ICD-10-CM (D57.x sans D57.3x) diagnosis codes. Our approach follows a validated
algorithm that has been previously been described in detail.
30
This identification method
68
relies solely on administrative ICD-9-CM and ICD-10-CM diagnosis codes; Reeves et al.
(2014) found using Michigan Medicaid claims that the addition of SCD-related
medications to diagnosis algorithms resulted in “no appreciable improvement” in
diagnosis accuracy over methods utilizing diagnosis codes.
31
Due to well-documented
difficulties in assigning genotype based solely on administrative diagnosis codes, we
only consider patients with SCD in aggregate.
32,33
See Appendix Table 4.2 for a detailed
list of diagnosis codes and corresponding descriptions.
Patients in our data were required to have at least one inpatient (IP) claim or two
or more outpatient (OP) claims (spaced at least 30 days apart) with a diagnosis of SCD
in any slot. We also required a minimum of one year of continuous enrollment in both
medical and prescription drug coverage. Each patient’s index date was designated the
date of their first SCD-related claim for patients with SCD, or their first date of eligibility
for patients without SCD (controls). See Figure 4.1 for a schematic of the observation
period.
Figure 4.1 Study observation period for patients with SCD and matched controls.
Abbreviations: SCD, sickle cell disease.
69
We do not make any distinction between incident and prevalent cases, and
hence do not use a “clean” period. A primary assumption is that as a hereditary
condition, incidence is at birth. For this reason, older patients with recent index dates in
our data should be considered newly identified prevalent cases. Ages in our analysis
are calculated based on age at index, and hence should be interpreted as age at first
disease-related claim for patients in the SCD group, and age at first eligibility for
patients in the control group.
Crises and disease severity
Among individuals with SCD that require medical care, a chief complaint is commonly
acute vaso-occlusive pain episodes (previously described as sickle cell crisis).
34,35
These episodes are also known as vaso-occlusive crises (VOC). The trigger to most of
these episodes is unknown, but some contributory factors are HbS polymerization,
sickle cell blood viscosity, sickle cell adherence to vascular endothelium, and
endothelial cell activation among others.
3,36-42
The incidence of painful episodes
requiring medical attention has been described as 1 to 2.59 events per year.
34,43,44
In
our study, we used ICD-9-CM and ICD-10-CM diagnosis codes from claims as a proxy
for VOC in patients with SCD.
Consistent with methods the literature, we identified claims with a primary or
secondary diagnosis of sickle cell disease with crisis.
45
In ICD-9-CM this included codes
282.42, 282.62, 282.64, and 282.69 which encompass various forms of SCD, all with
crisis. In ICD-10-CM codes are further stratified by complication type. These include
crisis (unspecified), acute chest syndrome, and splenic sequestration. Hence, crisis
70
codes in ICD-10-CM included D57.00, D57.219, D57.419, and D57.819. We counted
multiple VOC-related codes on a claim as a single episode. Multiple claims with a VOC-
related event without at least a three-day gap were also considered a single episode.
See Appendix Table 4.2 for a detailed list of all ICD-9-CM and ICD-10-CM diagnosis
codes used in this study. We used the annualized number of VOCs faced by a patient
as a proxy for disease severity in our study. We then stratified patients into unaffected
(control), mild SCD (zero crises per year), moderate SCD (more than zero but fewer
than two crises per year), and severe SCD (two or more crises per year).
Propensity score matching
Propensity score matching (PSM) is a commonly used statistical method for balancing
groups on observable covariates.
46
A propensity score is defined as the probability of
receiving some exposure, given observable baseline characteristics.
47
Hence,
𝑝 𝑖 = 𝑃 (𝑧 𝑖 = 1|𝑿 𝑖 ) (1)
Where 𝑝 𝑖 is the propensity score, 𝑧 𝑖 is the exposure status (0 or 1), and 𝑿 𝑖 is a vector
of observable covariates, all for patient 𝑖 . While 𝑝 𝑖 is not known, we can estimate it by
regressing exposure on our observable covariates in a binary outcome model (logit for
instance), ensuring the estimate is between zero and one:
𝑝 𝑖 = 𝑃 (𝑧 𝑖 = 1|𝑿 𝑖 ) =
exp(𝑿 𝑖 𝜷 )
1 + exp(𝑿 𝑖 𝜷 )
(2)
71
We can solve for 𝛽 ̂
𝑀𝐿𝐸 using maximum likelihood estimation, which in effect
provides us with an estimate for the propensity score, the predicted probability, 𝑝 ̂
𝑖 .
Affected patients can then be matched to unaffected patients or “controls” to achieve
balance in observable characteristics across groups. In the context of this study, we use
PSM to identify patients with similar characteristics to patients with SCD who are
unaffected by the disease. This allows us to rank control patients based on their
likelihood of having the disease, match them to those affected by SCD, and determine
incremental values of our study outcomes within pairs. PSM also helps us avoid the
issue of calculating all-inclusive healthcare costs for all patients in the database, which
is computationally intractable.
For this study, we matched patients with SCD to controls using 1:1 PSM without
replacement with a greedy nearest neighbor algorithm and caliper of 1/4 the standard
deviation of the propensity score.
48
Propensity scores were generated using logistic
regression on covariates gender, race, education
xix
, geographic division, year of birth,
index year, and plan characteristics. We then used a threshold of less than 10%
absolute standard mean difference (SMD) as a diagnostic for balance in the covariates
across groups post-matching.
49
In a matching sensitivity analysis we employ exact
matching on the same set of variables.
Total direct costs of healthcare including IP, OP, professional (PF), ancillary
(AN), and pharmacy (RX) services were calculated first year post-index (Y1), second
year post-index (Y2), third year post-index (Y3), and overall (annualized) post-index.
Second- and third-year costs were only calculated for 48.3% (3,097 of 6,416) and
xix
Education was measured using an indicator variable for any college attainment.
72
24.2% (1,553 of 6,416) of the original matched sample due to attrition, respectively. We
calculated annualized costs by aggregating all costs after index and dividing by the
remaining number of years the patient was eligible for coverage. Average incremental
costs (AIC) are defined as the mean difference in costs between patients with SCD and
respective controls. We report unadjusted AIC by age group on the patient’s index date,
inflated to 2018 US dollars. We employed statistical tests on our unadjusted outcomes
of interest that account for the matched nature of the data. These include paired t-tests
on paired total cost outcomes or the non-parametric Wilcoxon signed-rank test in the
case of non-normality.
50
Regression analysis
In addition to unadjusted aggregate estimates, we report regression estimated costs
across subgroups including age, gender, and disease severity. Healthcare cost
outcomes are non-negative and often positively skewed, necessitating caution when
estimating costs using regression analysis.
51
To address this, we estimated regression-
adjusted costs using generalized linear models (GLM) with robust standard errors. For
our model we regressed annualized healthcare costs on severity level (control, mild
SCD, moderate SCD, or severe SCD), age at index date, and age interacted with
severity, with separate regressions by gender. We used GLM with log-link and gamma
family which performed the best regarding goodness of fit relative to combinations of log
or square root link and gamma, Gaussian, or Poisson families. See Appendix Table 4.3
for Akaike information criterion (AIC) and Bayesian information criterion (BIC) values by
73
specification. We report our resulting regression estimates including coefficients,
significance levels, and predicted values by subgroup.
Lifetime analysis
To generate a pooled severity estimate of lifetime cost of chronic management for SCD,
we first recover predicted values from GLM regressions. In this specification we
combine all SCD severities into one group. Note that each gender is estimated
separately, and we drop the 𝑖 index for notational convenience.
𝑌𝐸𝐴𝑅𝐿𝑌𝐶𝑂𝑆𝑇 ̂
= exp(𝛽 ̂
0
+ 𝛽 ̂
1
𝑆𝐶𝐷 + 𝛽 ̂
2
𝐴𝐺𝐸 + 𝛽 ̂
3
𝑆𝐶𝐷 ∗ 𝐴𝐺𝐸 ) (3)
We then estimate predicted values for each gender, age, and SCD status. The
predicted values for each subgroup allow us to estimate lifetime costs using death rates
from the medical literature in a simple compartmental model with two states: alive or
dead. We use mortality rates for black or African American individuals for the control
patients and disease-adjusted mortality rates for patients with SCD.
52
We report lifetime
cost and life years by gender and disease status. We discount costs and life years at
3% per year. Our lifetime cost model inputs are available in Appendix Table 4.5.
Human subjects approval and data analysis
Our analysis is limited to de-identified data not collected for the purposes of this study.
For this reason, this study was approved by the University of Southern California’s
University Park Institutional Review Board (UPIRB). All analyses were performed in
74
SAS 9.4 software (SAS Institute, Raleigh, NC) and Stata 16.0 software (StataCorp,
College Station, TX).
Results
Pre-match patient characteristics
Using ICD-9-CM and ICD-10-CM diagnosis codes, we identified N=24,187 patients with
a diagnosis of SCD on any claim in any position. Of these, N=13,804 had at least one
diagnosis associated with an inpatient stay or two or more associated with an outpatient
visit, spaced at least 30 days apart. Lastly, we excluded patients with less than one year
of continuous enrollment following their index date. Our final pre-match sample of
patients with SCD included N=6,428 patients. See Figure 4.2 for a flow chart of included
and excluded patients at each step.
Baseline characteristics pre- and post-PSM were similar for patients with SCD.
Prior to matching, our sample was 60.2% female, with an average age of 37.4 (SD =
22.6), 50.0% black or African American, and 5.1% Hispanic.
75
Figure 4.2 Inclusion and exclusion flow chart. Abbreviations: SCD, sickle cell disease.
Patient characteristics
We identified 6,416 patients with SCD and 6,416 unique matched controls from 2007 to
2017. Our match rates were 99.8% (6,416/6,428) and 97.1% (6,239/6,428) in the base
case and sensitivity analysis, respectively. In the base case, absolute standardized
mean differences were less than 5% between the two groups on all matching variables
xx
(Appendix Table 4.4). See Appendix Figures 4.1a-e for empirical distributions of all
matching variables in patients with sickle cell disease and controls. Our matched
xx
These were all zero in the sensitivity analysis (exact matching) as by definition all variable
values were matched exactly.
76
sample of patients with SCD was 60.3% female, with an average age of 37.3 (SD =
22.6), 50.1% black or African American, and 5.1% Hispanic. The south-Atlantic, east-
south central, and west-south central geographic divisions accounted for 64.2% of
patients with SCD in our data. These divisions comprise the US states/territories of DE,
DC, FL, GA, MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, and TX. Annualized
average total costs (sum of IP, OP, PF, AN, and RX costs) for patients with SCD were
greater at all age categories, relative to controls. Patient characteristics and annualized
incremental costs were similar in our exact matching specification. For this reason, we
report results from the baseline specification (propensity score matching) for the
remainder of the paper. Detailed baseline characteristics by group and specification are
reported in Table 4.1.
77
Table 4.1 Baseline characteristics and annualized healthcare costs by specification and study arm
Specification (1) Propensity score matching* (2) Exact matching
†
Control (N=6,416) SCD (N=6,416) Control (N=6,239) SCD (N=6,239)
Variable Mean/% (SD or N) Mean/% (SD or N) Mean/% (SD or N) Mean/% (SD or N)
Age on index date 37.2 (22.7) 37.3 (22.6) 37.3 (22.7) 37.3 (22.7)
Index year 2011 (3.2) 2010.9 (3.2) 2010.9 (3.2) 2010.9 (3.2)
Enrollment post-index (years) 3.4 (2.4) 3.5 (2.3) 3.4 (2.4) 3.5 (2.3)
Gender
Female 60.1 (3856) 60.3 (3869) 60.3 (3762) 60.3 (3762)
Race/ethnicity
Asian 3.3 (212) 2.5 (163) 2.4 (148) 2.4 (148)
Black/African American 50.2 (3221) 50.1 (3214) 49.8 (3107) 49.8 (3107)
Hispanic 6 (387) 5.1 (329) 5.1 (319) 5.1 (319)
White 27.3 (1752) 29.1 (1867) 29.9 (1865) 29.9 (1865)
Unknown 13.2 (847) 11.3 (725) 12.9 (805) 11.2 (699)
Education
No college attainment 40 (2563) 39.2 (2514) 38.7 (2412) 38.7 (2412)
Product type
Commercial 69.9 (4485) 69.3 (4446) 70.5 (4398) 70.5 (4398)
Medicare 30.1 (1931) 30.7 (1970) 29.5 (1841) 29.5 (1841)
Geography
New England 2.7 (176) 2.6 (168) 2.5 (153) 2.5 (153)
Mid-Atlantic 9.1 (584) 9 (575) 8.8 (551) 8.8 (551)
East-North Central 9.9 (635) 10.1 (648) 10.1 (630) 10.1 (630)
West-North Central 5.3 (342) 5 (321) 4.9 (304) 4.9 (304)
South-Atlantic 45 (2887) 45.3 (2906) 46.1 (2876) 46.1 (2876)
East-South Central 4.9 (312) 4.8 (305) 4.5 (282) 4.5 (282)
West-South Central 14 (898) 14.1 (905) 14.2 (886) 14.2 (886)
Mountain 2.9 (185) 2.7 (170) 2.6 (160) 2.6 (160)
Pacific 5.7 (363) 5.9 (376) 5.9 (369) 5.9 (369)
Unknown 0.5 (32) 0.7 (42) 0.5 (29) 0.5 (29)
Plan types
ASO 39.6 (2541) 39 (2502) 39.4 (2458) 39.4 (2458)
78
EPO 14.9 (956) 13.7 (879) 14.1 (880) 13.7 (855)
HMO 24 (1540) 24.5 (1572) 23.9 (1491) 24.3 (1516)
IND 0.2 (11) 0.4 (24) 0.2 (11) 0.2 (12)
POS 43.6 (2797) 43.4 (2785) 44.3 (2764) 44.5 (2776)
PPO 4.8 (307) 5 (323) 5.1 (316) 4.9 (304)
OTH 12.5 (802) 13 (834) 12.5 (780) 12.5 (780)
CDHP (HRA) 6.3 (401) 7 (451) 6.2 (389) 7.1 (441)
CDHP (HSA) 8.8 (564) 6.9 (440) 7.7 (481) 6.9 (429)
CDHP (Neither) 85 (5454) 86.1 (5524) 86.1 (5372) 86.1 (5372)
Annualized costs (2018 USD)
Total 12521 (38204) 56791 (160922) 12784 (43637) 55765 (161498)
Inpatient 3137 (23409) 20345 (49230) 2926 (16930) 19977 (49076)
Outpatient 3229 (11187) 16294 (46570) 3413 (14403) 15970 (46372)
Professional and ancillary 3946 (12191) 14778 (122960) 4239 (20476) 14507 (124042)
Pharmacy 2209 (8315) 5373 (18421) 2205 (9875) 5311 (18419)
* 1:1 propensity score matching with greedy nearest neighbor algorithm without replacement and caliper of ¼ SD of the
propensity score.
†
1:1 exact matching on same set of variables as specification (1). Abbreviations: SCD, sickle cell
disease; SD, standard deviation; ASO, administrative services only; EPO, exclusive provider organization; HMO, health
maintenance organization; IND, indemnity; OTH, other; POS, point of service; PPO; preferred provider organization;
CDHP, consumer driven health plan; HRA, health reimbursement account; CDHP, consumer driven health plan; HSA,
health savings account; USD, united states dollar.
79
Unadjusted costs
Annualized AIC, i.e. costs attributable to management of SCD, ranged from 20,606 (SD
= 87,594) for ages 0-9 to $70,968 (SD = 503,342) for ages 70-79. First-year AIC were
greater than second and third-year AIC across all age groups except 10-19 years.
Annualized, first, and second-year AIC were all statistically significantly different from
zero for every age group (P ≤ 0.01). Inpatient visits accounted for a plurality of SCD-
related costs for the pooled sample and in seven of the nine age brackets (Figure 4.3).
Details on AIC of SCD management across all consecutive age ranges are available in
Table 4.2.
Figure 4.3 Composition of annualized average incremental cost by age at first disease-
related claim. * P ≤ 0.05; ** P ≤ 0.01; *** P ≤ 0.001.
80
Table 4.2 Annualized and yearly average incremental cost of SCD by age group and time post-index (unadjusted)
AIC of SCD in 2018 USD, mean (SD)
Age at index
date
Annualized
(N = 6,416)
Year 1
(N = 6,416)
Year 2
(N = 3,097)
Year 3
(N = 1,553)
<10 20,606 (87,594)*** 24,635 (126,896)*** 19,385 (77,594)*** 18,529 (62,123)***
10-19 40,535 (101,788)*** 35,857 (85,220)*** 38,818 (179,084)*** 36,226 (130,415)***
20-29 51,841 (109,499)*** 52,406 (108,709)*** 39,596 (87,303)*** 34,120 (83,211)***
30-39 41,206 (102,377)*** 46,837 (108,831)*** 32,215 (89,754)*** 33,884 (69,693)***
40-49 47,281 (123,851)*** 46,073 (111,145)*** 36,728 (99,181)*** 33,571 (83,560)***
50-59 47,393 (127,138)*** 50,658 (129,096)*** 40,366 (131,171)*** 45,197 (139,832)***
60-69 53,616 (128,257)*** 63,284 (148,716)*** 25,766 (92,284)*** 27,039 (72,787)***
70-79 70,968 (503,342)** 89,083 (658,282)** 27,103 (88,912)*** 26,835 (117,421)*
80+ 44,784 (126,658)*** 48,175 (101,298)*** 16,497 (57,798)** 16,962 (76,832)
Total 44,270 (164,203)*** 47,525 (196,575)*** 32,115 (109,137)*** 31,684 (98,298)***
Index date is defined as date of first SCD-related claim (patients with SCD), or first date of plan eligibility (control patients).
Abbreviations: AIC, average incremental cost; SCD, sickle cell disease; SD, standard deviation. * P ≤ 0.05; ** P ≤ 0.01; ***
P ≤ 0.001.
81
Regression results
In females, all coefficients for independent variables and their interactions are
statistically significantly different from zero at P ≤ 0.0001. Severity and age coefficients
have the expected positive signs, and we observe that the effect of age on cost varies
depending on the patient’s level of severity. Among males, the coefficients for mild SCD
(β1, P<0.0001), moderate SCD (β2, P=0.0163), severe SCD (β3, P<0.0001), and age
(β4, P=0.0064) are statistically significantly different from zero. None of the interaction
terms are statistically significant for males. See Tables 4.3.a and 4.3.b for regression
results for females and males, respectively.
82
Table 4.3.a Generalized linear model of annualized total costs, gamma family with
log-link, females
Variable Coeff SE T-Stat P-Val CI low CI high
Mild SCD (β₁) 1.58 0.14 11.36 <0.0001 1.30 1.85
Moderate SCD (β₂) 1.78 0.15 11.58 <0.0001 1.48 2.09
Severe SCD (β₃) 3.16 0.15 21.59 <0.0001 2.88 3.45
Age (β₄) 0.03 0.00 18.07 <0.0001 0.03 0.04
No SCD * Age (base) - - - - - -
Mild SCD * Age (β₅) -0.01 0.00 -3.89 0.0001 -0.02 -0.01
Moderate SCD * Age
(β₆)
-0.01 0.00 -4.32 <0.0001 -0.02 -0.01
Severe SCD * Age
(β₇)
-0.02 0.00 -6.68 <0.0001 -0.03 -0.02
Constant (β₀) 8.05 0.09 94.30 <0.0001 7.88 8.22
Number of obs = 7722
Residual df = 7714
Scale parameter = 5.050005
(1/df) Deviance = 1.84633
(1/df) Pearson = 5.050005
AIC = 22.05741
BIC = -54811.82
Log pseudolikelihood = -85155.65896
Abbreviations: Coeff, coefficient; SE, standard error; T-stat, T-statistic; P-Val; P-value;
CI, 95% confidence interval; SCD, sickle cell disease; obs, observations; df, degrees of
freedom; AIC, Akaike information criterion; BIC, Bayesian information criterion.
83
Table 4.3.b Generalized linear model of annualized total costs, gamma family with
log-link, males
Variable Coeff SE T-Stat P-Val CI low CI high
Mild SCD (β₁) 2.13 0.49 4.31 <0.0001 1.16 3.09
Moderate
SCD (β₂)
1.14 0.47 2.40 0.0163 0.21 2.07
Severe SCD
(β₃)
2.36 0.46 5.13 <0.0001 1.46 3.27
Age (β₄) 0.02 0.01 2.72 0.0064 0.01 0.04
No SCD *
Age (base)
- - - - - -
Mild SCD *
Age (β₅)
-0.02 0.01 -1.69 0.0910 -0.03 0.00
Moderate
SCD * Age
(β₆)
0.01 0.01 0.57 0.5684 -0.02 0.03
Severe SCD *
Age (β₇)
-0.01 0.01 -0.58 0.5616 -0.02 0.01
Constant (β₀) 8.39 0.45 18.56 <0.0001 7.51 9.28
Number of obs = 5110
Residual df = 5102
Scale parameter = 35.26815
(1/df) Deviance = 2.362032
(1/df) Pearson = 35.26815
AIC = 22.03777
BIC = -31514.66
Log pseudolikelihood = -56298.50372
Abbreviations: Coeff, coefficient; SE, standard error; T-stat, T-statistic; P-Val; P-value;
CI, 95% confidence interval; SCD, sickle cell disease; obs, observations; df, degrees of
freedom; AIC, Akaike information criterion; BIC, Bayesian information criterion.
We utilize the regression estimates to generate predicted values for patients by gender,
age, and control/severity status (Figures 4.4.a and b). In these regression estimates,
predicted mean annualized healthcare costs increase steadily by age for both control
patients and patients with SCD. Between the ages of 1 and 88 for instance, we estimate
predicted annual mean costs for patients with SCD of any severity to increase from
$34,973, 95% CI: (30,119 - 39,826) [females] and $32,152 (25,711 - 38,592) [males] to
84
$94,021, 95% CI: (80,767 - 107,275) [females] and $129,993 (76,217 - 183,768)
[males]. For both genders predicted mean costs are lowest for control patients and
highest for patients of greatest SCD severity (characterized by having 2 or more crises
per year in our data).
85
Figure 4.4.a GLM (gamma, log-link) predictive margins of total annualized healthcare cost with 95% confidence intervals
by severity, females. Abbreviations: VOC, vaso-occlusive crisis; SCD, sickle cell disease; USD, United States dollar.
86
Figure 4.4.b GLM (gamma, log-link) predictive margins of total annualized healthcare cost with 95% confidence intervals
by severity, males. Abbreviations: VOC, vaso-occlusive crisis; SCD, sickle cell disease; USD, United States dollar.
87
Lifetime analysis results
We use our gender- and age-specific regression predicted costs and death rates from
the medical literature to estimate outcomes for hypothetical cohorts of patients with
SCD and comparable control patients unaffected by SCD. We assume both cohorts
have the same gender breakdown of 50% males and 50% females. We find patients
with SCD experience on average a total healthcare cost of $2,938,344 over a 54.9-year
lifespan. This figure is substantially higher than the costs estimated for control patients
of $1,108,135 despite a significantly longer 75.9-year lifespan. The present values of
these costs discounted at 3% per year are $1,191,354 and $272,109 for patients born
with SCD and comparable controls, respectively. This implies a discounted difference
attributable to the management of SCD over a lifetime of $919,245 for the average
patient. We find that costs do not differ substantially by gender. These estimates do not
account for potential cost heterogeneity by disease severity. A detailed summary is
available in Table 4.4.
Table 4.4 Pooled severity lifetime cost estimates of patients with SCD and controls
Life Years Direct Healthcare Costs
Discounted
*
SCD Controls Diff. SCD Controls Diff.
Total
†
26.2 29.9 -3.7 $1,191,354 $272,109 $919,245
Females 26.3 30.3 -4.0 $1,181,478 $276,539 $904,940
Males 26.1 29.5 -3.4 $1,201,230 $267,679 $933,551
Undiscounted
Total
†
54.9 75.9 -21.0 $2,938,344 $1,108,135 $1,830,209
Females 55.6 78.9 -23.3 $2,853,023 $1,247,881 $1,605,142
Males 54.3 73.0 -18.6 $3,023,665 $968,388 $2,055,277
*
Life years and costs are discounted at the same rate of 3% per year.
†
Based on an
assumed equal distribution of males and females at birth. Abbreviations: SCD, sickle
cell disease; Diff, difference.
88
Discussion and Limitations
In this study we conducted a retrospective database analysis of a large commercial
insurer in the US to estimate age-specific and lifetime costs of care for patients with
sickle cell disease. We identify and report unadjusted cost estimates for 6,416 patients
with SCD and 6,416 propensity score-matched controls. Among our patients with SCD,
the sample is disproportionately black/African American or Hispanic (55.2%), mean age
37.3, with an average of 3.5 years of follow-up. For these patients, the average
annualized incremental cost of managing the disease is large and consistently
statistically significantly different from zero by age group. These SCD-related costs are
primarily due to inpatient and outpatient visits and range from $20,606/year in the first
decade to $70,968/year in the eighth decade of life. Using regression analysis, we
compute costs by gender, patient severity, and age and use resulting estimates to
calculate a discounted present value of treatment cost over a lifetime. In a cohort of
50% females and 50% males we use life tables for black/African American individuals
and patients with SCD to estimate lifetime cost by gender and disease exposure. For
patients with SCD in aggregate, we calculate a discounted lifetime cost of $1,191,354
over 26.2 discounted life years.
xxi
For a cohort of control patients with the same gender
breakdown we find a discounted lifetime cost of $272,109 over 29.9 discounted life
years.
xxii
These estimates suggest the average patient with SCD will experience 3.7
fewer discounted life years at a $919,245 higher discounted lifetime cost.
xxi
This estimate is $2,938,344 over 54.9 life years when undiscounted.
xxii
This estimate is $1,108,135 over 75.9 life years when undiscounted.
89
Our study is not without limitations. We are using commercial claims data which
may skew our sample towards wealthier patients that receive their coverage through
their employer. However, CDM has been shown to have strong national coverage,
including the South, where we expect most cases to exist. We only require a minimum
one year of continuous enrollment for inclusion, but even among these patients the final
average length of enrollment exceeds three years. We did not use any procedures or
comorbidities in addition to diagnosis codes to identify patients with SCD. Michalik et al.
(2016) used an ICD-based algorithm and found positive predictive value (PPV) of at
least 95.8% and sensitivity of at least 98.3% when applied to two EHR systems in
Wisconsin.
30
In addition, our final sample may be biased toward healthier (or sicker) patients
based on unknown factors. Inclusion and exclusion criteria may bias toward sicker
patients due to the requirement of 1+ IP or 2+ OP disease-related claims. In our data it
is not feasible to stratify by genotype, which is likely linked to severity. In the United
States most cases are likely to be the most severe manifestation, HbSS, commonly
known sickle cell anemia (SCA). One study estimates the percent of HbSS to be
between 63 and 81 percent of SCD cases, depending on the data used.
5
Our
continuous enrollment requirement may result in a patient population that maintains
stable commercial coverage and is hence likely to be less severe on average.
We do not match patients with SCD to controls using a comorbidity score or
index (Charlson, Elixhauser, etc.) as many comorbidities are caused by sickle cell
disease. For this reason, matching on a comorbidity score would bias the sample
towards disproportionately unhealthy controls and as a result underestimate incremental
90
cost of SCD. We used matching without replacement as the benefit of sampling with
replacement is diminished when the ratio of controls to treated is very high, as in our
case.
50
In addition, there are additional analytical considerations when using control
patients more than once; samples cannot be considered independent. We utilized 1:1
matching rather than 1:K matching where K > 1. Matching each treated patient to K
controls is uncommon despite decreasing the variance of the estimator due to increase
in the potential bias. One critical appraisal of studies in the medical literature found that
of 47 studies utilizing PSM between 1996 and 2003, 39 (83%) used 1:1 matching.
50
We
confirm robustness of our matching procedure using a sensitivity check where we
employed exact matching on the same set of variables. Despite a lower match rate, we
found cost outcomes to be comparable. However, due to the lower match rate, we
proceeded with the initial specification. As more individuals are left unmatched there is
an increased risk for bias which may be larger than the bias from inexact matches.
53,54
Our data, like many commercial claims databases, lacks information on how
patients exit the data. For this reason, it is not possible to classify reasons for attrition,
e.g. plan switching, loss of employment/coverage, or death. As well, our lifetime
analysis assumes the mean patient receives the average care of patients across all
ages of patients in the database. Lastly, a large proportion of patients with SCD in US
are covered by Medicaid, not commercial plans. In Medicaid populations treatment
pathways, healthcare resource utilization, reimbursement rates, health related quality of
life, life expectancy, and hence direct lifetime cost of care are likely to differ from
commercial plans. Despite this, our analysis provides insight on a population that is less
91
studied than patients covered by state Medicaid or Children’s Health Insurance
Programs (CHIP).
Conclusions
The cost of disease management for commercially insured patients with SCD in the US
is substantial. We observe a rise in SCD-specific costs from childhood to early
adulthood. This may reflect the higher risk of disease-associated complications and
mortality seen during this period. Additionally, AIC first year after initial SCD-related
claim are consistently higher compared to AIC in years two and three. This finding may
be attributable to an increased utilization of healthcare services from previously
underinsured patients with SCD. Using regression analysis, we find costs rise steadily
with increased age across all subgroups including gender and by disease severity.
Future work should involve simulating estimated lifetime total indirect cost of healthcare
attributable to management of SCD for patients in the US.
92
References
1. Rees DC, Williams TN, Gladwin MT. Sickle-cell disease. Lancet.
2010;376(9757):2018-2031.
2. Steinberg MH. Pathophysiology of sickle cell disease. Baillieres Clin Haematol.
1998;11(1):163-184.
3. Vekilov PG. Sickle-cell haemoglobin polymerization: is it the primary pathogenic
event of sickle-cell anaemia? Br J Haematol. 2007;139(2):173-184.
4. Sundd P, Gladwin MT, Novelli EM. Pathophysiology of Sickle Cell Disease. Annu
Rev Pathol. 2019;14:263-292.
5. Hassell KL. Population estimates of sickle cell disease in the U.S. Am J Prev
Med. 2010;38(4 Suppl):S512-521.
6. Piel FB, Tatem AJ, Huang Z, Gupta S, Williams TN, Weatherall DJ. Global
migration and the changing distribution of sickle haemoglobin: a quantitative
study of temporal trends between 1960 and 2000. Lancet Glob Health.
2014;2(2):e80-89.
7. Benson JM, Therrell BL, Jr. History and current status of newborn screening for
hemoglobinopathies. Semin Perinatol. 2010;34(2):134-144.
8. National Heart, Lung, and Blood Institute. Evidence-based management of sickle
cell disease: expert panel report, 2014. Washington, DC: National Institutes of
Health. 2014.
9. Piel FB, Steinberg MH, Rees DC. Sickle Cell Disease. N Engl J Med.
2017;377(3):305.
93
10. Ngo D, Bae H, Steinberg MH, et al. Fetal hemoglobin in sickle cell anemia:
genetic studies of the Arab-Indian haplotype. Blood Cells Mol Dis. 2013;51(1):22-
26.
11. Steinberg MH. Predicting clinical severity in sickle cell anaemia. Br J Haematol.
2005;129(4):465-481.
12. Gardner K, Douiri A, Drasar E, et al. Survival in adults with sickle cell disease in
a high-income setting. Blood. 2016;128(10):1436-1438.
13. Quinn CT, Rogers ZR, McCavit TL, Buchanan GR. Improved survival of children
and adolescents with sickle cell disease. Blood. 2010;115(17):3447-3452.
14. Adams RJ, Brambilla DJ, Granger S, et al. Stroke and conversion to high risk in
children screened with transcranial Doppler ultrasound during the STOP study.
Blood. 2004;103(10):3689-3694.
15. Adams RJ, McKie VC, Hsu L, et al. Prevention of a first stroke by transfusions in
children with sickle cell anemia and abnormal results on transcranial Doppler
ultrasonography. N Engl J Med. 1998;339(1):5-11.
16. Falletta JM, Woods GM, Verter JI, et al. Discontinuing penicillin prophylaxis in
children with sickle cell anemia. Prophylactic Penicillin Study II. J Pediatr.
1995;127(5):685-690.
17. Gaston MH, Verter JI, Woods G, et al. Prophylaxis with oral penicillin in children
with sickle cell anemia. A randomized trial. N Engl J Med. 1986;314(25):1593-
1599.
94
18. John AB, Ramlal A, Jackson H, Maude GH, Sharma AW, Serjeant GR.
Prevention of pneumococcal infection in children with homozygous sickle cell
disease. Br Med J (Clin Res Ed). 1984;288(6430):1567-1570.
19. Nkouwap I DJ, Noyon I, Etienne-Julan M, Merault L. Is there any alternative to
oral penicillin in antibioprophylaxis for children with sickle cell disease? [French]
Y a-t-il une alternative a la penicilline orale dans l'antibioprophylaxie chez les
enfants drepanocytaires? Med Mal Infect. 1999;29(2):111-116.
20. Kato GJ, Piel FB, Reid CD, et al. Sickle cell disease. Nat Rev Dis Primers.
2018;4:18010.
21. Vichinsky E, Hoppe CC, Ataga KI, et al. A Phase 3 Randomized Trial of
Voxelotor in Sickle Cell Disease. N Engl J Med. 2019;381(6):509-519.
22. Ataga KI, Kutlar A, Kanter J, et al. Crizanlizumab for the Prevention of Pain
Crises in Sickle Cell Disease. N Engl J Med. 2017;376(5):429-439.
23. Raphael JL, Mei M, Mueller BU, Giordano T. High resource hospitalizations
among children with vaso-occlusive crises in sickle cell disease. Pediatr Blood
Cancer. 2012;58(4):584-590.
24. Kauf TL, Coates TD, Huazhi L, Mody-Patel N, Hartzema AG. The cost of health
care for children and adults with sickle cell disease. Am J Hematol.
2009;84(6):323-327.
25. Carroll CP, Haywood C, Jr., Fagan P, Lanzkron S. The course and correlates of
high hospital utilization in sickle cell disease: Evidence from a large, urban
Medicaid managed care organization. Am J Hematol. 2009;84(10):666-670.
95
26. Blinder MA, Duh MS, Sasane M, Trahey A, Paley C, Vekeman F. Age-Related
Emergency Department Reliance in Patients with Sickle Cell Disease. J Emerg
Med. 2015;49(4):513-522.e511.
27. Blinder MA, Vekeman F, Sasane M, Trahey A, Paley C, Duh MS. Age-related
treatment patterns in sickle cell disease patients and the associated sickle cell
complications and healthcare costs. Pediatr Blood Cancer. 2013;60(5):828-835.
28. Optum.com. Optum research data assets.
https://www.optum.com/content/dam/optum/resources/productSheets/5302_Data
_Assets_Chart_Sheet_ISPOR.pdf. Accessed August 4, 2019.
29. Solovieff N, Hartley SW, Baldwin CT, et al. Ancestry of African Americans with
sickle cell disease. Blood Cells Mol Dis. 2011;47(1):41-45.
30. Michalik DE, Taylor BW, Panepinto JA. Identification and Validation of a Sickle
Cell Disease Cohort Within Electronic Health Records. Acad Pediatr.
2017;17(3):283-287.
31. Reeves S, Garcia E, Kleyn M, et al. Identifying sickle cell disease cases using
administrative claims. Acad Pediatr. 2014;14(5 Suppl):S61-67.
32. Amendah DD, Mvundura M, Kavanagh PL, Sprinz PG, Grosse SD. Sickle cell
disease-related pediatric medical expenditures in the U.S. Am J Prev Med.
2010;38(4 Suppl):S550-556.
33. Snyder AB, Lane PA, Zhou M, Paulukonis ST, Hulihan MM. The accuracy of
hospital ICD-9-CM codes for determining Sickle Cell Disease genotype. J Rare
Dis Res Treat. 2017;2(4):39-45.
96
34. Field JJ. Five lessons learned about long-term pain management in adults with
sickle cell disease. Hematology Am Soc Hematol Educ Program.
2017;2017(1):406-411.
35. Yusuf HR, Atrash HK, Grosse SD, Parker CS, Grant AM. Emergency department
visits made by patients with sickle cell disease: a descriptive study, 1999-2007.
Am J Prev Med. 2010;38(4 Suppl):S536-541.
36. Schmalzer EA, Lee JO, Brown AK, Usami S, Chien S. Viscosity of mixtures of
sickle and normal red cells at varying hematocrit levels. Implications for
transfusion. Transfusion. 1987;27(3):228-233.
37. Setty BN, Betal SG. Microvascular endothelial cells express a phosphatidylserine
receptor: a functionally active receptor for phosphatidylserine-positive
erythrocytes. Blood. 2008;111(2):905-914.
38. Zhou Z, Thiagarajan P, Udden M, Lopez JA, Guchhait P. Erythrocyte membrane
sulfatide plays a crucial role in the adhesion of sickle erythrocytes to
endothelium. Thromb Haemost. 2011;105(6):1046-1052.
39. Closse C, Seigneur M, Renard M, et al. Influence of hypoxia and hypoxia-
reoxygenation on endothelial P-selectin expression. Thromb Res.
1997;85(2):159-164.
40. Setty BN, Chen D, O'Neal P, Littrell JB, Grossman MH, Stuart MJ. Eicosanoids in
sickle cell disease: potential relevance of 12(S)-hydroxy-5,8,10,14-
eicosatetraenoic acid to the pathophysiology of vaso-occlusion. J Lab Clin Med.
1998;131(4):344-353.
97
41. Setty BN, Stuart MJ, Dampier C, Brodecki D, Allen JL. Hypoxaemia in sickle cell
disease: biomarker modulation and relevance to pathophysiology. Lancet.
2003;362(9394):1450-1455.
42. Wood K, Russell J, Hebbel RP, Granger DN. Differential expression of E- and P-
selectin in the microvasculature of sickle cell transgenic mice. Microcirculation.
2004;11(4):377-385.
43. Brousseau DC, Owens PL, Mosso AL, Panepinto JA, Steiner CA. Acute care
utilization and rehospitalizations for sickle cell disease. JAMA.
2010;303(13):1288-1294.
44. Platt OS, Thorington BD, Brambilla DJ, et al. Pain in sickle cell disease. Rates
and risk factors. N Engl J Med. 1991;325(1):11-16.
45. Shah N, Bhor M, Xie L, Paulose J, Yuce H. Sickle cell disease complications:
Prevalence and resource utilization. PLoS One. 2019;14(7):e0214355.
46. Austin PC. An Introduction to Propensity Score Methods for Reducing the Effects
of Confounding in Observational Studies. Multivariate Behav Res.
2011;46(3):399-424.
47. Rosenbaum PR, Rubin DB. The central role of the propensity score in
observational studies for causal effects. Biometrika. 1983;70(1):41-55.
48. Stuart EA, Rubin DB. Matching With Multiple Control Groups With Adjustment for
Group Differences. Journal of Educational and Behavioral Statistics.
2008;33(3):279-306.
98
49. Austin PC. Goodness-of-fit diagnostics for the propensity score model when
estimating treatment effects using covariate adjustment with the propensity
score. Pharmacoepidemiol Drug Saf. 2008;17(12):1202-1217.
50. Austin PC. A critical appraisal of propensity-score matching in the medical
literature between 1996 and 2003. Stat Med. 2008;27(12):2037-2049.
51. Briggs A, Gray A. The distribution of health care costs and their statistical
analysis for economic evaluation. J Health Serv Res Policy. 1998;3(4):233-245.
52. Lubeck D, Agodoa I, Bhakta N, et al. Estimated Life Expectancy and Income of
Patients With Sickle Cell Disease Compared With Those Without Sickle Cell
Disease. JAMA Netw Open. 2019;2(11):e1915374.
53. Stuart EA. Matching methods for causal inference: A review and a look forward.
Stat Sci. 2010;25(1):1-21.
54. Rosenbaum PR, Rubin DB. Constructing a control group using multivariate
matched sampling methods that incorporate the propensity score. The American
Statistician. 1985;39(1):33-38.
55. Fingar KR, Owens PL, Reid LD, Mistry KB, Barrett ML. Characteristics of
Inpatient Hospital Stays Involving Sickle Cell Disease, 2000-2016: Statistical
Brief #251. In: Healthcare Cost and Utilization Project (HCUP) Statistical Briefs.
Rockville (MD): Agency for Healthcare Research and Quality (US); 2019.
99
Appendix
Appendix Table 4.1 Patient level variables, values, and definitions
Variable Values Description
Patient ID Number Encrypted number that identifies individual
across multiple groups/policies
Year of Birth Year Member's year of birth, capped at 90 years
of age
Gender Male Male Gender
Female Female Gender
Unknown Unknown Gender
Race/Ethnicity Asian Asian Race/Ethnicity
Black/African
American
Black/African American Race/Ethnicity
Hispanic Hispanic Race/Ethnicity
White White Race/Ethnicity
No College Yes/No Indicator of any college attainment
Eligibility Effective Date Date coverage is effective
Eligibility End Date Date coverage ended
Administrative
Services Only
Yes/No Identifies if financial arrangement is
administrative services only
Business Product COM Commercial Plan
MCR Medicare Advantage Plan
Consumer Driven
Health Plan
(CDHP)
CDHP: HSA CDHP with health savings account
CDHP: HRA CDHP with health reimbursement account
CDHP: Neither
Plan Type ALL National Ancillaries Contracted for All
Products
EPO Exclusive Provider Organization
GPO Group Purchasing Organization
HMO Health Maintenance Organization
IND Indemnity
IPP Individual Program Plan
NONE No Industry Product Code
OTH Other
POS Point of Service
PPO Preferred Provider Organization
SPN State Policy Network
UNK Unknown
Geographic
Division
New England CT, ME, MA, NH, RI, VT
Middle Atlantic NJ, NY, PA
East North
Central
IL, IN, MI, OH, WI
West North
Central
IA, KS, MN, MO, NE, ND, SD
South Atlantic DE, DC, FL, GA, MD, NC, SC, VA, WV
100
East South
Central
AL, KY, MS, TN
West South
Central
AR, LA, OK, TX
Mountain AZ, CO, ID, MT, NV, NM, UT, WY
Pacific AK, CA, HI, OR, WA
Unknown Other
Abbreviations: ID, identification document; CT, Connecticut; ME, Maine; MA,
Massachusetts; NH, New Hampshire; RI, Rhode Island; VT, Vermont; NJ, New Jersey;
NY, New York; PA, Pennsylvania; IL, Illinois; IN, Indiana; MI, Michigan; OH, Ohio; WI,
Wisconsin; IA, Iowa; KS, Kansas; MN, Minnesota; MO, Missouri; NE, Nebraska; ND,
North Dakota; SD, South Dakota; DE, Delaware; DC, District of Columbia; FL, Florida;
GA, Georgia; MD, Maryland; NC, North Carolina; SC, South Carolina; VA, Virginia; WV,
West Virginia; AL, Alabama; KY, Kentucky; MS, Mississippi; TN, Tennessee; AR,
Arkansas; LA, Louisiana; OK, Oklahoma; TX, Texas; AZ, Arizona; CO, Colorado; ID,
Idaho; MT, Montana; NV, Nevada; NM; New Mexico; UT, Utah; WY, Wyoming; AK,
Arkansas; CA, California; HI, Hawaii; OR, Oregon; WA, Washington.
101
Appendix Table 4.2 Diagnosis codes of sickle cell disease, ICD-9-CM and ICD-10-CM
ICD-9-CM code Description
282.42 Sickle-cell thalassemia with crisis
282.62 Hb-SS disease with crisis
282.64 Sickle-cell/Hb-C disease with crisis
282.69 Other sickle-cell disease with crisis
282.41 Sickle-cell thalassemia without crisis
282.6 Sickle-cell disease, unspecified
282.61 Hb-SS disease without crisis
282.63 Sickle-cell/Hb-C disease without crisis
282.68 Other sickle-cell disease without crisis
ICD-10-CM code Description
D57.00 Hb-SS disease with crisis, unspecified
D57.01 Hb-SS disease with acute chest syndrome
D57.02 Hb-SS disease with splenic sequestration
D57.211 Sickle-cell/Hb-C disease with acute chest syndrome
D57.212 Sickle-cell/Hb-C disease with splenic sequestration
D57.219 Sickle-cell/Hb-C disease with crisis, unspecified
D57.411 Sickle-cell thalassemia with acute chest syndrome
D57.412 Sickle-cell thalassemia with splenic sequestration
D57.419 Sickle-cell thalassemia with crisis, unspecified
D57.811 Other sickle-cell disorders with acute chest syndrome
D57.812 Other sickle-cell disorders with splenic sequestration
D57.819 Other sickle-cell disorders with crisis, unspecified
D57.1 Sickle-cell disease without crisis
D57.20 Sickle-cell/Hb-C disease without crisis
D57.40 Sickle-cell thalassemia without crisis
D57.80 Other sickle-cell disorders without crisis
Adapted from Fingar et al. (2019).
55
Abbreviations: ICD-9-CM, International
Classification of Diseases, Ninth Revision, Clinical Modification; ICD-10-CM,
International Classification of Diseases, Tenth Revision, Clinical Modification; Hb-C,
hemoglobin C; Hb-SS, hemoglobin sickle cell.
102
Appendix Table 4.3 Goodness of fit comparisons for generalized linear models
Specification N Log-likelihood df AIC BIC
Log-link, gamma 12,832 -141,575 8 283,167 283,226
Sqrt-link, gamma 12,832 -141,618 8 283,251 283,311
Log-link, Gaussian 12,832 -167,773 6 335,557 335,602
Sqrt-link, Gaussian 12,832 -167,735 8 335,486 335,546
Log-link, Poisson 12,832 -408,095,352 8 816,190,719 816,190,779
Sqrt-link, Poisson 12,832 -408,332,061 8 816,664,139 816,664,198
Abbreviations: df, degrees of freedom; AIC, Akaike information criterion; BIC, Bayesian
information criterion; Sqrt, square-root.
103
Appendix Table 4.4 Standardized mean differences for matching variables by group
Control (N=6,416) SCD (N=6,416)
Variable Mean SD Mean SD Std. Diff.
Gender: Male 0.3992 0.48976 0.3973 0.48937 0.00382
Gender: Female 0.6008 0.48976 0.6027 0.48937 -0.00382
Race: Asian 0.03304 0.17876 0.02541 0.15736 0.04535
Race: Black 0.5025 0.50003 0.5006 0.50004 0.00374
Race: Hispanic 0.06032 0.23809 0.05128 0.22058 0.03939
Race: White 0.2726 0.44533 0.2913 0.4544 -0.04157
Education: No College 0.3995 0.48983 0.3918 0.4882 0.01562
Division: New England 0.02743 0.16335 0.02618 0.1597 0.00772
Division: Mid Atlantic 0.09102 0.28766 0.08962 0.28566 0.00489
Division: East North
Central
0.09897 0.29865 0.1008 0.30114 -0.00624
Division: West North
Central
0.0533 0.22466 0.05003 0.21803 0.01479
Division: South Atlantic 0.45 0.49753 0.4532 0.49785 -0.00658
Division: East South
Central
0.04863 0.21511 0.04754 0.2128 0.0051
Division: West South
Central
0.1403 0.3473 0.1409 0.34794 -0.00179
Division: Mountain 0.02883 0.16735 0.0265 0.16062 0.01425
Division: Pacific 0.05658 0.23105 0.0586 0.2349 -0.0087
Division: Unknown 0.004988 0.070452 0.006546 0.080649 -0.02058
Year of Birth 1974 22.275 1974 22.175 0.00854
Index Year 2011 3.2349 2011 3.2012 0.02562
Plan: ASO 0.3957 0.48904 0.3897 0.48771 0.01245
Plan: HMO/EPO 0.3892 0.4876 0.3825 0.48603 0.01377
Plan: PPO/POS 0.4843 0.49979 0.4843 0.49979 0
Plan: Other 0.1266 0.3325 0.1333 0.33988 -0.01993
Plan: No CDHP 0.8496 0.3575 0.8611 0.34584 -0.03279
Plan: Commercial 0.6992 0.45865 0.6931 0.46124 0.01322
Plan: Medicare 0.3008 0.45865 0.3069 0.46124 -0.01322
Abbreviations: SCD, sickle cell disease; SD, standard deviation; ASO, administrative
services only; EPO, exclusive provider organization; HMO, health maintenance
organization; POS, point of service; PPO; preferred provider organization; CDHP,
consumer driven health plan.
104
Appendix Table 4.5 Regression predicted annualized cost by exposure and subgroup (2018 US dollars)
Females Males
Controls SCD Controls SCD
Age Mean LB UB Mean LB UB Mean LB UB Mean LB UB
1 3,236 2,704 3,768 34,973 30,119 39,826 4,530 590.5 8,469 32,152 25,711 38,592
2 3,340 2,801 3,878 35,372 30,556 40,189 4,639 683.6 8,594 32,672 26,320 39,025
3 3,447 2,902 3,993 35,777 30,998 40,556 4,751 780.8 8,721 33,201 26,940 39,463
4 3,558 3,005 4,110 36,186 31,446 40,926 4,865 882.3 8,848 33,738 27,571 39,906
5 3,672 3,113 4,231 36,600 31,900 41,299 4,983 988.1 8,977 34,285 28,213 40,356
6 3,790 3,224 4,356 37,018 32,359 41,677 5,103 1,099 9,107 34,840 28,866 40,813
7 3,911 3,338 4,484 37,441 32,824 42,058 5,226 1,214 9,238 35,404 29,531 41,276
8 4,037 3,457 4,616 37,869 33,295 42,443 5,352 1,334 9,371 35,977 30,206 41,747
9 4,166 3,580 4,752 38,302 33,772 42,832 5,481 1,459 9,504 36,559 30,892 42,226
10 4,300 3,707 4,892 38,740 34,255 43,225 5,613 1,589 9,638 37,151 31,589 42,713
11 4,438 3,839 5,037 39,183 34,743 43,622 5,749 1,725 9,773 37,752 32,296 43,209
12 4,580 3,975 5,186 39,631 35,238 44,024 5,888 1,866 9,909 38,363 33,012 43,715
13 4,727 4,116 5,339 40,084 35,738 44,430 6,030 2,013 10,047 38,984 33,738 44,231
14 4,879 4,261 5,497 40,542 36,244 44,840 6,175 2,165 10,185 39,615 34,472 44,759
15 5,035 4,411 5,659 41,006 36,756 45,255 6,324 2,324 10,324 40,257 35,214 45,299
16 5,197 4,567 5,827 41,474 37,274 45,675 6,477 2,490 10,464 40,908 35,963 45,853
17 5,364 4,727 6,000 41,948 37,797 46,100 6,633 2,661 10,605 41,570 36,718 46,423
18 5,536 4,893 6,178 42,428 38,326 46,530 6,793 2,840 10,746 42,243 37,478 47,009
19 5,713 5,065 6,362 42,913 38,862 46,965 6,957 3,025 10,889 42,927 38,241 47,613
20 5,897 5,242 6,551 43,404 39,402 47,405 7,125 3,218 11,032 43,622 39,006 48,237
21 6,086 5,425 6,747 43,900 39,949 47,851 7,297 3,417 11,176 44,328 39,772 48,884
22 6,281 5,614 6,948 44,402 40,500 48,303 7,473 3,625 11,320 45,046 40,537 49,555
23 6,483 5,809 7,156 44,909 41,057 48,761 7,653 3,840 11,466 45,775 41,298 50,252
24 6,691 6,011 7,370 45,423 41,620 49,226 7,838 4,063 11,612 46,516 42,054 50,978
25 6,905 6,219 7,591 45,942 42,187 49,697 8,027 4,295 11,758 47,269 42,803 51,734
26 7,127 6,434 7,820 46,467 42,760 50,175 8,220 4,535 11,906 48,034 43,543 52,524
27 7,356 6,656 8,055 46,998 43,337 50,660 8,419 4,784 12,053 48,811 44,274 53,349
28 7,591 6,884 8,299 47,536 43,919 51,153 8,622 5,041 12,202 49,602 44,993 54,210
29 7,835 7,120 8,550 48,079 44,505 51,653 8,830 5,308 12,351 50,404 45,699 55,110
105
30 8,086 7,363 8,810 48,629 45,095 52,162 9,043 5,584 12,501 51,220 46,393 56,048
31 8,346 7,613 9,078 49,185 45,689 52,680 9,261 5,870 12,652 52,049 47,073 57,026
32 8,614 7,871 9,356 49,747 46,286 53,208 9,484 6,166 12,803 52,892 47,739 58,045
33 8,890 8,137 9,643 50,316 46,887 53,744 9,713 6,471 12,955 53,748 48,393 59,103
34 9,175 8,411 9,940 50,891 47,490 54,292 9,947 6,787 13,108 54,618 49,034 60,202
35 9,470 8,693 10,246 51,473 48,096 54,849 10,187 7,113 13,262 55,502 49,663 61,341
36 9,773 8,983 10,564 52,061 48,704 55,418 10,433 7,449 13,418 56,401 50,281 62,520
37 10,087 9,281 10,893 52,656 49,313 55,999 10,685 7,795 13,575 57,314 50,889 63,738
38 10,411 9,588 11,233 53,258 49,924 56,593 10,943 8,152 13,734 58,241 51,488 64,995
39 10,745 9,904 11,586 53,867 50,535 57,199 11,207 8,518 13,895 59,184 52,077 66,291
40 11,089 10,228 11,951 54,483 51,147 57,819 11,477 8,894 14,060 60,142 52,660 67,625
41 11,445 10,561 12,330 55,106 51,759 58,452 11,754 9,280 14,228 61,116 53,235 68,997
42 11,812 10,903 12,722 55,736 52,371 59,100 12,037 9,673 14,402 62,105 53,804 70,407
43 12,191 11,255 13,128 56,373 52,982 59,764 12,328 10,074 14,582 63,110 54,367 71,854
44 12,583 11,616 13,550 57,017 53,593 60,442 12,625 10,481 14,770 64,132 54,925 73,339
45 12,986 11,986 13,987 57,669 54,202 61,136 12,930 10,890 14,970 65,170 55,478 74,862
46 13,403 12,366 14,440 58,329 54,810 61,847 13,242 11,300 15,183 66,225 56,027 76,423
47 13,833 12,757 14,909 58,995 55,417 62,574 13,561 11,707 15,416 67,297 56,572 78,022
48 14,277 13,157 15,396 59,670 56,022 63,317 13,888 12,105 15,672 68,386 57,114 79,659
49 14,735 13,568 15,902 60,352 56,626 64,078 14,224 12,489 15,958 69,493 57,653 81,334
50 15,208 13,990 16,426 61,042 57,228 64,855 14,567 12,853 16,280 70,618 58,188 83,048
51 15,696 14,422 16,969 61,740 57,830 65,650 14,918 13,192 16,644 71,761 58,721 84,802
52 16,199 14,866 17,532 62,446 58,429 66,462 15,278 13,501 17,055 72,923 59,251 86,595
53 16,719 15,322 18,116 63,159 59,028 67,290 15,647 13,778 17,515 74,103 59,778 88,429
54 17,255 15,789 18,722 63,881 59,626 68,137 16,024 14,024 18,024 75,303 60,303 90,303
55 17,809 16,269 19,349 64,612 60,224 69,000 16,411 14,240 18,581 76,522 60,826 92,218
56 18,380 16,761 20,000 65,350 60,820 69,880 16,807 14,431 19,183 77,761 61,346 94,175
57 18,970 17,266 20,674 66,097 61,417 70,778 17,212 14,598 19,826 79,019 61,864 96,175
58 19,579 17,784 21,373 66,853 62,013 71,693 17,627 14,745 20,510 80,298 62,379 98,218
59 20,207 18,316 22,098 67,617 62,610 72,625 18,053 14,874 21,231 81,598 62,892 100,304
60 20,855 18,862 22,849 68,390 63,207 73,574 18,488 14,987 21,989 82,919 63,403 102,435
61 21,524 19,422 23,626 69,172 63,805 74,540 18,934 15,087 22,782 84,261 63,912 104,611
62 22,215 19,998 24,432 69,963 64,403 75,523 19,391 15,173 23,609 85,625 64,418 106,833
106
63 22,928 20,589 25,267 70,763 65,003 76,523 19,859 15,247 24,471 87,011 64,921 109,102
64 23,663 21,195 26,132 71,572 65,604 77,540 20,338 15,309 25,366 88,420 65,422 111,418
65 24,423 21,818 27,027 72,390 66,206 78,574 20,829 15,361 26,297 89,851 65,920 113,782
66 25,206 22,457 27,955 73,218 66,810 79,625 21,331 15,400 27,262 91,305 66,415 116,196
67 26,015 23,114 28,916 74,055 67,416 80,693 21,846 15,429 28,262 92,783 66,907 118,660
68 26,850 23,788 29,911 74,901 68,024 81,779 22,373 15,447 29,298 94,285 67,396 121,175
69 27,711 24,480 30,941 75,757 68,633 82,882 22,913 15,454 30,371 95,812 67,881 123,742
70 28,600 25,192 32,009 76,624 69,245 84,002 23,465 15,449 31,481 97,362 68,363 126,362
71 29,518 25,922 33,113 77,499 69,860 85,139 24,031 15,433 32,630 98,938 68,842 129,035
72 30,465 26,672 34,257 78,385 70,477 86,294 24,611 15,405 33,817 100,540 69,317 131,763
73 31,442 27,443 35,442 79,282 71,096 87,467 25,205 15,365 35,045 102,168 69,787 134,548
74 32,451 28,234 36,668 80,188 71,719 88,657 25,813 15,312 36,314 103,821 70,254 137,389
75 33,492 29,047 37,938 81,105 72,344 89,866 26,436 15,246 37,626 105,502 70,716 140,287
76 34,567 29,882 39,252 82,032 72,972 91,092 27,074 15,166 38,981 107,210 71,174 143,245
77 35,676 30,739 40,612 82,970 73,603 92,336 27,727 15,073 40,381 108,945 71,627 146,263
78 36,820 31,620 42,021 83,918 74,237 93,599 28,396 14,965 41,827 110,709 72,075 149,343
79 38,002 32,525 43,479 84,877 74,875 94,880 29,081 14,842 43,320 112,501 72,517 152,484
80 39,221 33,454 44,988 85,848 75,515 96,180 29,782 14,703 44,862 114,322 72,954 155,690
81 40,479 34,409 46,550 86,829 76,159 97,499 30,501 14,548 46,454 116,172 73,385 158,960
82 41,778 35,390 48,167 87,822 76,807 98,837 31,237 14,376 48,097 118,053 73,810 162,296
83 43,119 36,397 49,840 88,826 77,458 100,194 31,990 14,187 49,794 119,964 74,229 165,699
84 44,502 37,432 51,572 89,841 78,112 101,571 32,762 13,979 51,545 121,906 74,641 169,170
85 45,930 38,495 53,365 90,868 78,770 102,967 33,552 13,752 53,353 123,879 75,046 172,712
86 47,404 39,587 55,220 91,907 79,432 104,382 34,362 13,506 55,218 125,884 75,444 176,324
87 48,924 40,708 57,141 92,958 80,097 105,818 35,191 13,239 57,143 127,922 75,834 180,009
88 50,494 41,860 59,128 94,021 80,767 107,275 36,040 12,951 59,129 129,993 76,217 183,768
Abbreviations: SCD, sickle cell disease; LB, 95% confidence interval lower bound; UB, 95% confidence interval upper
bound.
107
Appendix Figure 4.1.a Empirical distributions of matching variables in patients with sickle cell disease and controls
108
Appendix Figure 4.1.b Empirical distributions of matching variables in patients with sickle cell disease and controls
Abbreviations: DIV, geographic division.
109
Appendix Figure 4.1.c Empirical distributions of matching variables in patients with sickle cell disease and controls
Abbreviations: DIV, geographic division; ASO, administrative services only; EPO, exclusive provider organization; HMO,
health maintenance organization.
110
Appendix Figure 4.1.d Empirical distributions of matching variables in patients with sickle cell disease and controls
Abbreviations: POS, point of service; PPO; preferred provider organization; CDHP, consumer driven health plan.
111
Appendix Figure 4.1.e Empirical distributions of matching variables in patients with sickle cell disease and controls
Abbreviations: Unk, unknown; edu, education.
112
Appendix Figure 4.2.a Linear predictive margins with 95% confidence intervals by severity, females
Abbreviations: VOC, vaso-occlusive crisis; SCD, sickle cell disease; USD, United States dollar.
113
Appendix Figure 4.2.b Linear predictive margins with 95% confidence intervals by severity, males
Abbreviations: VOC, vaso-occlusive crisis; SCD, sickle cell disease; USD, United States dollar.
114
Chapter 5
Cost-Effectiveness of a Hypothetical Durable
Treatment for Sickle Cell Disease
Abstract
Background: Sickle cell disease (SCD) is a group of inherited genetic conditions
associated with lifelong complications and increased healthcare resource utilization.
There are an estimated 100,000 patients living with the disease in the United States.
Standard treatment for SCD in the US varies based on stage of the disease and
observed clinical severity. In this study we aim to evaluate the potential cost-
effectiveness of a durable cure for sickle cell disease from the US healthcare sector
perspective.
Methods: We developed a Markov model to evaluate the cost-effectiveness of a
hypothetical single-administration durable treatment (DT) for SCD provided at birth,
relative to standard of care (SOC), over a lifetime. We informed model inputs including
direct healthcare costs, health state utility weights, transition probabilities, and mortality
rates using a retrospective database analysis of commercially insured individuals and
the medical literature. Our primary outcome of interest was the incremental cost-
effectiveness ratio (ICER) of DT versus SOC evaluated at a base case willingness to
pay (WTP) threshold of $150,000 per quality-adjusted life year. We tested robustness of
115
our base case findings through scenario, deterministic sensitivity (DSA), and
probabilistic sensitivity analyses (PSA).
Results: In the base case analysis, treatment with DT was cost-effective with an ICER
of $140,877/QALY relative to SOC for a hypothetical cohort of 47% females. By gender,
both males (ICER of $135,574/QALY) and females (ICER of $146,511/QALY) were
similarly cost-effective to treat. In univariate DSA the base case ICER was most
sensitive to the costs of treating males, DT treatment cost, the discount rate, and the
health utility for children/adolescents with SCD. In PSA, DT was cost-effective in 32.7%,
66.0%, and 92.6% of 10,000 simulations at WTP values of $100,000, $150,000, and
$200,000 per QALY, respectively. A scenario analysis showed cost-effectiveness of DT
is highly contingent on assumed lifetime durability of the cure.
Conclusions: A hypothetical cell or gene therapy cure for SCD is likely to be cost-
effective from the US healthcare sector perspective. Large upfront costs of a single
administration cure are offset by significant downstream gains in health for patients
treated early in life. We find cost-effectiveness outcomes do not vary substantially by
gender; however, several model parameters including assumed durability and upfront
cost of DT are likely to influence cost-effectiveness findings.
116
Introduction
Sickle cell disease (SCD) is an inherited group of disorders characterized by the
presence of hemoglobin S. Individuals inherit the disease from parents that carry the
trait or may even exhibit the disease. The clinical manifestation of SCD involves
damaged and abnormally shaped red blood cells which result in hemolytic anemia,
small blood vessel obstruction, and other resulting complications.
1-3
The most common
form of SCD, sickle cell anemia (SCA), is also the most severe.
1
Informally, SCA is
often used interchangeably with SCD. SCD is associated with reduced life expectancy,
significant quality of life detriments, and increased healthcare resource utilization.
4
Severe pain episodes are common in many patients and often require costly
hospitalization.
5,6
There are an estimated 100,000 individuals currently living with SCD
in the United States.
7
Current treatment paradigms for SCD in the US vary by age group and generally
involve both palliative and preventive care. Adverse events include painful and
resource-intensive crises and other acute and chronic complications. Common
interventions for these adverse events supported by evidence include vaccination,
penicillin, hydroxyurea, blood transfusion, and opioids.
5
Few disease modifying
therapies are available. Hematopoietic stem cell transplant (HSCT) is a potential
curative treatment usually reserved for those with severe progressive disease despite
optimal medical management. However, HSCT is not widely available due to substantial
barriers to treatment, possible transplant rejection, and various long-term adverse
outcomes.
8,9
Two recent treatments voxelotor and crizanlizumab have been shown in
clinical trials to exhibit superior hemoglobin (Hb) response rate over placebo and reduce
117
the frequency of vaso-occlusive crises (VOC) over placebo, respectively.
10,11
These
trials were cited as evidence in US Food and Drug Administration (FDA) approvals for
each of these interventions.
12,13
In addition to an increasing number of treatments for chronic management of
disease-related symptoms, a gene therapy cure aimed at helping patients with SCD
produce normal red blood cells has shown promise, and may reach US markets within
years.
14
With this comes the potential to treat patients with SCD with a single or short-
term administration therapy and avoid disease progression for up to a lifetime, resulting
in downstream improvements in life expectancy, quality of life, and treatment cost. The
cost-effectiveness of a hypothetical cure for SCD from the US healthcare sector
perspective is unknown. In this study we estimate the cost-effectiveness of a
hypothetical treatment for SCD provided at birth with varying durability (ranging from 10
years to a lifetime) free of the disease and its related complications.
Methods
Study design
We built a cohort Markov model to conduct a cost-effectiveness analysis of a
hypothetical durable treatment for sickle cell disease relative to standard of care (SOC).
The model horizon was lifetime in yearly cycles with a 3% yearly discount on costs and
health outcomes. We conducted the analysis from the US healthcare sector
perspective. Costs included direct medical costs incurred by patients and providers and
our measure of health utility was quality-adjusted life years (QALYs). Our primary
outcome of interest was the incremental cost-effectiveness ratio (ICER) between
118
durable therapy and SOC. We evaluated this outcome against a willingness to pay
(WTP) threshold of $150,000 per QALY in 2018 US dollars; An ICER below our
threshold means the intervention is cost-effective. We tested robustness of our base
case through univariate, probabilistic, and scenario analyses. This included a value of
information (VOI) analysis to quantify the expected value of perfect information (EVPI)
surrounding treatment decisions.
Model
We created a decision analytic Markov model using Microsoft Excel, Visual Basic for
Applications (VBA) (Microsoft Corporation, Redmond, WA), and the R programming
language (R Foundation for Statistical Computing). Our model evaluates the cost-
effectiveness of a hypothetical durable treatment for SCD relative to a current SOC
bundle for affected patients in the US. We follow a theoretical cohort over a 100-year
horizon with one-year cycles to approximate discounted outcomes over a lifetime. As
recommended by the Second Panel on Cost-Effectiveness in Health and Medicine, our
healthcare sector perspective includes formal medical costs borne by payers and
patients.
15
This includes both current and future medical costs related and unrelated to
the management of SCD.
Treatment and management of SCD in the United States is complex and varies
based on disease severity, patient age, and setting. Despite recent FDA approvals for
non-curative treatments voxelotor and crizanlizumab for the management of SCD, the
real-world long-term effectiveness, uptake, and patient access to these medications are
unknown. For these reasons, we define the SOC as the typical treatment regimen and
119
disease management services received by subgroup in a sample of commercially
insured individuals between 2007 and 2017. SOC includes but is not limited to
treatment with: antibiotics, vaccinations, pain-relief medications, hydroxyurea, blood
transfusions, and stem cell transplants. This data-driven definition for SOC allows for
the treatment bundle cost to vary across age, gender, and disease severity levels.
Within this group we do not differentiate patients based on specific treatments received,
for instance patients treated with hydroxyurea versus those who are not.
For the durable treatment arm, we assume the intervention is a fully effective
single administration provided at birth. We define a “fully effective” cure as one that
completely suppresses disease-related complications and costs and restores life
expectancy and health-related quality of life (HRQoL) to that of a comparable individual
unaffected by the disease. Our analysis assumes a cure that is 100% effective across
predetermined time ranges (10 years, 20 years, and lifetime). We take an intention-to-
treat (ITT) approach in that patients unsuccessful on the durable treatment after any
period are assumed to move to treatment with the SOC bundle. For this reason, our
formal comparison can be stated as two distinct treatment strategies:
• Administration of durable treatment (DT) at birth, with subsequent standard of
care management after effectiveness period (henceforth “DT” arm)
• Standard of care management only (henceforth “SOC” arm)
Our model is informed by various data sources including direct medical costs from a
commercial claims retrospective database analysis and other data from the published
medical literature. Consistent with recommendations from the second panel on cost-
120
effectiveness in health and medicine, we discount health outcomes and costs at the
same rate of 3% per year.
15
An ICER is often used in health economic evaluation to determine the additional
resources necessary to obtain one additional year of perfect health, or a quality-
adjusted life year (QALY). In the case of two interventions, A and B, the ICER of
treatment A relative to B is defined as the ratio of average incremental costs incurred
under treatment A over the average incremental QALYs gained under treatment A, both
relative to treatment B.
𝐼𝐶𝐸 𝑅 𝐴 =
𝐶𝑂 𝑆𝑇
𝐴 − 𝐶𝑂𝑆 𝑇 𝐵 𝑄𝐴𝐿𝑌 𝑠 𝐴 − 𝑄𝐴𝐿𝑌 𝑠 𝐵 (1)
Our primary outcome of interest is the ICER between the DT and SOC treatment
strategies. We consider a treatment to be cost-effective if it generates an ICER of less
than $150,000 per QALY, which is between roughly two and three times the current US
gross domestic product (GDP) per capita.
16
Our reporting complies with the
Consolidated Health Economic Evaluation Reporting Standards (CHEERS) guidelines
where applicable.
17
Markov process approach
Markov models are commonly used mathematical tools to simulate recurring events
over an extended period of time. Markov models exhibit the “memoryless” property
which means transition probabilities depend only on the current state. However, it is
121
possible to create a Markov model in which rewards and transition probabilities vary
over time (for example by age) by creating a set of states for each point in time. We
employ this form of Markov model with annual cycles informed by data from a private
insurer and the published medical literature to estimate the lifetime direct cost and
health-related quality of life under management of SCD (Figure 5.1).
Figure 5.1 Markov process for lifetime management of SCD. Death (not pictured) can
occur from any state. Abbreviations: SCD, sickle cell disease; mild, mild disease (zero
crises per year); mod, moderate disease (<2 crises per year); sev, severe disease (≥ 2
crises per year).
Informing a discrete-time Markov process involves providing multiple model
inputs. In a health economic evaluation context, these include: health states, costs, a
measure of health outcome (life years [LYs], quality-adjusted life years [QALYs], events
avoided, etc.), and transition probabilities.
18
We outline our methods for obtaining each
of these below. A detailed summary of model inputs is available in Table 5.1.
122
Table 5.1 Model input parameters and probabilistic sensitivity analysis distributions
Parameter Value (95% CI)* Distribution (parameters) Source
Yearly discount rate 3 (2.4, 3.6) Uniform (a = 2.4, b = 3.6)
†
Assumption
Number of theoretical patients 10,000 Not varied in DSA or PSA Assumption
Percent female 51 (41.1, 60.9) Beta (α = 49, β = 47) Optum 2007-2017
Initial SCD severities, females
Percent mild 50.8 Optum 2007-2017
Percent moderate 22.1 Dirichlet (α₁ = 227, α₂ = 99, α₃ = 121) Optum 2007-2017
Percent severe 27.1
Optum 2007-2017
Initial SCD severities, males
Percent mild 48.3
Optum 2007-2017
Percent moderate 23.1 Dirichlet (α₁ = 224, α₂ = 107, α₃ = 133) Optum 2007-2017
Percent severe 28.7
Optum 2007-2017
Initial treatment assumptions, females
Percent mild treated 100 Not varied in PSA or DSA Assumption
Percent moderate treated 100 Not varied in PSA or DSA Assumption
Percent severe treated 100 Not varied in PSA or DSA Assumption
Initial treatment assumptions, males
Percent mild treated 100 Not varied in PSA or DSA Assumption
Percent moderate treated 100 Not varied in PSA or DSA Assumption
Percent severe treated 100 Not varied in PSA or DSA Assumption
Utility weights, females
Control QALY, ages 1-44 0.89 (0.87, 0.91) Beta (α = 870.42, β = 107.58)
19,20
Control QALY, ages 45-54 0.87 (0.85, 0.89) Beta (α = 983.1, β = 146.9)
19,20
Control QALY, ages 55-64 0.84 (0.82, 0.86) Beta (α = 1128.12, β = 214.88)
19,20
Control QALY, ages 65-74 0.84 (0.82, 0.86) Beta (α = 1128.12, β = 214.88)
19,20
Control QALY, ages 75+ 0.82 (0.8, 0.84) Beta (α = 1209.5, β = 265.5)
19,20
Utility weights, males
Control QALY, ages 1-44 0.89 (0.87, 0.91) Beta (α = 870.42, β = 107.58)
19,20
Control QALY, ages 45-54 0.88 (0.86, 0.9) Beta (α = 928.4, β = 126.6)
19,20
Control QALY, ages 55-64 0.86 (0.84, 0.88) Beta (α = 1034.58, β = 168.42)
19,20
Control QALY, ages 65-74 0.87 (0.85, 0.89) Beta (α = 983.1, β = 146.9)
19,20
Control QALY, ages 75+ 0.85 (0.83, 0.87) Beta (α = 1082.9, β = 191.1)
19,20
123
Utility weights, gender invariant
SCD QALY, ages 1-18 0.69 (0.57, 0.80) Normal (μ = 0.688, σ = 0.058)
§
20,21
SCD QALY, ages 19+ 0.68 (0.67, 0.69) Normal (μ = 0.682, σ = 0.005)
§
20,21
Transition probabilities, both genders
Yearly relapse probability 0.00 Not varied in PSA or DSA Assumption
Ordered logit transition probability regression coefficients, females
Moderate SCD (α₁) 1.07 (0.64, 1.51) Normal (μ = 1.074, σ = 0.223) Optum 2007-2017
Severe SCD (α₂) 2.63 (2.19, 3.06) Normal (μ = 2.625, σ = 0.22) Optum 2007-2017
Age (α₃) -0.03 (-0.03, -0.02) Normal (μ = -0.026, σ = 0.004) Optum 2007-2017
Moderate SCD * Age (α₄) 0.01 (0, 0.02) Normal (μ = 0.007, σ = 0.006) Optum 2007-2017
Severe SCD * Age (α₅) 0.02 (0.01, 0.03) Normal (μ = 0.021, σ = 0.005) Optum 2007-2017
Cut Point 1 (κ₁) 1.04 (0.76, 1.32) Normal (μ = 1.04, σ = 0.141) Optum 2007-2017
Cut Point 2 (κ₂) 1.86 (1.57, 2.15) Normal (μ = 1.858, σ = 0.147) Optum 2007-2017
Ordered logit transition probability regression coefficients, males
Moderate SCD (α₁) 0.7 (0.23, 1.17) Normal (μ = 0.701, σ = 0.24) Optum 2007-2017
Severe SCD (α₂) 1.98 (1.53, 2.42) Normal (μ = 1.976, σ = 0.225) Optum 2007-2017
Age (α₃) -0.03 (-0.04, -0.02) Normal (μ = -0.03, σ = 0.004) Optum 2007-2017
Moderate SCD * Age (α₄) 0.01 (0, 0.03) Normal (μ = 0.012, σ = 0.007) Optum 2007-2017
Severe SCD * Age (α₅) 0.03 (0.02, 0.05) Normal (μ = 0.034, σ = 0.007) Optum 2007-2017
Cut Point 1 (κ₁) 0.79 (0.49, 1.08) Normal (μ = 0.787, σ = 0.151) Optum 2007-2017
Cut Point 2 (κ₂) 1.6 (1.3, 1.91) Normal (μ = 1.605, σ = 0.158) Optum 2007-2017
Cost parameters
Cost of single administration durable
therapy
2.1M (1.68M,
2.52M)
Uniform (a = 1680000, b = 2520000) Assumption
GLM gamma log-link cost regression coefficients, females
Intercept (β₀) 8.05 (7.88, 8.22) Normal (μ = 8.051, σ = 0.085) Optum 2007-2017
Mild SCD (β₁) 1.58 (1.3, 1.85) Normal (μ = 1.575, σ = 0.139) Optum 2007-2017
Moderate SCD (β₂) 1.78 (1.48, 2.09) Normal (μ = 1.785, σ = 0.154) Optum 2007-2017
Severe SCD (β₃) 3.16 (2.88, 3.45) Normal (μ = 3.165, σ = 0.147) Optum 2007-2017
Age (β₄) 0.03 (0.03, 0.04) Normal (μ = 0.032, σ = 0.002) Optum 2007-2017
Mild SCD * Age (β₅) -0.01 (-0.02, -0.01) Normal (μ = -0.01, σ = 0.003) Optum 2007-2017
Moderate SCD * Age (β₆) -0.01 (-0.02, -0.01) Normal (μ = -0.014, σ = 0.003) Optum 2007-2017
Severe SCD * Age (β₇) -0.02 (-0.03, -0.02) Normal (μ = -0.023, σ = 0.003) Optum 2007-2017
124
GLM gamma log-link cost regression coefficients, males
Intercept (β₀) 8.39 (7.51, 9.28) Normal (μ = 8.395, σ = 0.452) Optum 2007-2017
Mild SCD (β₁) 2.13 (1.16, 3.09) Normal (μ = 2.126, σ = 0.494) Optum 2007-2017
Moderate SCD (β₂) 1.14 (0.21, 2.07) Normal (μ = 1.14, σ = 0.475) Optum 2007-2017
Severe SCD (β₃) 2.36 (1.46, 3.27) Normal (μ = 2.364, σ = 0.46) Optum 2007-2017
Age (β₄) 0.02 (0.01, 0.04) Normal (μ = 0.024, σ = 0.009) Optum 2007-2017
Mild SCD * Age (β₅) -0.02 (-0.03, 0) Normal (μ = -0.016, σ = 0.009) Optum 2007-2017
Moderate SCD * Age (β₆) 0.01 (-0.02, 0.03) Normal (μ = 0.006, σ = 0.011) Optum 2007-2017
Severe SCD * Age (β₇) -0.01 (-0.02, 0.01) Normal (μ = -0.005, σ = 0.009) Optum 2007-2017
*Confidence intervals omitted for variables drawn from multivariate distributions.
†
Not varied in PSA.
‡
Not varied in DSA.
§
Truncated normal distribution on [0,1]. Abbreviations: PSA, probabilistic sensitivity analysis; DSA, deterministic sensitivity
analysis; GLM, generalized linear model.
125
Health states
Our de novo decision model contains the following health states:
• Healthy, in remission
• Mild SCD, under SOC treatment
• Moderate SCD, under SOC treatment
• Severe SCD, under SOC treatment
• Dead
For all patients, we assume incidence of sickle cell disease is at birth. We also
assume disease-related complications begin to manifest immediately after birth. Given
SCD is an inherited genetic condition, these assumptions are consistent with the
physiology and subsequent manifestation of the disease.
22
The patient’s initial health
state is determined based on treatment arm and subsequent transitions depend on the
assumed durability of the single administration therapy, the probability of disease
progression, and the probability of death. Initial SCD severity for patients in the SOC
arm is based on the observed distributions of mild, moderate, and severe patients
observed in our data before age 10. In our base case we assume all patients in the DT
arm are treated regardless of gender or disease severity. Death is the only absorbing
state in the model.
Costs
Sickle cell disease is associated with substantial healthcare resource utilization and
events for patients managed in the US. Our analysis incorporates regression-estimated
126
direct costs of healthcare associated with managing SCD. For patients in the model
treated with SOC, we estimate direct healthcare costs across consecutive age ranges
from a retrospective analysis of a large commercial claims database that includes
Medicare Advantage patients (see Chapter 4 for additional details). For patients in the
model who entered with SCD but were cured and are hence “in remission” we estimate
direct costs using data from propensity score matched controls (matched on gender,
race, geographic division, year of birth, index year, plan characteristics, and education)
unaffected by the disease. Data on these patients are from the same database. To
adjust costs, we employ a generalized linear model (GLM) with a log-link and gamma
family, and hence our yearly cost estimates are defined by:
𝑌𝐸𝐴𝑅𝐿𝑌𝐶𝑂𝑆𝑇 ̂
= exp(𝛽 ̂
0
+ 𝛽 ̂
1
𝑆𝐶 𝐷 𝑀𝐼𝐿𝐷 + 𝛽 ̂
2
𝑆𝐶 𝐷 𝑀𝑂𝐷 + 𝛽 ̂
3
𝑆𝐶 𝐷 𝑆𝐸𝑉 + 𝛽 ̂
4
𝐴𝐺𝐸 + 𝛽 ̂
5
𝑆𝐶 𝐷 𝑀𝐼𝐿𝐷 ∗ 𝐴𝐺𝐸 + 𝛽 ̂
6
𝑆𝐶 𝐷 𝑀𝑂𝐷 ∗ 𝐴𝐺𝐸 + 𝛽 ̂
7
𝑆𝐶 𝐷 𝑆𝐸𝑉 ∗ 𝐴𝐺𝐸 )
Note that each gender is estimated separately, and we drop the i index for notational
convenience. See chapter 4 for further discussion on our cost estimation procedures.
For cured patients we also consider the direct cost of the durable treatment
administration. As there are few cell and gene therapies currently available, and none
for SCD, there is limited precedent on the potential cost of the product and
administration. The most recently approved curative therapy approved by the US Food
and Drug Administration for pediatric patients was in spinal muscular atrophy (SMA).
23
The product, Zolgensma
®
(onasemnogene abeparvovec-xioi), is an adeno-associated
127
(AAV9) virus vector-based gene therapy. It carries a list price of roughly $2.1 million
USD as of May 2019.
24
We use this price of $2.1M as a base case figure for a potential
gene therapy in SCD. We vary this price by ± 20% in deterministic (DSA) and
probabilistic sensitivity analyses (PSA).
Deceased patients are assumed to incur zero costs. We report all costs in 2018
USD, adjusted by the consumer price index (CPI) or medical component of the CPI
when necessary.
25
Future costs related or unrelated to the management of SCD are
accounted for through the cohort nature of the model. Every yearly cycle, the patients
age an additional year and face modified stage rewards (i.e. costs and QALYs)
representative of their current state. All future streams are discounted at a rate of 3%
per year. See Table 5.1 for a summary of cost inputs including GLM cost regression
coefficients by gender.
Health outcomes
Our primary health outcome of interest is average QALYs by treatment arm. We
generate QALYs by taking estimated life years lived in each health state and adjusting
by a factor within [0,1], also known as a health utility or QALY weight. For our study, we
use health utilities (QALY weights) from the published literature. Lubeck et al. (2019)
provides utilities which vary by age for both patients currently living with SCD and
comparable matched controls.
20
For patients in the “healthy, in remission” health state, we utilize age-specific
QALY weights provided in Lubeck et al. which were calculated in Fryback et al. (2007)
using the EuroQol-5 Dimensions (EQ-5D) instrument to represent a normative US
128
population.
19
For patients in any “affected, under SOC treatment” health state we apply
mean health utilities reported in Lubeck et al. for children/adolescents and adults.
20
The
authors generated these weights by mapping visual analog pain scale (VAS) scores for
patients with SCD to the EQ-5D.
21
We assign deceased patients a QALY weight of zero
and discount all future QALYs at a rate of 3% per year. See Table 5.1 for a summary of
utility inputs by treatment arm and age.
Transition probabilities
Patients begin in either the healthy, in remission state (DT arm) or in an affected, on
SOC treatment state (SOC arm); see illustration in Figure 5.1. Patients in the DT arm
continue in remission with a predetermined probability of experiencing relapse. Patients
that experience relapse transition to management under SOC with mild disease. These
patients may then experience disease progression or death but cannot return to the
remission state. In our base case where we assume the durable treatment is effective
for a lifetime, the probability of relapse is equal to zero.
Patients that begin in the SOC arm stay in an affected state, with zero probability
of moving to the healthy state. We use the initial distribution of mild, moderate, and
severe disease from a retrospective database analysis of commercially insured patients
to approximate disease severity in the first year of life. All patients in the model can die
after any cycle, with probabilities determined using simulated life tables for black or
African American individuals (healthy/remission state) or simulated SCD risk-adjusted
life tables (any affected state). We obtain these annual probabilities of death for both
arms from a lifetime simulation analysis of patients with and without SCD in the US by
129
Lubeck et al. (2019).
20
The probabilities of disease maintenance or progression are
conditional on survival and hence are applied only to patients surviving the cycle.
For patients with SCD, we estimated the probability of moving across disease
severities using observed transitions in our data. Using disease severity in years one
(severity1) and two (severity2) we regressed severity2 on severity1 and age, including
interactions. We ran these regressions separately by gender. Our severity
classifications are created using an underlying annualized count of vaso-occlusive
crises: mild disease (zero crises per year), moderate disease (>0 but <2 crises per
year), and severe disease (≥ 2 crises per year). For this reason, our health states are
ordinal and estimation using conditional logistic regression is appropriate.
26
This method
for estimating transition probabilities for ordinal health states has been described in
detail elsewhere.
27
130
Table 5.2.a Ordered logistic regression of second year severity, females
Variable Coeff SE z P-Val CI low CI high
Mild SCD (base) - - - - - -
Moderate SCD (α₁) 1.07 0.22 4.81 <0.0001 0.64 1.51
Severe SCD (α₂) 2.63 0.22 11.92 <0.0001 2.19 3.06
Age (α₃) -0.03 0.00 -7.35 <0.0001 -0.03 -0.02
Mild SCD * Age (base) - - - - - -
Moderate SCD * Age
(α₄)
0.01 0.01 1.25 0.2106 0.00 0.02
Severe SCD * Age
(α₅)
0.02 0.01 3.76 0.0002 0.01 0.03
Cut Point 1 (κ₁) 1.04 0.14 7.36 <0.0001 0.76 1.32
Cut Point 2 (κ₂) 1.86 0.15 12.63 <0.0001 1.57 2.15
Number of obs = 2,650
LR chi2(5) = 1203.85
Prob > chi2 = 0.0000
Pseudo R2 = 0.2679
Log likelihood = -1644.73
Abbreviations: Coeff, coefficient; SE, standard error; T-stat, T-statistic; P-Val; P-value;
CI, 95% confidence interval; SCD, sickle cell disease; obs, observations; LR, likelihood
ratio; prob, probability.
Table 5.2.b Ordered logistic regression of second year severity, males
Variable Coeff SE z P-Val CI low CI high
Mild SCD (base) - - - - - -
Moderate SCD (α₁) 0.70 0.24 2.92 0.0035 0.23 1.17
Severe SCD (α₂) 1.98 0.23 8.77 <0.0001 1.53 2.42
Age (α₃) -0.03 0.00 -6.99 <0.0001 -0.04 -0.02
Mild SCD * Age (base) - - - - - -
Moderate SCD * Age
(α₄)
0.01 0.01 1.73 0.0832 0.00 0.03
Severe SCD * Age
(α₅)
0.03 0.01 5.22 <0.0001 0.02 0.05
Cut Point 1 (κ₁) 0.79 0.15 5.20 <0.0001 0.49 1.08
Cut Point 2 (κ₂) 1.60 0.16 10.17 <0.0001 1.30 1.91
Number of obs = 1,727
LR chi2(5) = 728.33
Prob > chi2 = 0.0000
Pseudo R2 = 0.2334
Log likelihood = -1196.01
Abbreviations: Coeff, coefficient; SE, standard error; T-stat, T-statistic; P-Val; P-value;
CI, 95% confidence interval; SCD, sickle cell disease; obs, observations; LR, likelihood
ratio; prob, probability.
131
Using these regression estimates, we calculated the transition probabilities by
recovering predicted values for each age, disease severity, and gender:
𝑃 (𝑆𝑡𝑎𝑡𝑒 𝑀𝑖𝑙 𝑑 𝑇 +1
|𝑆𝑡𝑎𝑡𝑒 𝑖 𝑇 ) =
1
1 + exp(𝑍 ̂
𝑖 − 𝜅 ̂
1
)
(2)
𝑃 (𝑆𝑡𝑎𝑡𝑒 𝑀𝑜 𝑑 𝑇 +1
|𝑆𝑡𝑎𝑡𝑒 𝑖 𝑇 ) =
1
1 + exp(𝑍 ̂
𝑖 − 𝜅 ̂
2
)
−
1
1 + exp(𝑍 ̂
𝑖 − 𝜅 ̂
1
)
(3)
𝑃 (𝑆𝑡𝑎𝑡𝑒 𝑆𝑒 𝑣 𝑇 +1
|𝑆𝑡𝑎𝑡𝑒 𝑖 𝑇 ) = 1 −
1
1 + exp(𝑍 ̂
𝑖 − 𝑘 ̂
2
)
(4)
Where i is mild, moderate, or severe
and
𝑍 𝑖 = 𝛼 1
𝟏 𝑀𝑜𝑑 + 𝛼 2
𝟏 𝑆𝑒𝑣 + 𝛼 3
𝐴𝑔𝑒 + 𝛼 4
(𝟏 𝑀𝑜𝑑 ∗ 𝐴𝑔𝑒 ) + 𝛼 5
(𝟏 𝑆𝑒𝑣 ∗ 𝐴𝑔𝑒 )
See Appendix Figures 5.1.a, b, and c for base case transition probabilities by gender,
severity, and age.
Scenario and sensitivity analyses
In addition to our base case model in which we compare a fully effective durable
treatment over a lifetime relative to SOC, we conduct scenario and sensitivity analyses
to test robustness of our results. The primary scenario analyses involve variation in the
132
assumed effectiveness period of the DT. We consider scenarios in which the DT is
expected to last for 10 and 20 years for the median patient, after which effects dissipate
and disease progression returns to that observed under current SOC. To employ these
scenario analyses, we transform the 10- and 20-year probabilities to annual probability
of relapse (conditional on staying alive) assuming constant exponential rate. Our
equation is:
𝑃 𝐴𝑛𝑛𝑢𝑎𝑙 = 1 − (1 − 𝑃 𝑇 𝑦𝑒𝑎𝑟𝑠 )
1
𝑇 (5)
Hence for median 10-year duration,
𝑃 𝐴𝑛𝑛𝑢𝑎𝑙 = 1 − (1 − 0.50)
1
10
= 0.06697
and for 20-years
𝑃 𝐴𝑛𝑛𝑢𝑎𝑙 = 1 − (1 − 0.50)
1
20
= 0.03406
In these scenarios the annual probabilities of relapse to mild SCD for patients in the DT
arm are 0.06697 and 0.03406, respectively.
In univariate deterministic sensitivity analysis (DSA) we vary each input
parameter individually, holding all others constant. We vary parameters within their 95%
confidence intervals, or by ± 20% when unavailable. We evaluate impact on the ICER of
DT relative to SOC and report results in a tornado diagram. Using the results from
univariate DSA, we find the top three nongendered parameters to which the model is
most sensitive and vary them by pairs in two-way sensitivity analyses.
133
In probabilistic sensitivity analysis we fit likely probability distributions to each
parameter and conduct 10,000 iterations in a Monte Carlo simulation. Each iteration
uses a vector of K independent draws, one for each of the K input variable’s respective
distribution and evaluates them in the model. Following the simulation, we count the
number of iterations for which the DT ICER is less than our threshold of $150,000 per
QALY and divide by 10,000. We consider this value the percent of the time DT is cost-
effective. In addition, we use the results of PSA to compute and graph the cost-
effectiveness acceptability curve (CEAC), frontier (CEAF), and the EVPI.
28,29
The EVPI is a common metric in VOI analysis. It is defined as the difference
between expected net benefit (NB) under perfect information and the expected NB
under current information (or simply expected monetary value). This can also be written
as:
𝐸𝑉𝑃𝐼 = 𝐸 𝜃 max
𝑗 𝑁𝐵 (𝑗 , 𝜃 ) − max
𝑗 𝐸 𝜃 𝑁𝐵 (𝑗 , 𝜃 )
Where j is treatment given current information 𝜃 .
In the context of health care, it measures the amount a healthcare system would be
willing to pay to eliminate uncertainty regarding a treatment decision. For our purposes,
we calculate EVPI to understand the additional cost related to treatment decision
uncertainty for a range of willingness to pay values for a QALY.
Human subjects approval and data analysis
Our database analysis component was limited to de-identified data not collected for the
purposes of the study. For this reason, the proposal for that study was approved by the
University of Southern California’s University Park Institutional Review Board (UPIRB).
134
Our cost-effectiveness study does not utilize human subjects or additional protected
health information. Our retrospective database analysis and statistical procedures were
conducted in SAS 9.4 software (SAS Institute, Raleigh, NC) and Stata 16.0 software
(StataCorp, College Station, TX), respectively. The Markov model and Monte Carlo
simulations were conducted in Excel and VBA (Microsoft Corporation, Redmond, WA).
Graphs were created and our VOI analyses were conducted using the R programming
language (R Foundation for Statistical Computing).
Results
Base case and scenario analysis results
In the base case analysis, we find treatment with DT results in an ICER of $140,877 per
QALY and is hence cost-effective at our predetermined WTP threshold of $150K per
QALY. DT generated an average of 26.4 discounted QALYs at a discounted cost of
$2,372,482 per-patient. SOC resulted in 17.9 discounted QALYs at a discounted cost of
$1,175,566 per-patient. Base case results did not differ notably by gender. Females
gained an additional 0.6 discounted incremental QALYs relative to males (8.8 vs 8.2)
but incurred greater discounted incremental costs of $156,289 over a lifetime relative to
males after being treated with DT ($1,279,485 vs $1,123,195). DT ICERs among
females and males separately were $146,511 per QALY and $135,574 per QALY,
respectively. The health state movements of our theoretical patients are shown in
Figure 5.2 and detailed base case and scenario analysis results are reported in Table
5.3.
135
Our scenario analyses involved loosening the assumption that DT will last a
lifetime for all patients treated. In our median 20-year duration scenario the health
outcomes of DT dropped to an average of 21.8 QALYs
xxiii
gained with an increase in
cost
xxiv
to $2,761,601 due to unsuccessful patients moving to treatment with SOC. The
ICER
xxv
in this scenario was $410,607 per QALY. A scenario in which DT lasts a
median of 10 years further decreases QALYs, increases costs, and results in an ICER
of $740,058 per QALY. We do not consider DT to be cost-effective in either of these
scenarios.
xxiii
Discounted at 3% per year.
xxiv
Ibid.
xxv
Ibid.
136
Table 5.3 Discounted and undiscounted base case and scenario analysis results, per-patient
DT SOC
Discounted Results* ICER
†
Cost QALY LY Cost QALY LY
Base case $140,877 $2,372,482 26.4 29.9 $1,175,566 17.9 26.2
Females $146,511 $2,377,583 26.8 30.3 $1,098,098 18.0 26.3
Males $135,574 $2,367,928 26.1 29.5 $1,244,733 17.9 26.1
DT lasts 20-years
‡
$410,607 $2,761,601 21.8 27.2 $1,175,566 17.9 26.2
Females $409,161 $2,703,690 22.0 27.4 $1,098,098 18.0 26.3
Males $411,937 $2,813,307 21.7 27.0 $1,244,733 17.9 26.1
DT lasts 10-years
‡
$740,058 $2,914,175 20.3 26.5 $1,175,566 17.9 26.2
Females $730,453 $2,836,566 20.4 26.7 $1,098,098 18.0 26.3
Males $748,848 $2,983,469 20.2 26.4 $1,244,733 17.9 26.1
Undiscounted Results
Base case $15,332 $3,210,182 66.2 75.8 $2,770,348 37.6 54.9
Females $24,825 $3,364,554 68.5 78.9 $2,607,231 38.0 55.6
Males $5,778 $3,072,350 64.2 73.0 $2,915,988 37.2 54.3
DT lasts 20-years
‡
$162,116 $4,154,316 46.1 59.6 $2,770,348 37.6 54.9
Females $163,939 $4,054,067 46.8 60.7 $2,607,231 38.0 55.6
Males $160,380 $4,243,825 45.4 58.7 $2,915,988 37.2 54.3
DT lasts 10-years
‡
$419,576 $4,411,570 41.5 56.3 $2,770,348 37.6 54.9
Females $411,772 $4,257,316 42.0 57.1 $2,607,231 38.0 55.6
Males $426,872 $4,549,296 41.0 55.6 $2,915,988 37.2 54.3
*All outcomes discounted at 3% per year.
†
ICER units are 2018 US dollars per QALY.
‡
Expected median duration.
Abbreviations: DT, durable therapy; SOC, standard of care; ICER, incremental cost-effectiveness ratio; QALY, quality-
adjusted life year; LY, life year.
137
Figure 5.2 Markov trace for SOC and DT arms (N=10,000 theoretical patients).
Abbreviations: SOC, standard of care; DT, durable therapy; SCD, sickle cell disease.
Deterministic sensitivity analysis
In DSA we first varied parameters one at a time within their 95% confidence intervals (±
20% when unavailable) to study impact on the base case ICER ($140,877/QALY). We
find the model to be most sensitive to variations in two of the coefficients in the cost
regression of treating males, DT treatment cost, the discount rate, and the QALY weight
for children and adolescents with SCD. Univariately, the model is most sensitive to the
regression coefficients 𝛽 ̂
0
and 𝛽 ̂
1
in the GLM regressions of cost on observable
characteristics for males. Decreasing (increasing) 𝛽 ̂
0
within its 95% CI results in an
ICER for DT of $176,589 ($54,207) per QALY. Similarly, decreasing (increasing) 𝛽 ̂
1
within its 95% CI results in an ICER for DT of $169,491 ($65,582) per QALY. These
results are intuitive – all else equal, increasing 𝛽 ̂
0
increases the predicted cost gap
between male control patients and male patients with SCD, of any severity. Similarly,
138
increasing 𝛽 ̂
1
increases the cost gap between male control patients and specifically
male patients with mild SCD. In both scenarios the cost-effectiveness of a durable cure
relative to standard of care increases.
Decreasing (increasing) the DT treatment cost (base value $2.1M USD) by 20
percent resulted in an ICER for DT of $91,443/QALY ($190,311/QALY). Changes in the
DT price result in pronounced impact on the model ICER due to the nature of upfront
payment; in the model patients are treated immediately and hence the cost is not
subject to discounting over future periods. Similarly, decreasing (increasing) the
discount rate (base value 3%) by 20 percent resulted in an ICER for DT of
$105,118/QALY ($181,407/QALY). This result suggests that a health system that is
increasingly myopic is less likely to find DT to be cost-effective. This is because single
administration costs are upfront while benefits are disproportionately longer term. Lastly,
setting the SCD QALY weight for children/adolescents at its lower and upper 95% CI
bounds resulted in ICERs for DT of $118,559 per QALY and $173,546 per QALY,
respectively. All else equal, patients with SCD who have their condition managed well
may be less cost-effective to treat than more severe cases. This is especially true for
children and adolescents as health gains during the first two decades of life are
discounted the least in the model. A tornado diagram of the top ten parameters in DSA
is shown in Figure 5.3. See Appendix Figure 5.4 for a complete tornado diagram and
Appendix Table 5.1 for a corresponding table of DSA results including variation ranges.
139
Figure 5.3 Tornado diagram of deterministic sensitivity analysis results (top ten).
Parameters were varied univariately within their 95% CIs, or ± 20% when unavailable.
Abbreviations: CI, confidence interval; SOC, standard of care; DT, durable therapy;
SCD, sickle cell disease; Yr, year; QALY, quality-adjusted life year; GLM, generalized
linear model; M, males; F, females; O. Logit, ordered logistic regression; Cont., control
patients; Prop., proportion.
Using the results from univariate DSA, we identified the top three non-gender
specific parameters to which the model ICER was most sensitive. These include the
upfront cost of DT, the discount rate, and the QALY weight for children and adolescents
with SCD. We then varied these in three different two-way deterministic sensitivity
analyses: DT price versus discount rate, DT price versus pediatric/adolescent QALY
weight, and discount rate versus pediatric/adolescent QALY weight. Findings were
similar in all three comparisons. Slight simultaneous increases in these parameters
often resulted in an ICER for DT between $150,000 and $200,000/QALY. In all three
140
two-way DSAs, DT was cost-effective at WTP of $250K or higher per QALY regardless
of parameter values. See Figure 5.4 for detailed graphs of two-way DSA.
141
Figure 5.4 Two-way deterministic sensitivity analyses of select model parameters. (A) DT upfront cost versus yearly
discount rate; (B) DT upfront cost versus pediatric QALY weight; (C) Yearly discount rate versus pediatric QALY weight.
Parameters were varied simultaneously within their 95% CIs, or ± 20% when unavailable. Abbreviations: DT, durable
therapy; QALY, quality-adjusted life year; CI, confidence interval; CE, cost-effective; WTP, willingness to pay; K,
thousand.
142
Probabilistic sensitivity analysis and value of information
Our primary outcome of interest in PSA was the ICER for DT relative to SOC. In 10,000
iterations, DT was cost-effective relative to SOC 13.6%, 32.7%, 66.0%, and 92.6% of
the time at WTP values of $50K, $100K, $150K, and $200K/QALY, respectively (Figure
5.5). Across genders, DT was much more likely to be cost-effective at $100K WTP per
QALY for males (42.5%) than for females (11.1%). See Appendix Figures 5.5.a and b
for gender-specific results.
In value of information analysis the expected value of perfect information (EVPI)
for the average patient is highest at the WTP associated with the highest decision
uncertainty. In our model, this occurs at a WTP per QALY of $115K where both arms
have an equal probability of being cost-effective. At this WTP, the EVPI is $224.1K
suggesting this is the most a healthcare system should be willing to pay to eliminate
model and parameter uncertainty. At our predetermined WTP threshold of
$150K/QALY, the EVPI is $88.9K per patient (Figures 5.4a and b).
143
Figure 5.5 Probabilistic sensitivity analysis results (A) Cost-effectiveness acceptability
curves (CEAC) and frontier (CEAF); (B) Expected value of perfect information (EVPI).
144
Discussion and Limitations
Based on currently available information on the economic and health-related quality of
life burdens, our findings suggest that a durable treatment or cure for sickle cell disease
would be cost-effective to the US healthcare sector. Assuming a $2.1 million USD all-
inclusive price for treatment, the average patient with SCD cured at birth stands to gain
8.5 QALYs and 3.7 LYs over a lifetime, at an incremental cost of $1,196,917 (all values
discounted). This results in a base case ICER from our model of $140,877 per QALY for
an initial distribution of 47% females (51% mild SCD, 22% moderate SCD, 27% severe
SCD), and 53% males (48% mild SCD, 23% moderate SCD, 29% severe SCD). These
initial gender and severity distributions were approximated using our data.
We do not find substantial differences in base case cost-effectiveness results
across subgroups. The ICERs for curing females and males in our base case are
$146,511 per QALY and $135,574 per QALY, respectively. However, in scenario
analyses, we find that cost-effectiveness is highly contingent on the single
administration therapy being fully effective for the lifetime of the patient. For a scenario
in which patients have a 50% chance of relapse (conditional on survival) by 20 years,
the overall ICER increases to $410,607 per QALY. If we change this value to 10 years,
the ICER further increases to $740,058 per QALY.
Our sensitivity analysis results suggest the likelihood DT is not cost-effective
decreases dramatically beyond willingness to pay of $250K/QALY. In univariate DSA,
the largest ICER we obtained was $190,311 per QALY from increasing the durable
treatment price to $2.52 million USD. In two-way DSA we find that slight simultaneous
increases our baseline discount rate and upfront cost of DT would result in an ICER for
145
DT of over $150K/QALY; however, no comparisons resulted in an ICER above
$250K/QALY. The same result holds true for pairwise comparisons of upfront cost of DT
versus the pediatric/adolescent SCD QALY weight and discount rate versus the
pediatric/adolescent SCD QALY weight. In PSA, DT was cost-effective 32.7%, 66.0%,
and 92.6% of the time at WTP values of $100K, $150K, and $200K/QALY, respectively.
Through VOI analysis we find that a health system would be willing to pay up to
approximately $224.1K per-patient to eliminate all uncertainty regarding durable
treatment decisions.
Our cost-effectiveness study is not without limitations. We assume treatment for
the durable therapy arm is provided at birth, which may not align with true eligible
population. Given the treatment is hypothetical, it is unclear who will qualify for the
treatment and subsequently receive it. In addition, we assume the cure is single
administration, and patients cannot receive it twice. A recently FDA approved gene
therapy in SMA, onasemnogene abeparvovec-xioi, is a one-time administration through
IV available to children younger than two years.
30
We assume the treatment is 100%
effective during the durability period. This period is lifetime in our base case, but we
vary it between a median of 10 and 20 years in our scenario analyses.
The data on costs in our study are appropriate for our purposes but carry
limitations. The direct costs are estimated from a retrospective database analysis of the
commercially insured, and do not vary based on time from death. They are based on
estimated allowed amounts, or ultimately what the insurer agrees to pay for services. A
societal perspective analysis would include other direct costs not borne by the
healthcare sector. Our assumed cost for a single administration durable therapy is
146
based on the most recently available comparable product. We use Zolgensma
®
(onasemnogene abeparvovec-xioi), a gene therapy approved for the treatment of
pediatric SMA, which carries a list price of $2.1 million USD. This price may not account
for discounts, rebates, or other costs associated with treatment administration.
In addition, the use of Zolgensma’s price may complicate our analysis if it was
determined by market forces and the presumptive value it brings to patients with SMA.
To address this, we note that an independent review by the Institute for Clinical and
Economic Review (ICER) found that Zolgensma’s incremental cost-effectiveness
ratio
xxvi
for treating SMA type I would be roughly $243,000 per QALY at a hypothetical
placeholder price of $2.0 million USD.
31
If we assume the current market price was
informed by this analysis, a $2.1 million USD price tag (+$100,000 to their placeholder
price), would imply the price was justified based on an ICER of roughly $252,000 per
QALY
xxvii
. For our purposes, an ICER of $252,000 per QALY would correspond with a
single administration total cost of $3.04 million USD for a durable SCD cure; however,
we would not consider the therapy to be cost-effective at this implied price point.
We do not incorporate indirect costs of managing SCD. These may include costs
such as uncompensated caregiver burden, lost income to presenteeism or
absenteeism, and other indirect costs borne by the healthcare or other sectors.
However, indirect costs are likely to be larger under SOC than in the DT arm due to the
well-documented humanistic burden associated with managing SCD. This means our
ICER likely underestimates the true benefit of a durable therapy; we would expect
xxvi
Relative to “best supportive care (BSC).”
xxvii
Zolgensma generated 11.77 QALYs at an increased cost of $2.87M USD over BSC. If price
increases $100K, the new ICER for Zolgensma is $2.97M/11.77 QALYs or approximately
$252K/QALY.
147
significant productivity benefits from the cure of a chronic disease. In addition, we would
also expect greater lifetime income from DT due to longer lifespan and additional
working years.
Our base case analysis uses a 3% discount rate on both costs and health
outcomes, a number based on the estimated real consumption rate of interest. This
discounting procedure is recommended by the Second Panel on Cost Effectiveness in
Health and Medicine.
15
However, appropriate methods for discounting in health
economic evaluation are still a subject of debate.
32,33
As confirmed in our case using
sensitivity analysis results, a high discount rate has a disproportionate impact on the
incremental cost-effectiveness ratio of therapies for which (i) the majority of costs are
upfront and (ii) the majority of health benefits are downstream. Despite a likely large
upfront cost, a durable therapy for SCD would provide significant health gains for
patients over the course of an extended lifespan. We acknowledge the impact discount
rate has on study results and for this reason also report undiscounted costs, life years,
quality-adjusted life years, and ICERs for additional interpretation.
Conclusions
We used a decision analytic model to explore the cost-effectiveness of a hypothetical
durable cure for sickle cell disease in the US. Our base case model suggests treatment
is likely cost-effective, contingent on durability of the cure. In probabilistic sensitivity
analysis durable therapy is cost effective 32.7%, 66.0%, and 92.6% of the time at WTP
values of $100,000, $150,000, and $200,000 per QALY, respectively. We find durable
treatment would be cost-effective at a minimum willingness to pay (WTP) of $150,000
148
per QALY at single administration costs of $713K, $1.09M, and $2.18M, for cure
durations of median 10 years, median 20 years, and lifetime, respectively. We
acknowledge our findings may not be applicable to all patients affected by SCD in the
US. There is substantial heterogeneity in SCD manifestation and subsequent treatment.
Despite this, understanding potential uptake and impacts prior to treatment availability is
difficult. Like HSCT, gene therapy treatment may be reserved for the most severe
patients who have failed numerous lines of previous therapies. We build our model
based on available data with the potential to provide refined estimates for specific
groups as a cure becomes likely. We hope additional data on the economic and clinical
burdens of managing sickle cell disease in the US are forthcoming.
149
References
1. Rees DC, Williams TN, Gladwin MT. Sickle-cell disease. Lancet.
2010;376(9757):2018-2031.
2. Sundd P, Gladwin MT, Novelli EM. Pathophysiology of Sickle Cell Disease. Annu
Rev Pathol. 2019;14:263-292.
3. Vekilov PG. Sickle-cell haemoglobin polymerization: is it the primary pathogenic
event of sickle-cell anaemia? Br J Haematol. 2007;139(2):173-184.
4. Platt OS, Brambilla DJ, Rosse WF, et al. Mortality in sickle cell disease. Life
expectancy and risk factors for early death. N Engl J Med. 1994;330(23):1639-
1644.
5. National Heart, Lung, and Blood Institute. Evidence-based management of sickle
cell disease: expert panel report, 2014. Washington, DC: National Institutes of
Health. 2014.
6. Platt OS, Thorington BD, Brambilla DJ, et al. Pain in sickle cell disease. Rates
and risk factors. N Engl J Med. 1991;325(1):11-16.
7. Hassell KL. Population estimates of sickle cell disease in the U.S. Am J Prev
Med. 2010;38(4 Suppl):S512-521.
8. Kapoor S, Little JA, Pecker LH. Advances in the Treatment of Sickle Cell
Disease. Mayo Clin Proc. 2018;93(12):1810-1824.
9. Kassim AA, Sharma D. Hematopoietic stem cell transplantation for sickle cell
disease: The changing landscape. Hematol Oncol Stem Cell Ther.
2017;10(4):259-266.
150
10. Vichinsky E, Hoppe CC, Ataga KI, et al. A Phase 3 Randomized Trial of
Voxelotor in Sickle Cell Disease. N Engl J Med. 2019;381(6):509-519.
11. Ataga KI, Kutlar A, Kanter J, et al. Crizanlizumab for the Prevention of Pain
Crises in Sickle Cell Disease. N Engl J Med. 2017;376(5):429-439.
12. U.S. Food and Drug Administration. FDA approves voxelotor for sickle cell
disease. 2019; https://www.fda.gov/drugs/resources-information-approved-
drugs/fda-approves-voxelotor-sickle-cell-disease. Accessed December 2, 2019.
13. U.S. Food and Drug Administration. FDA approves first targeted therapy to treat
patients with painful complication of sickle cell disease. 2019;
https://www.fda.gov/news-events/press-announcements/fda-approves-first-
targeted-therapy-treat-patients-painful-complication-sickle-cell-disease.
Accessed December 2, 2019.
14. Rubin R. Gene Therapy for Sickle Cell Disease Shows Promise. JAMA.
2019;321(4):334-334.
15. Neumann PJ, Sanders GD, Russell LB, Siegel JE, Ganiats TG. Cost
effectiveness in health and medicine. Second edition. ed. Oxford ; New York:
Oxford University Press; 2017.
16. Neumann PJ, Cohen JT, Weinstein MC. Updating cost-effectiveness--the curious
resilience of the $50,000-per-QALY threshold. N Engl J Med. 2014;371(9):796-
797.
17. Husereau D, Drummond M, Petrou S, et al. Consolidated Health Economic
Evaluation Reporting Standards (CHEERS)--explanation and elaboration: a
151
report of the ISPOR Health Economic Evaluation Publication Guidelines Good
Reporting Practices Task Force. Value Health. 2013;16(2):231-250.
18. Denton BT. Handbook of healthcare operations management : methods and
applications. New York: Springer; 2013.
19. Fryback DG, Dunham NC, Palta M, et al. US norms for six generic health-related
quality-of-life indexes from the National Health Measurement study. Med Care.
2007;45(12):1162-1170.
20. Lubeck D, Agodoa I, Bhakta N, et al. Estimated Life Expectancy and Income of
Patients With Sickle Cell Disease Compared With Those Without Sickle Cell
Disease. JAMA Netw Open. 2019;2(11):e1915374.
21. Anie KA, Grocott H, White L, Dzingina M, Rogers G, Cho G. Patient self-
assessment of hospital pain, mood and health-related quality of life in adults with
sickle cell disease. BMJ Open. 2012;2(4).
22. Hall JB, Kress J, Schmidt GA. Principles of Critical Care, 4th edition. McGraw-Hill
Education; 2015.
23. U.S. Food and Drug Administration. FDA approves innovative gene therapy to
treat pediatric patients with spinal muscular atrophy, a rare disease and leading
genetic cause of infant mortality. 2019; https://www.fda.gov/news-events/press-
announcements/fda-approves-innovative-gene-therapy-treat-pediatric-patients-
spinal-muscular-atrophy-rare-disease. Accessed January 17, 2020.
24. Rosenmayr-Templeton L. Industry update for May 2019. Ther Deliv.
2019;10(9):555-561.
152
25. U.S. Bureau of Labor Statistics. Medical care in U.S. city average. CPI-All Urban
Consumers (Current Series) 2019;
https://data.bls.gov/timeseries/CUUR0000SAM?output_view=pct_12mths.
Accessed November 30, 2019.
26. Cameron AC, Trivedi PK. Microeconometrics using Stata. College Station, Tex.:
Stata Press; 2010.
27. Jung J. Estimating Markov transition probabilities between health states in the
HRS dataset. 2006.
28. van Hout BA, Al MJ, Gordon GS, Rutten FF. Costs, effects and C/E-ratios
alongside a clinical trial. Health Econ. 1994;3(5):309-319.
29. Fenwick E, Claxton K, Sculpher M. Representing uncertainty: the role of cost-
effectiveness acceptability curves. Health Econ. 2001;10(8):779-787.
30. Zolgensma - one-time gene therapy for spinal muscular atrophy. Med Lett Drugs
Ther. 2019;61(1577):113-114.
31. Pearson SD, Thokala P, Stevenson M, Rind D. The Effectiveness and Value of
Treatments for Spinal Muscular Atrophy: A Summary from the Institute for
Clinical and Economic Review’s New England Comparative Effectiveness Public
Advisory Council. Journal of managed care & specialty pharmacy.
2019;25(12):1300-1306.
32. Attema AE, Brouwer WBF, Claxton K. Discounting in Economic Evaluations.
Pharmacoeconomics. 2018;36(7):745-758.
153
33. Claxton K, Paulden M, Gravelle H, Brouwer W, Culyer AJ. Discounting and
decision making in the economic evaluation of health-care technologies. Health
Econ. 2011;20(1):2-15.
154
Appendix
Appendix Table 5.1 Univariate deterministic sensitivity analysis results on ICER,
ordered by base case model sensitivity
Parameter -% +% Range Base Variation
GLM β0, M $176,589 $54,207 $122,381 8.39 10.56
GLM β1, M $169,491 $65,582 $103,909 2.13 45.52
DT, Admin. Cost $91,443 $190,311 $98,868 2100000
.00
20.00
Discount Rate, Yr $105,118 $181,407 $76,289 0.03 20.00
QALY, SCD, 1-18 $118,559 $173,546 $54,988 0.69 16.64
GLM β3, M $156,301 $102,850 $53,451 2.36 38.17
GLM β5, M $107,394 $156,913 $49,518 -0.02 115.96
GLM β4, M $157,938 $111,699 $46,240 0.02 71.94
GLM β7, M $127,321 $148,280 $20,959 -0.01 337.68
GLM β3, F $148,670 $130,489 $18,181 3.16 9.08
GLM β0, F $147,896 $132,580 $15,315 8.05 2.08
GLM β1, F $146,900 $132,974 $13,926 1.58 17.25
O. Logit κ2, F $134,389 $146,335 $11,946 1.86 15.52
GLM β2, M $144,086 $132,739 $11,347 1.14 81.61
O. Logit α2, F $144,920 $135,702 $9,218 2.63 16.44
QALY, M, Cont., 1-44 $145,361 $136,662 $8,699 0.89 2.29
GLM β5, F $136,116 $144,633 $8,518 -0.01 50.38
O. Logit κ2, M $136,338 $144,504 $8,166 1.60 19.28
QALY, F, Cont., 1-44 $144,902 $137,070 $7,832 0.89 2.29
GLM β7, F $136,636 $144,344 $7,708 -0.02 29.32
O. Logit α3, F $137,182 $143,779 $6,597 -0.03 26.66
GLM β6, M $142,842 $136,344 $6,499 0.01 343.60
GLM β4, F $143,522 $138,015 $5,507 0.03 10.85
O. Logit α3, M $137,732 $142,873 $5,141 -0.03 28.05
O. Logit α2, M $142,660 $138,343 $4,317 1.98 22.35
O. Logit α1, F $142,670 $138,586 $4,084 1.07 40.79
O. Logit α5, F $142,666 $138,646 $4,020 0.02 52.08
QALY, SCD, 19+ $138,979 $142,828 $3,850 0.68 1.39
O. Logit α5, M $142,183 $138,969 $3,215 0.03 37.55
O. Logit κ1, F $139,200 $142,376 $3,176 1.04 26.64
GLM β2, F $142,038 $139,307 $2,731 1.78 16.92
Initial Prop. Female $139,832 $141,912 $2,081 0.47 20.12
O. Logit α4, F $141,644 $139,907 $1,737 0.01 156.54
O. Logit κ1, M $141,738 $140,051 $1,687 0.79 37.70
O. Logit α1, M $141,451 $139,991 $1,460 0.70 67.13
GLM β6, F $140,188 $141,428 $1,241 -0.01 45.37
O. Logit α4, M $141,320 $140,178 $1,142 0.01 113.13
QALY, M, Cont., 45-54 $141,263 $140,493 $770 0.88 2.30
QALY, F, Cont., 45-54 $141,235 $140,521 $714 0.87 2.32
QALY, M, Cont., 55-64 $141,137 $140,618 $519 0.86 2.34
155
QALY, F, Cont., 55-64 $141,126 $140,629 $497 0.84 2.39
QALY, F, Cont., 65-74 $141,040 $140,714 $326 0.84 2.39
QALY, M, Cont., 65-74 $141,034 $140,720 $314 0.87 2.32
QALY, F, Cont., 75+ $141,011 $140,743 $268 0.82 2.44
QALY, M, Cont., 75+ $140,980 $140,774 $206 0.85 2.37
Base case ICER was $140,877/QALY. ICER units are dollars per QALY. Parameters
were varied univariately within their 95% CIs, or ± 20% when unavailable.
Abbreviations: ICER, incremental cost-effectiveness ratio; CI, confidence interval; SOC,
standard of care; DT, durable therapy; SCD, sickle cell disease; Yr, year; QALY, quality-
adjusted life year; GLM, generalized linear model; M, males; F, females; O. Logit,
ordered logistic regression; Cont., control patients; Prop., proportion.
156
Appendix Figure 5.1.a Ordered logit predicted probabilities of transitioning to mild SCD by gender, age, and state
Abbreviations: SCD, sickle cell disease.
157
Appendix Figure 5.1.b Ordered logit predicted probabilities of transitioning to moderate SCD by gender, age, and state
Abbreviations: SCD, sickle cell disease.
158
Appendix Figure 5.1.c Ordered logit predicted probabilities of transitioning to severe SCD by gender, age, and state
Abbreviations: SCD, sickle cell disease.
159
Appendix Figure 5.2.a Markov trace for SOC and DT arms, females
Abbreviations: SOC, standard of care; DT, durable therapy; SCD, sickle cell disease.
160
Appendix Figure 5.2.b Markov trace for SOC and DT arms, males
Abbreviations: SOC, standard of care; DT, durable therapy; SCD, sickle cell disease.
161
Appendix Figure 5.3.a Monte Carlo draws (N=10,000), Cost parameters and coefficients for GLM log-link gamma family
annualized total cost regressions
Abbreviations: GLM, generalized linear model; B, beta coefficient; M, males; F, females; DT, durable therapy; MIL,
millions (2018 USD).
162
Appendix Figure 5.3.b Monte Carlo draws (N=10,000), Utility parameters
Abbreviations: QALY, quality-adjusted life year; F, females; CONT, control patients, M, males; SCD, sickle cell disease.
163
Appendix Figure 5.3.c Monte Carlo draws (N=10,000), Initial condition and transition probability parameters: coefficients
for ordered logit transition probability regressions.
Abbreviations: A, alpha coefficient; F, females; M, males; mod, moderate; sev, severe.
164
Appendix Figure 5.4 Tornado diagram of deterministic sensitivity analysis results
Parameters were varied univariately within their 95% CIs, or ± 20% when unavailable. Abbreviations: CI, confidence
interval; SOC, standard of care; DT, durable therapy; SCD, sickle cell disease; Yr, year; QALY, quality-adjusted life year;
GLM, generalized linear model; M, males; F, females; O. Logit, ordered logistic regression; Cont., control patients; Prop.,
proportion.
165
Appendix Figure 5.5.a PSA results among females (A) Cost-effectiveness acceptability
curves (CEAC) and frontier (CEAF); (B) Expected value of perfect information (EVPI)
Abbreviations: PSA, probabilistic sensitivity analysis; QALY, quality-adjusted life year.
166
Appendix Figure 5.5.b PSA results among males (A) Cost-effectiveness acceptability
curves (CEAC) and frontier (CEAF); (B) Expected value of perfect information (EVPI)
Abbreviations: PSA, probabilistic sensitivity analysis; QALY, quality-adjusted life year.
167
Chapter 6
Conclusions and Policy Implications
Conclusions
In this dissertation we presented a current overview of the economics of sickle cell
disease (SCD) in the United States. We find that the literature on healthcare resource
utilization and costs for the management of SCD is limited, with no current estimates for
patients of all ages covered by commercially insured plans. We then surveyed methods
for determining a lifetime cost estimate for managing the condition and subsequently to
determine a counterfactual under a case where the condition is cured at birth. In our
retrospective database study, we find that the current direct healthcare-related costs of
managing SCD in a commercially insured are substantial and greatly exceed previous
estimates in a Medicaid population.
1
Lastly, through a cost-effectiveness analysis we show that a hypothetical cell or
gene therapy cure would be cost-effective from the US healthcare sector perspective. At
a willingness to pay (WTP) of $150,000 per quality-adjusted life year (QALY), we find
durable therapy is cost-effective at single administration upfront price points of $713K,
$1.09M, and $2.18M, for full-effectiveness expected durations of 10 years
xxviii
, 20
years
xxix
, and lifetime, respectively. Despite uncertainty around base case estimates, in
a Monte Carlo simulation we find a single administration lifetime cure for SCD provided
xxviii
Median expected duration, conditional on survival.
xxix
Ibid.
168
at birth to be cost-effective 66.0% of the time at WTP of $150K per QALY up to 92.6%
of the time at WTP of $200K per QALY.
Aims and Hypotheses Revisited
In chapter one we outlined three aims of this dissertation and three hypotheses we
intended to test. Here we briefly reiterate each statement and evaluate our findings.
Our first aim was to “determine the magnitude and composition of direct
healthcare costs faced by patients with sickle cell disease in the US. This includes an
age-specific analysis to inform how disease management costs change by stage of life.”
We addressed this aim in multiple chapters. In chapter two we conducted a review of
the medical cost literature; we found few lifetime estimates and none for the
commercially insured. In chapter three, through a cross-sectional analysis of the
Healthcare Cost and Utilization Project (HCUP) Kids' Inpatient Database (KID) 2016
edition, we showed there were an estimated 37,000 SCD-related hospitalizations
among pediatric patients in 2016. The average cost of these hospitalizations was
$9,141, 95% confidence interval (CI) ($8,538, 9,744), resulting in a total financial impact
of $338.2M, 95% CI ($281.8M, 394.6M) to the US healthcare sector. While their
average cost per hospitalization was the lowest, SCD-related stays accounted for 67%
of all stays and the second highest total cost among the eleven orphan diseases we
studied.
In chapter four we utilized patient-level real world evidence on patients affected
with SCD to understand age-specific costs in a commercially insured population. Using
the Optum Clinformatics Data Mart (a large commercial pharmacy and medical claims
169
database) across 2007 to 2017, we identified 6,416 patients affected by SCD from birth
to the ninth decade of life. In the same database, we identified the same number of
comparable controls unaffected by the disease using matching procedures. From this
analysis we find that the unadjusted incremental costs of care for patients with SCD
relative to comparable controls are substantial. These incremental costs are persistent
and statistically significantly different from zero
xxx
across all age groups studied. In our
regression estimates, predicted mean annualized healthcare costs for patients with
SCD increase steadily by age at first disease-related claim. For instance from $34,973,
95% CI: (30,119 - 39,826) for females and $32,152 (25,711 - 38,592) for males at age 1
to $94,021, 95% CI: (80,767 - 107,275) for females and $129,993 (76,217 - 183,768) for
males at age 88 (severities pooled).
The second aim of this dissertation was to “determine the hypothetical lifetime
burden for a patient born with sickle cell disease in the US.” Using the age-specific
results in chapter four, we used a compartmental model to estimate the hypothetical
lifetime burden for a patient cohort born with SCD in the US (50% male) and a cohort of
comparable control patients. Transitions in our two-state model (alive or dead) were
informed by death rates from the medical literature.
2
With this model we calculated
present values finding lifetime costs of $2,938,344 for patients with SCD ($1,191,354
discounted), $1,108,135 for controls ($272,109 discounted), and a difference of
$1,830,209 ($919,245 discounted) attributable to managing SCD over a lifetime. In
chapter five, we refined and applied this procedure to a hypothetical cohort with
characteristics observed in our data. We used the direct costs from our retrospective
xxx
Statistical significance varied depending on how we classified age groups; however, P<0.01
for each decade of life.
170
database analysis combined with disease progression information and survival data
from the medical literature to inform a cohort Markov decision model. Using this model
and initial conditions observed in our data, we find the average hypothetical patient with
SCD treated with standard of care lives 54.9 years (26.2 discounted
xxxi
), experiences
37.6 quality-adjusted life years (17.9 discounted
xxxii
), and incurs $2.77M in direct
healthcare costs ($1.18M discounted
xxxiii
).
The third and final aim of this dissertation was to “estimate the cost-effectiveness
of a hypothetical cure for sickle cell disease provided to patients at birth [as well as]
determine uncertainty associated in cost-effectiveness estimates and analyze the value
in obtaining additional data on sickle cell disease burden.” We addressed this aim
entirely in chapter five. Using the Markov model informed by our retrospective database
analysis and other data from the medical literature, we found a single administration
treatment is likely to be cost-effective by traditional economic evaluation standards. For
a cohort of 10,000 hypothetical individuals with SCD and similar initial characteristics to
patients in our database analysis, lifetime durable treatment is cost-effective relative to
standard of care at a minimum willingness to pay of $140,877/QALY. However, we find
that this result is sensitive to many parameters. In deterministic sensitivity analysis
(DSA) we showed that our estimated ICER varies between as much as $54,207 per
QALY and $190,311 per QALY when we vary inputs univariately. In two-way DSA we
found simultaneous increases in pairwise comparisons of three non-gender specific
parameters often result in ICERs for DT of over $150,000/QALY but always below
xxxi
Discounted at 3% per year.
xxxii
Ibid.
xxxiii
Ibid.
171
$250,000/QALY. In probabilistic sensitivity analysis we showed that the durable therapy
is either dominant or cost-effective in 66.0% of 10,000 iterations at our base case WTP
of $150K/QALY. In value of information analysis, we showed the maximum expected
value of perfect information (EVPI)
xxxiv
is $224.1K per patient at a WTP of $115K/QALY.
At our base case WTP of $150K/QALY the EVPI is $88.9K/patient.
Our hypotheses for this dissertation were threefold. The first stated, “there is a
substantial direct economic burden associated with chronic management of sickle cell
disease over a lifetime for patients born and treated in the US.” Under this point we also
hypothesized that current estimates likely underestimate the true burden for various
reasons. We tested this hypothesis in chapter four, finding that the incremental cost
attributable to management of SCD is large and persistent across all age groups. Our
second hypothesis stated: “a hypothetical cure provided at birth is likely to result in
increased life years, increased quality-adjusted life years, and decreased direct
healthcare expenditures over a lifetime.” In chapter five, we used a Markov model to
show average incremental life years, quality-adjusted life years, and direct healthcare
expenditures are 3.7 years
xxxv
, 8.5 QALYs
xxxvi
, and $1.20M
xxxvii
(2018 USD),
respectively, for patients cured at birth. Lastly, we hypothesized “a cure for sickle cell
disease will be cost-effective to the healthcare sector at willingness to pay of
$150,000/QALY, given consideration to variability in incremental costs, quality of life,
and mortality.” In chapter five we showed a durable therapy is cost-effective (ICER of
xxxiv
The EVPI, in short, is the most a decision maker would be willing to pay to eliminate
decision uncertainty regarding the treatment decision.
xxxv
Discounted at 3% per year.
xxxvi
Ibid.
xxxvii
Ibid.
172
$140,877/QALY relative to SOC) given our assumptions in the base case. In sensitivity
and scenario analyses the model was most sensitive to the expected duration of
therapy (lifetime, median 20 years, or median 10 years), two of the coefficients in the
cost regression for males, the cost of the durable therapy (mean $2.1M per-patient,
range: 1.68M to 2.52M) and the model discount rate (mean 3%, range 2.4 to 3.6%).
Policy Implications
Sickle cell disease is a chronic condition for which treatment in the United States is
complicated by various economic and sociopolitical factors. Often considered a disease
of “double disadvantage”, patients are significantly more likely to come from politically
underrepresented, racially disadvantaged, and lower socioeconomic status
backgrounds.
3
In the United States, it is unknown what proportion of patients are
currently covered by publicly funded insurance; however, one study by Boulet et al.
(2010) found using the 1997–2005 National Health Interview Survey (NHIS) Child
Sample Core that 56.2% of children with SCD were insured by either Medicaid or the
State Children’s Health Insurance Program (SCHIP).
4
Despite a large reliance on public
coverage in this population, our analysis which utilizes estimated allowed amounts for
paid medical and pharmacy claims provides insight into the true cost of managing SCD
in a commercially insured setting. By combining these data with other sources and
current methodological approaches for estimating lifetime outcomes, we find that a
durable gene therapy cure for SCD would result in large health outcome gains and
standard of care cost reductions for those treated early in life. These health
173
improvements and cost reductions are persistent across subgroups; we do not find that
cost-effectiveness outcomes vary substantially by gender.
As we showed in this dissertation, the cost-effectiveness of a hypothetical cure
will depend largely on the price of the single administration therapy. In addition, access
and widespread adoption is contingent on the US healthcare system having appropriate
financial mechanisms to absorb large upfront costs. In relatively higher prevalence
disease areas, these costs are driven by both the high price of the cure (P) and the
large quantity of patients eligible for treatment (Q). Sickle cell disease is not unique in
this regard; one published study estimates that 30 to 60 cell or gene therapy products
will have treated over 300,000 US patients by 2030.
5
Currently, there are few cell and
gene therapy treatments approved by the Food and Drug Administration (FDA)
available in the US market. Examples include Zolgensma for the treatment of spinal
muscular atrophy (approved May 2019), Luxturna for the treatment of vision loss due to
Leber congenital amaurosis or retinitis pigmentosa (approved December 2017),
Kymriah for the treatment of B-cell acute lymphoblastic leukemia (approved August
2017), and Yescarta for the treatment of relapsed or refractory large B-cell lymphoma
(approved October 2017).
6
List prices of these therapies, which do not necessarily
characterize all costs of administration or rebates, are nonetheless helpful for
understanding the magnitude of potential expenditures. For these therapies they range
between $373,000 for select indications of Kymriah and Yescarta to $2.1 million for
Zolgensma per administration.
7,8
From the perspective of society, curing patients with SCD early in life is likely to
be significantly more cost-effective than our base case healthcare sector perspective
174
analysis suggests. The many burdens of SCD are pervasive and difficult to quantify for
direct inclusion in a modeling exercise. These include impacts at the patient level on
longevity, psychological and physical well-being, workplace productivity, and medical
and non-medical financial costs. In addition, caregivers, parents, and other family
members of individuals with SCD have been shown to suffer similar impacts along with
substantial uncompensated efforts.
9-11
Our healthcare sector analysis focuses primarily
on factors for which high-quality data across all age groups exist. These include direct
medical costs, longevity, and health-related quality of life. For these reasons, we believe
our estimated base case ICER of $140,877/QALY can serve as an upper bound for an
ICER generated from the US societal perspective. Inclusion of additional evidence on
SCD would likely show cost benefits in other sectors
xxxviii
.
Clinically, patients are likely to benefit greatly from the availability of a durable
cure for SCD. Preliminary phase I trial results for bluebird bio product LentiGlobin™
(NCT02140554) showed a 99% reduction
xxxix
in annualized vaso-occlusive crises (VOC)
and acute chest syndrome (ACS).
12
Any treatment that becomes available to the US
market will have satisfied FDA standards for safety. Despite this, elective uptake at the
patient level is uncertain. The black and African American community has been shown
to exhibit high levels of medical mistrust, often attributed to a vast history of
marginalization and medical exploitation in the United States.
13
In sickle cell disease
specifically, there is evidence that a common barrier to allogenic hematopoietic stem
xxxviii
Increased labor force participation, increased productivity, and decreased uncompensated
labor by caregivers, for instance.
xxxix
In their trial’s “Group C” patients with a history of VOC, acute chest syndrome (ACS), and at
least 6-months of follow-up. Groups A and B saw a lesser reduction at two-years after
treatment.
175
cell transplant (HSCT)
xl
clinical trial participation is medical mistrust of healthcare
providers.
14
Cell or gene therapy uptake for appropriate patients may require
informational campaigns or other strategies to directly address mistrust in the
healthcare system.
15
The impact on US payers of a cure for SCD will vary. As previously described,
the majority of patients with SCD in the US are covered by publicly funded state or
federal programs. State Medicaid programs in southern US states with disproportionate
numbers of prevalent cases may feel considerable financial strain. This may result in
Medicaid plans reimbursing the cell or gene therapy at a lower rate than commercial
plans.
16
Given the life expectancy at birth of less than 55 years in this condition,
Medicare may have only a small prevalent population to cure. As well, given the typical
coverage path of individuals with SCD: Medicaid early in life, an increase in commercial
or other plan usage later in life, and then Medicare at age 65, it is unlikely Medicare will
have many patients to cure past the prevalent population. Commercial payers may be
reluctant to cover expensive single administration therapies. Concerns may include: the
large bolus of upfront cost due to treatment of the prevalent population, patient plan
mobility post-treatment, and product performance concerns.
17
Even under demonstrated
long term cost-effectiveness of a cure, costs and benefits are likely to fall on various
stakeholders in the fragmented US healthcare system. Creative financing tools for cures
in various disease areas have been proposed; their implementation may be a necessary
catalyst to widespread adoption.
18-21
As an historically neglected condition with serious
xl
HSCT is currently the only available cure for SCD. The procedure is considered “high-risk” and
is currently only viable for a small fraction of patients with SCD.
176
societal impact, we hope a potential cure for SCD is developed and becomes widely
accessible in the coming years.
177
References
1. Kauf TL, Coates TD, Huazhi L, Mody-Patel N, Hartzema AG. The cost of health
care for children and adults with sickle cell disease. Am J Hematol.
2009;84(6):323-327.
2. Lubeck D, Agodoa I, Bhakta N, et al. Estimated Life Expectancy and Income of
Patients With Sickle Cell Disease Compared With Those Without Sickle Cell
Disease. JAMA Netw Open. 2019;2(11):e1915374.
3. Hijmans CT, Fijnvandraat K, Oosterlaan J, Heijboer H, Peters M, Grootenhuis
MA. Double disadvantage: a case control study on health-related quality of life in
children with sickle cell disease. Health Qual Life Outcomes. 2010;8:121.
4. Boulet SL, Yanni EA, Creary MS, Olney RS. Health status and healthcare use in
a national sample of children with sickle cell disease. Am J Prev Med. 2010;38(4
Suppl):S528-535.
5. Quinn C, Young C, Thomas J, Trusheim M, Group MNFW. Estimating the
Clinical Pipeline of Cell and Gene Therapies and Their Potential Economic
Impact on the US Healthcare System. Value Health. 2019;22(6):621-626.
6. American Society of Gene and Cell Therapy (ASGCT). Gene and Cell Therapy
FAQ's. 2020; https://www.asgct.org/education/more-resources/gene-and-cell-
therapy-faqs. Accessed May 20, 2020.
7. Salzman R, Cook F, Hunt T, et al. Addressing the Value of Gene Therapy and
Enhancing Patient Access to Transformative Treatments. Mol Ther.
2018;26(12):2717-2726.
178
8. Rosenmayr-Templeton L. Industry update for May 2019. Ther Deliv.
2019;10(9):555-561.
9. Brown RT, Connelly M, Rittle C, Clouse B. A longitudinal examination predicting
emergency room use in children with sickle cell disease and their caregivers. J
Pediatr Psychol. 2006;31(2):163-173.
10. Moskowitz JT, Butensky E, Harmatz P, et al. Caregiving time in sickle cell
disease: Psychological effects in maternal caregivers. Pediatr Blood Cancer.
2007;48(1):64-71.
11. Barakat LP, Patterson CA, Tarazi RA, Ely E. Disease-related parenting stress in
two sickle cell disease caregiver samples: Preschool and adolescent. Families,
Systems, & Health. 2007;25(2):147-161.
12. Walters MC, Tisdale JF, Kwiatkowski JL, et al. Exploring the Drivers of Potential
Clinical Benefit in Initial Patients Treated in the Hgb-206 Study of Lentiglobin for
Sickle Cell Disease (SCD) Gene Therapy. Blood.
2019;134(Supplement_1):2061-2061.
13. Alsan M, Wanamaker M. Tuskegee and the Health of Black Men*. The Quarterly
Journal of Economics. 2017;133(1):407-455.
14. Omondi NA, Ferguson SES, Majhail NS, et al. Barriers to hematopoietic cell
transplantation clinical trial participation of african american and black youth with
sickle cell disease and their parents. J Pediatr Hematol Oncol. 2013;35(4):289-
298.
179
15. Stevens EM, Patterson CA, Li YB, Smith-Whitley K, Barakat LP. Mistrust of
Pediatric Sickle Cell Disease Clinical Trials Research. Am J Prev Med. 2016;51(1
Suppl 1):S78-86.
16. Fulton BD, Felton MC, Pareja C, Potischman A, Scheffler RM. Coverage, cost-
control mechanisms, and financial risk-sharing alternatives of high-cost health
care technologies. 2009.
17. Barlow JF, Yang M, Teagarden JR. Are Payers Ready, Willing, and Able to
Provide Access to New Durable Gene Therapies? Value Health. 2019;22(6):642-
647.
18. Basu A, Subedi P, Kamal-Bahl S. Financing a Cure for Diabetes in a Multipayer
Environment. Value Health. 2016;19(6):861-868.
19. Barlow J, Teagarden R, Trusheim M. Precision Financing of Durable, Potentially
Curative Therapies. PharmExec. 2019.
20. Daniel G, Leschly N, Marrazzo J, McClellan MB. Advancing Gene Therapies And
Curative Health Care Through Value-Based Payment Reform. Health Affairs
Blog. 2017.
21. Trusheim MR, Bach P. Next Steps For Netflix: Can ‘All You Can Treat’ Contracts
Work Beyond Hepatitis C? Health Affairs Blog. 2019.
Abstract (if available)
Abstract
Sickle cell disease (SCD) is an inherited blood disorder with significant healthcare resource utilization. It is commonly associated with frequent complications including pain crises, infections, acute chest syndrome, and stroke, all of which lead to reduced life expectancy. Treatment for SCD has evolved significantly over past decades in the United States (US), with gene therapy treatment currently in trials and showing promise. Lifetime costs for managing SCD are uncertain. This dissertation encompasses three aims. These include determining: (1) the magnitude and composition of direct healthcare costs faced by patients with SCD in the US, (2) a hypothetical lifetime economic burden for managing SCD, and (3) the potential cost-effectiveness of a cure provided at birth, and value of information (VOI) of obtaining additional data on the burden of SCD. We propose addressing these aims using a large database of private insurance claims data (Optum’s de-identified Clinformatics® Data Mart Database) supplemented by other data sources. We estimate age-specific healthcare costs using real-world evidence (RWE) of patients with SCD relative to propensity-score matched control patients. We then use these estimates to simulate the incremental economic burden associated with the disease over a lifetime. The cost-effectiveness of a durable treatment can be evaluated based on these lifetime costs and quality-adjusted life years, assuming the cure is provided at birth. VOI analysis provides additional inference regarding uncertainty around these figures and may inform future RWE generation efforts in SCD.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Evaluating treatment options for metastatic, castration-resistant prostate cancer: a comprehensive value assessment
PDF
The disease burden among aging people living with HIV/AIDS
PDF
Crohn’s disease: health outcomes and resource utilization in the biologic era
PDF
Impact of medication persistence on clinical and health care cost outcomes in patients with major depressive disorder using retrospective claims data
PDF
Advances and applications for economic evaluation methods in health technology assessment (HTA)
PDF
A series of longitudinal analyses of patient reported outcomes to further the understanding of care-management of comorbid diabetes and depression in a safety-net healthcare system
PDF
Three essays in health economics
PDF
Essays on the economics of infectious diseases
PDF
Optimizing healthcare decision-making: Markov decision processes for liver transplants, frequent interventions, and infectious disease control
PDF
Physician behavior and low-value care
PDF
Three essays on emerging issues in hemophilia care
Pdf
Alzheimer's disease: risk factors, value, and alternative payment models
PDF
Disease modifying treatments for Alzeimer's disease: modeling, value assessment and eligible patients
PDF
Documenting the impact of under-treatment for chronic disorders
PDF
Evaluating cancer treatments with health economics methods
PDF
Essays in pharmaceutical and health economics
PDF
Calibration uncertainty in model-based analyses for medical decision making with applications for ovarian cancer
PDF
Value in oncology care and opportunities for improvement
PDF
Value in health in the era of vertical integration
PDF
Economic aspects of obesity
Asset Metadata
Creator
Salcedo, Jonathan
(author)
Core Title
Essays in health economics
School
School of Pharmacy
Degree
Doctor of Philosophy
Degree Program
Health Economics
Publication Date
07/14/2020
Defense Date
05/05/2020
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
cost-effectiveness analysis,costs and cost analyses,curative therapy,economic evaluation,genetic therapy,health care costs,Markov model,Monte Carlo simulation,OAI-PMH Harvest,retrospective database analysis,sickle cell anemia,sickle cell disease,single or short-term therapy,standard of care
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Lakdawalla, Darius Noshir (
committee chair
), Padula, William Vincent (
committee member
), Suen, Sze-chuan (
committee member
)
Creator Email
jonathan.salcedo.js@gmail.com,salcedoj@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c89-328444
Unique identifier
UC11663410
Identifier
etd-SalcedoJon-8676.pdf (filename),usctheses-c89-328444 (legacy record id)
Legacy Identifier
etd-SalcedoJon-8676.pdf
Dmrecord
328444
Document Type
Dissertation
Rights
Salcedo, Jonathan
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
cost-effectiveness analysis
costs and cost analyses
curative therapy
economic evaluation
genetic therapy
health care costs
Markov model
Monte Carlo simulation
retrospective database analysis
sickle cell anemia
sickle cell disease
single or short-term therapy
standard of care