University of Southern California Dissertations and Theses

Advances and applications for economic evaluation methods in health technology assessment (HTA)

(USC Thesis Other)

Advances and applications for economic evaluation methods in health technology assessment (HTA)

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content i

Advances and applications for economic evaluation
methods in Health Technology Assessment (HTA)

by
Yao Ding

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PHARMACEUTICAL ECONOMICS AND POLICY)

August 2015

ii

DEDICATION

To my parents for their endless love, support, and sacrifice;
To my baby boy, Dean, for having you in my life and bringing us the greatest joy and
happiness;
To my husband for always being there to share my frustration, to support and
encourage me.

iii

ACKNOWLEDGEMENTS

I would like to sincerely thank my advisor and the dissertation committee chair, Dr.
Joel W. Hay, for his continuous support academically and personally. Dr. Hay not only
taught me how to conduct a cost-effectiveness analysis in the medical field, but also
introduced me to the UCLA Center for Vaccine Research and initiated collaboration with
the experts there, which subsequently resulted in two publications in clinical journals.
He offered me invaluable advice, guidance and encouragement in every aspect
throughout my graduate study at USC. I’ll bring his enthusiasm for research to my
career in the next chapter of my life.
I would also like to thank my thesis committee members, Dr. Ken Zangwill and Dr.
Neeraj Sood. Dr. Zangwill refined my clinical understanding and provided erudite
suggestions about the hospital-based postpartum vaccination programs. I would never
have been able to publish the first two studies of my dissertation without his meticulous
edits and revisions. Dr. Sood taught me the econometric methods in class and how to
apply those models in a real-world setting. I’m extremely grateful for his insightful
suggestions and innovative ideas to refine the control group for the sensitivity analyses
in my third study. Special thanks to Dr. Vivian Wu and Dr. Geoffrey Joyce for their
helpful suggestions and invaluable feedback to elucidate my proposal.
I would also like to extend my appreciation to Dr. Jeffrey McCombs who is always
trying to help me in career development. I wanted to thank all of my classmates and
friends from the bottom of my heart for making my time at USC a truly excellent
experience.
iv

TABLE OF CONTENTS

ACKNOWLEDGEMENTS ............................................................................................... iii
LIST OF TABLES ............................................................................................................ vi
LIST OF FIGURES ......................................................................................................... vii
ABSTRACT ..................................................................................................................... 1
CHAPTER 1: INTRODUCTION ....................................................................................... 2
1.1. Background ......................................................................................................... 2
1.2. Overview of Economic Evaluation Methods ........................................................ 4
1.3. Purpose of this Study .......................................................................................... 9
1.4. Chapter-One References .................................................................................. 11
CHAPTER 2: COST-BENEFIT ANALYSIS OF IN-HOSPITAL INFLUENZA
VACCINATION OF POSTPARTUM WOMEN .............................................................. 15
Chapter-Two Abstract .............................................................................................. 15
2.1. Chapter-Two Background ................................................................................. 17
2.2. Chapter-Two Methods ....................................................................................... 18
2.2.1. Data Source ............................................................................................... 18
2.2.2. Economic Model and Study Population ...................................................... 18
2.2.3. Probabilities ................................................................................................ 20
2.2.4. Healthcare Resource Costs ........................................................................ 22
2.3. Chapter-Two Results ......................................................................................... 24
2.4. Sensitivity Analyses ........................................................................................... 25
2.5. Chapter-Two Discussion ................................................................................... 28
2.6. Conclusions ....................................................................................................... 30
2.7. Tables and Figures ............................................................................................ 32
2.8. Chapter-Two References .................................................................................. 38
CHAPTER 3: COST–BENEFIT ANALYSIS OF HOSPITAL-BASED POSTPARTUM
VACCINATION WITH COMBINED TETANUS TOXOID, REDUCED DIPHTHERIA
TOXOID, AND ACELLULAR PERTUSSIS VACCINE (Tdap) ....................................... 42
Chapter-Three Abstract ............................................................................................ 42
3.1. Chapter-Three Background ............................................................................... 44
v

3.2. Chapter-Three Materials and Methods .............................................................. 45
3.2.1. Data Collection and Economic Model ......................................................... 45
3.2.2. Outcome Measure ...................................................................................... 47
3.2.3. Probabilities ................................................................................................ 47
3.2.4. Costs .......................................................................................................... 49
3.3. Chapter-Three Results ...................................................................................... 52
3.3.1. Base-case Results ..................................................................................... 52
3.3.2. Sensitivity Analyses .................................................................................... 52
3.4. Chapter-Three Discussion ................................................................................. 54
3.5. Conclusions ....................................................................................................... 57
3.6. Tables and Figures ............................................................................................ 58
3.7. Chapter-Three References ................................................................................ 64
CHAPTER 4: DIRECT MEDICAL COSTS ASSOCIATED WITH PEYRONIE'S
DISEASE IN THE UNITED STATES ............................................................................. 68
Chapter-Four Abstract .............................................................................................. 68
4.1. Chapter-Four Background ................................................................................. 70
4.2. Chapter-Four Materials and Methods ................................................................ 72
4.2.1. Data Source ............................................................................................... 72
4.2.2. Study Population in the Base Case ............................................................ 73
4.2.3. Comparative Analyses and Statistical Tests ............................................... 76
4.2.4. Model Structure .......................................................................................... 76
4.2.5. Tests for Attrition Bias ................................................................................ 78
4.2.6. Sensitivity Analyses with Alternative Control Groups ................................. 79
4.3. Chapter-Four Results ........................................................................................ 81
4.3.1. Base-Case Results ..................................................................................... 81
4.3.2. Results for Sensitivity Analyses .................................................................. 85
4.4. Chapter-Four Discussion ................................................................................... 86
4.5. Tables and Figures ............................................................................................ 91
4.6. Chapter-Four Appendix ................................................................................... 100
4.7. Chapter-Four References ................................................................................ 108
CHAPTER 5: CONCLUSIONS .................................................................................... 113
vi

LIST OF TABLES

Table 1.1. Types of economic evaluation methods used in HTA.................................... 8
Table 2.1. Probability variables used in the model......................................................... 32
Table 2.2. Base-case parameters for direct and indirect costs....................................... 33
Table 2.3. One-way sensitivity analysis of selected parameters.................................... 34
Table 2.4. Two-way sensitivity analyses of annual attack rate and vaccine efficacy...... 35
Table 3.1. Probability variables used in the model…………………………….................. 58
Table 3.2. Base-case parameters for direct and indirect costs....................................... 59
Table 3.3. Base-case results from a societal or healthcare system perspective............ 60
Table 3.4. Two-way sensitivity analyses of annual incidence and vaccine efficacy
from a societal perspective............................................................................................. 61
Table 4.1. Baseline characteristics and comorbidities.................................................... 91
Table 4.2.a. Resource use during the 3-year study period............................................. 93
Table 4.2.b. Selected drug use at baseline…………………........................................... 94
Table 4.3.a. Distribution of healthcare costs in Medicare cohort (unadjusted)............... 95
Table 4.3.b. Distribution of costs in commercial cohort (unadjusted)............................. 96
Table A.4.1. List of ICD-9 codes and CPT procedure codes…………………………...... 100
Table A.4.2. Cost comparison between PD and physician office visit controls……....... 101
Table A.4.3. Cost comparison between PD and benign prostatic hypertrophy
patients........................................................................................................................... 102
Table A.4.4. Subgroup analysis of the base-case excluded PD patients....................... 104
Table A.4.5. Seemingly unrelated estimation results evaluating difference in
coefficients of GLM model between full and completing samples.................................. 105
Table A.4.6. Cross-sectional probit regressions to predict the probability of attrition on
characteristics variables and lagged outcomes.............................................................. 106
Table A.4.7. Becketti, Gould, Lilard and Welch (BGLW) test for random attrition.......... 107

vii

LIST OF FIGURES

Figure 2.1. Overall structure of a postpartum influenza vaccination strategy.............. 36
Figure 2.2. Tornado diagram of one-way sensitivity analyses..................................... 37
Figure 3.1. Model structure of a postpartum Tdap strategy......................................... 62
Figure 3.2. Tornado diagram of one-way sensitivity analyses..................................... 63
Figure 4.1. Base case study scheme.......................................................................... 97
Figure 4.2. Adjusted incremental total costs per 6 months, 95% CIs.......................... 98
Figure 4.3.a. Adjusted incremental outpatient costs.................................................... 99
Figure 4.3.b. Adjusted incremental prescription drug costs......................................... 99
Figure A.4.1. Adjusted incremental outpatient costs in PD patients over benign
prostatic hypertrophy patients……………………………………………………………... 103

1

ABSTRACT

Health technology assessment (HTA) is a multidisciplinary process that assesses
and evaluates the economic, clinical, social, and ethical impacts of health care
technologies. HTA incorporates number of areas of expertise including clinical research,
epidemiology, health services research, economics, and psychometrics. The field has
rapidly expanded in the last decade and played a crucial role in improvement of the
quality of healthcare.
This three-paper dissertation demonstrates two common approaches of economic
evaluation in HTA: Paper 1 and 2 are cost-benefit analyses (CBAs), with cost data
obtained from cost-of-illness (COI) studies and health outcomes data from systematic
evidence review or clinical trials, to compare between different healthcare programs or
interventions. Paper 3 is a COI study conducted from a third-party payer’s perspective
with cost data directly collected from administrative claims databases used for health
care payment. The cost data from one or more such sources often are combined with
data from primary clinical studies, epidemiological studies, and other sources to conduct
the CBAs, cost-effectiveness analyses and other analyses that involve weighing health
and economic impacts of a health technology. Compared to the COI analysis, CBA
provides additional evidence that can be used to determine the best prevention
intervention or treatment with respect to the disease studied, which can assist policy
makers to fund the interventions or to evaluate completing public health programs (e.g.,
immunization, newborn screening, and water purification).

2

CHAPTER 1: INTRODUCTION

Background
Health technology assessment (HTA) is a multidisciplinary process that assesses
and evaluates the economic, clinical, social, and ethical impacts of health care
technologies. Such medical technologies or interventions include but are not limited to
prescription drugs, diagnostic tests, medical devices, surgical procedures and
healthcare programs. HTA has rapidly developed in the last decade and has played a
crucial role in improving the quality of healthcare delivery system (Banta, 2009; Kunkle,
1995; Margolis and Guston, 2003). This field incorporates different areas of expertise
including clinical research, epidemiology, economics, psychometrics, and biostatistics.
The Institute of Medicine defined HTA as “any process of examining and reporting
properties of a medical technology used in health care, such as safety, efficacy,
feasibility, and indications for use, cost, and cost-effectiveness, as well as social,
economic, and ethical consequences” (Institute of Medicine, 1985). The National
Information Center on Health Services Research and Health Care Technology (NICHSR)
defined HTA as the systematic evaluation of properties, effects or other impacts of a
health technology (National Institutes of Health, 2014).

HTAs have been widely used to inform technology-related policies and decisions
at the individual, the health care provider, the payer or the national levels. For example,
clinicians may use HTA data to advise patients on the appropriate health care
interventions for their specific clinical needs (Glasziou, 2011; Straus et al., 2011);
Pharmaceutical or medical device companies may require HTA to make decisions for
3

new product development and marketing; Payers including health insurance plans, drug
formularies and employers may conduct the HTA to decide the coverage or
reimbursement of an intervention. Furthermore, regulatory agencies (e.g., U.K. National
Institute for Health and Care Excellence) need HTA evidence for decision making on
new drugs or devices approaching market approval. In addition, government health and
disease control agencies may apply HTA to evaluate specific public health programs,
such as immunization for infectious diseases, universal newborn screening for genetic
disorders, water purification system, and environmental protection programs (Kuperman
et al., 2010).

Among the above applications of HTA, a main contribution is to provide health
economic and outcomes-based evidence to clinicians, public health professionals,
regulatory officials, payers and other policy makers for informed decision making on
medical technologies (Banta, 2009; Banta and Luce, 1993; Hailey and Menon, 1999).
Economic attributes or impacts of health technologies can be macroeconomic and
microeconomic. For example, macroeconomic impacts include national health care
costs, resource allocation across private and public health care sectors, international
trade, and policy changes that affect technological innovation, adoption, diffusion, and
use (National Institutes of Health, 2014). On the other hand, microeconomic implications
that pertain to health technologies include costs, charges, payments associated with
individual technologies, and impacts of alternative health care interventions for a given
health problem. Such impacts can be evaluated by comparing resource requirements
(e.g., costs) and outcomes (or benefits) in a cost effectiveness analysis, cost utility
4

analysis and cost benefit analysis. In this dissertation, I will focus on evaluating the
microeconomic attributes of health technology. The following section provides a brief
overview of the economic analysis methods used to estimate the microeconomic
outcomes in HTA.

Overview of Economic Evaluation Methods
The required resources to provide health care, including equipment, materials,
facilities and staff, are scarce and can never be sufficient to meet all medical demands
(Brazier et al., 2007). In 2012, total health care spending increased by 3.7% to $2.8
trillion, as a percentage of nominal (i.e., not adjusted for inflation) gross domestic
product (GDP) of 17.2% in the U.S. (Martin et al., 2014). As a result of the continued
implementation of the Affordable Care Act (ACA) coverage expansions, faster projected
economic growth, and the aging of the population, health spending is projected to be
19.3% of GDP by 2023 with an average expected growth of 6.0% per year from 2015 to
2023 (Centers for Medicare & Medicaid Services, 2015). With the concerns about the
unmet medical needs and the pressures about the rising health care costs, economic
evaluation of health technologies is one of the most critical research issues now and for
the future. In order to optimize allocation of the limited health care resources and to
improve social welfare, economic evaluation assists policy makers to make efficient and
equitable decisions by comparing the costs and benefits of alternative health care
interventions (Brazier et al., 2007; Drummond et al., 2005). Generally, there are the
following types of analysis or approach to evaluate the economic outcomes in HTA
research:
5

Cost-of-illness analysis (COI) measures the economic impact of a disease or
condition and estimate the total health care costs that would potentially be saved if the
disease were to be eradicated (Segel, 2006). COI studies have been applied in public
and private health care sectors to decide which disease should require increased
allocation of prevention or treatment resources. For example, COI analysis can estimate
the economic burden a disease has on public programs, such as Medicaid and
Medicare (Finkelstein et al., 2003; Taylor and Sloan, 2000; Zhang et al., 1999).
Furthermore, costs due to an illness that affect business or industry suggest the impact
of their employee’s absenteeism and lost productivity (Thompson, Edelsberg et al. 1998;
Goetzel, Long et al. 2004). In addition, COI studies that provide crucial cost estimates to
cost-benefit and cost-effectiveness analyses are a first step in the economic evaluation
of prevention interventions (Finkelstein and Corso, 2003). Nonetheless, one important
limitation of COI studies is that they do not quantify the effectiveness of the disease-
related health care interventions (Koopmanschap, 1998).

Cost-effectiveness analysis (CEA) compares different interventions with costs
estimated from COI studies and outcomes measured in cost per unit of effect, for
example, cost per death averted or cost per positive cancer detected (Gold et al., 1996).
Cost-utility analysis (CUA) is a special form of CEA with outcomes measured in cost per
quality-adjusted life year (QALY). The QALY is a unit of health care outcome that
combines length of life with health-related quality of life on a single measure
representing ‘a year in full health’ (Drummond et al., 2005). The numbers of QALYs are
6

calculated by multiplying the value of the health-related quality of life for each health
state (i.e., health utility) by the duration of the health state (Torrance and Feeny, 1989).
The health utility is measured on a scale between zero and one, where zero represents
a health state equivalent to being dead and one is perfect health (Neumann et al., 2000).
QALYs have been used to estimate the burden of disease for oncology clinical trials, to
make economic comparisons of alternative health care interventions in cost-utility
analyses, and to evaluate the impact of health technologies by some HTA organizations,
e.g., the National Institute for Health and Care Excellence (NICE) in the UK (Gold et al.,
2002; Sloan et al., 2014).

Cost-benefit analysis (CBA) compares the costs with the benefits of alternative
interventions to prevent, diagnose, treat, or monitor a clinical condition, where all the
benefits measured in monetary terms (Brazier et al., 2007). For example, flu vaccine
can reduce the infection of flu virus, avert a case of flu, and reduce the mortality and
morbidity associated with flu (Ding et al., 2012). If the monetary valuation of all the
benefits exceeds or offsets the total costs of an intervention, then the intervention is
considered to be cost-beneficial (Brazier et al., 2007). Here, the terminology “cost-
beneficial” is different from “cost-saving”. A cost-saving intervention leads to lower total
costs than the alternative strategy and may or may not improve health outcomes;
whereas a cost-beneficial intervention may or may not raise costs but the reduced
medical expenditure and monetary value of health benefits offsets the costs of
intervention (Prosser et al., 2012).

7

Compared to CBA, CEA requires a given threshold of willingness to pay (WTP) for
QALYs to be logically consistent for decision making. However, once the decision
maker accepts a specific WTP for QALYs, Net Monetary Benefits (i.e., a form of CBA) is
a better approach because it deals with statistical issues such as estimating confidence
intervals better than CEA (Stinnett and Mullahy, 1998; Laska et al., 1999; Briggs, 2001;
Hay, 2015) As a result, the decisions made based on CEA with QALYs may be
inefficient to allocate scarce health care resources (Brazier et al., 2007). In addition,
CBA can capture non-health benefits of treatment such as the recovered productivity of
both patients and caregivers, as all the inputs are measured in monetary term and the
outcome measures are reported in terms of net benefits. On the other hand, concerns
regarding CBA include: (1) attaching a monetary value to the health of a person has
been less well-accepted by the medical community (Drummond et al., 2005; Prosser et
al., 2012); (2) some researchers proposed that CBA is insensitive to the impact of
income inequality, as CBA assesses the benefits and costs in monetary terms at an
individual level which results in it more efficient to improve the health of the rich rather
than the poor (Brazier et al., 2007).

Another technique of economic evaluation is called cost-minimization analysis
(CMA). CMA assumes the alternative interventions to produce identical outcomes and
seeks the least cost one. However, achieving equivalent results is rarely realistic as
there’s always uncertainty around the outcome measure in practice (Brazier et al.,
2007). Table 1.1 shows the differences and comparisons among these economic
analysis methods.
8

Table 1.1 Types of economic evaluation methods used in HTA
Economic
Analysis
Objective Valuation of
Health Effects
Measure of
Outcome
Cost-of-illness
(COI)
Estimate economic impact of a
disease
Morbidity and
mortality costs
Net cost
Cost-effectiveness
(CEA)
Compare interventions that
achieve clinical end points, e.g.,
deaths averted
Natural units Lowest cost per
case averted
Cost-utility
(CUA)
Compare interventions that have
morbidity and mortality outcomes
Quantified in a
health utility
index
Lowest cost per
quality-adjusted
life year (QALY)
Cost-benefit
(CBA)
Compare alternative programs
with both health and non-health
outcomes
Dollars Greatest net
benefit or Benefit-
to-cost ratio
Cost-minimization
(CMA)
Compare interventions that
produce identical outcomes
Assumed
equivalent
Lowest cost

All the above economic evaluation techniques vary by analytical time horizon and
perspective (Gold et al., 1996; Hodgson, 1994). Time horizon refers to over what time
frame are the comparative outcomes to be assessed (e.g., 90 days post-hospital
discharge, 5 years or life-time). Time frame of a study may influence the magnitude of
costs, health status and other outcomes of a health care intervention (Hay, 2004).
Perspective represents the decision-maker’s framework to measure the costs and
benefits of interventions. Studies conducted from different perspectives may generate
large potential differences in costs and outcome measures. For instance, in a third party
payer’s perspective, costs are restricted to direct payments by the third party payer.
Costs estimated from a healthcare system perspective include direct costs that
aggregate medical expenditures regardless of payer (Ouyang et al., 2009). Nonetheless,
economic analysis conducted from a societal perspective will cover both the direct and
indirect costs associated with a disease or health technology. Generally, direct costs
9

refer to the market-purchased resources associated with the intervention; indirect costs
capture the lost productivity and absenteeism due to an illness or death, including the
productivity losses due to mortality or disability, and opportunity costs associated with
goods and services consumed but not purchased in the market, such as the services of
uncompensated caregivers who missed time from other activities to care for patients
(Goetzel et al., 2004; Zhou et al., 2005). Typically, economic analyses from a third-party
payer or healthcare system perspective tend to result in lower costs than that from a
societal perspective (Gold et al., 1996; Hodgson, 1994).

Purpose of this Study
Economic evaluation methods in HTA facilitate health care decision making to
optimize the allocation of scarce resources, to address unmet medical needs, to
improve the quality of health care delivery and at the same time to control costs. This
three-paper dissertation demonstrates two common approaches of economic evaluation
to provide health economic and outcomes-based evidence for HTA. Paper 1 and 2 are
CBAs with cost data collected from COI studies and health outcomes data from
systematic review or clinical trials to assist decision makers to compare between
alternative healthcare programs or interventions (e.g., postpartum vaccination versus no
vaccination) in terms of the probabilities that certain events will occur and the costs of
the outcomes that would result from each event. Paper 3 is a COI study conducted from
a third-party payer’s perspective with cost data collected from administrative claims
databases used for health care payment. This study is concerned with how one actually
gathers and assesses the direct medical costs for a specific disease. In addition, cost
10

data from such sources typically are combined with data from clinical trials,
observational studies, case registries, public health statistics, and preference surveys to
conduct a CBA or CEA to aid decision making on interventions of the same disease
(Byford et al., 2000; Corso et al., 2004).

11

Chapter-One References
Banta, D. (2009). What is technology assessment? International journal of technology
assessment in health care 25 Suppl 1, 7-9.
Banta, H.D., and Luce, B.R. (1993). Health Care Technology and Its Assessment: An
International Perspective. New York, NY: Oxford University Press.
Brazier, J., Ratcliffe, J., Salomon, J.A., and Tsuchiya, A. (2007). Measuring and Valuing
Health Benefits for Economic Evaluation. Oxford, United Kingdom: Oxford University
Press.
Briggs, A.H. (2001). A bayesian approach to stochastic cost-effectiveness analysis.
International Journal of Technology Assessment in Health Care 17 (1), 69-82.
Byford, S., Torgerson, D.J., and Raftery, J. (2000). Economic note: cost of illness
studies. BMJ 320, 1335.
Centers for Medicare & Medicaid Services. Available at: http://www.cms.gov/Research-
Statistics-Data-and-Systems/Statistics-Trends-and-
Reports/NationalHealthExpendData/Downloads/Proj2013.pdf.
Corso, P., Grosse, S., and Finkelstein, E. (2004). The skinny on COI analysis. Obesity
research 12(7), 1189.
Ding, Y., Zangwill, K.M., Hay, J.W., Allred, N.J., and Yeh, S.H. (2012). Cost-benefit
analysis of in-hospital influenza vaccination of postpartum women. Obstetrics and
gynecology 119, 306-314.
Drummond, M., Sculpher, M., and Torrance, G.W. (2005). Methods for the Economic
Evaluation of Health Care Programmes. 3rd ed.. Oxford, United Kingdom: Oxford
University Press.
Finkelstein, E., and Corso, P. (2003). Cost-of-illness analyses for policy making: a
cautionary tale of use and misuse. Expert review of pharmacoeconomics & outcomes
research 3, 367-369.
12

Finkelstein, E.A., Fiebelkorn, I.C., and Wang, G. (2003). National medical spending
attributable to overweight and obesity: how much, and who's paying? Health affairs
Suppl Web Exclusives, W3-219-226.
Glasziou, P. (2011). What is EBM? Evidence-based medicine 16, 129-130.
Goetzel, R.Z., Long, S.R., Ozminkowski, R.J., Hawkins, K., Wang, S., and Lynch, W.
(2004). Health, absence, disability, and presenteeism cost estimates of certain physical
and mental health conditions affecting U.S. employers. Journal of occupational and
environmental medicine / American College of Occupational and Environmental
Medicine 46, 398-412.
Gold, M.R., Siegel, J.E., Russell, L.B., and Weinstein, M.C. (1996). Cost-Effectiveness
in Health and Medicine. New York, NY: Oxford University Press.
Gold, M.R., Stevenson, D., and Fryback, D.G. (2002). HALYS and QALYS and DALYS,
Oh My: similarities and differences in summary measures of population Health. Annual
review of public health 23, 115-134.
Hailey, D., and Menon, D. (1999). A short history of INAHTA. International Network of
Agencies for Health Technology Assessment. International journal of technology
assessment in health care 15, 236-242.
Hay, J.W. (2015). Cost Effectiveness Analysis with and Without Stochastic Uncertainty.
CESR-Schaeffer Working Paper No. 2015-005. Available at SSRN:
http://ssrn.com/abstract=2601147 or http://dx.doi.org/10.2139/ssrn.2601147
Hay, J.W. (2004). Evaluation and review of pharmacoeconomic models. Expert Opin
Pharmacother 5, 1867-1880.
Hodgson, T.A. (1994). Costs of illness in cost-effectiveness analysis. A review of the
methodology. PharmacoEconomics 6, 536-552.
Institute of Medicine. Assessing Medical Technologies. Washington, DC: National
Academy Press; 1985. http://www.nap.edu/openbook.php?record_id=607.
13

Koopmanschap, M.A. (1998). Cost-of-illness studies. Useful for health policy?
PharmacoEconomics 14, 143-148.
Kunkle, G. (1995). New challenges or the past revisited? The Office of Technology
Assessment in historical context Technology in Society 17(2), 175-196.
Kuperman, G.J., Blair, J.S., Franck, R.A., Devaraj, S., Low, A.F., and Group,
N.T.I.C.S.C.W. (2010). Developing data content specifications for the nationwide health
information network trial implementations. Journal of the American Medical Informatics
Association : JAMIA 17, 6-12.
Laska, E.M., Meisner, M., Siegel, C., Stinnett, A.A. (1999). Ratio-based and net benefit-
based approaches to health care resource allocation: proofs of optimality and
equivalence. Health Economics 8(2), 171-4.
Margolis, R.M., and Guston, D.H. (2003). The origins, accomplishments, and demise of
the Office of Technology Assessment. Science and Technology Advice for Congress
Washington, DC: Resources for the Future, 53-76.
Martin, A.B., Hartman, M., Whittle, L., Catlin, A., and National Health Expenditure
Accounts Team. (2014). National health spending in 2012: rate of health spending
growth remained low for the fourth consecutive year. Health Affairs 33(1), 67-77.
National Institutes of Health. National Information Center on Health Services Research
and Health Care Technology (NICHSR). Available from:
http://www.nlm.nih.gov/nichsr/hta101/ta10104.html#Heading2. Retrieved July 30, 2014.
Neumann, P.J., Goldie, S.J., and Weinstein, M.C. (2000). Preference-based measures
in economic evaluation in health care. Annual review of public health 21, 587-611.
Ouyang, L., Grosse, S.D., Amendah, D.D., and Schechter, M.S. (2009). Healthcare
expenditures for privately insured people with cystic fibrosis. Pediatric pulmonology 44,
989-996.
14

Prosser, L.A., Grosse, S.D., Kemper, A.R., Tarini, B.A., and Perrin, J.M. (2012).
Decision analysis, economic evaluation, and newborn screening: challenges and
opportunities. Genetics in medicine 14, 703-712
Segel, J.E. (2006). Cost-of-Illness studies-A primer. RTI international [RTI -UNC Center
of Excellence in Health Promotion Economics]. 1-39.
Sloan, J.A., Sargent, D.J., Novotny, P.J., Decker, P.A., Marks, R.S., and Nelson, H.
(2014). Calibration of quality-adjusted life years for oncology clinical trials. Journal of
pain and symptom management 47, 1091-1099.
Stinnett, A.A., Mullahy, J. (1998). Net health benefits: a new framework for the analysis
of uncertainty in cost-effectiveness analysis. Med Decis Making 18(2), S68-80.
Straus, S.E., Richardson, W.S., Glasziou, P., and Haynes, R.B. (2011). Evidence-
Based Medicine: How to Practice and Teach It. (4th ed. New York, NY: Churchill
Livingstone Elsevier).
Taylor, D.H., Jr., and Sloan, F.A. (2000). How much do persons with Alzheimer's
disease cost Medicare? Journal of the American Geriatrics Society 48, 639-646.
Torrance, G.W., and Feeny, D. (1989). Utilities and quality-adjusted life years.
International journal of technology assessment in health care 5, 559-575.
Zhang, X., Miller, L., Max, W., and Rice, D.P. (1999). Cost of smoking to the Medicare
program, 1993. Health care financing review 20, 179-196.
Zhou, F., Santoli, J., Messonnier, M.L., Yusuf, H.R., Shefer, A., Chu, S.Y., Rodewald, L.,
and Harpaz, R. (2005). Economic evaluation of the 7-vaccine routine childhood
immunization schedule in the United States, 2001. Archives of pediatrics & adolescent
medicine 159, 1136-1144.

15

CHAPTER 2: COST-BENEFIT ANALYSIS OF IN-HOSPITAL INFLUENZA
VACCINATION OF POSTPARTUM WOMEN
1

Chapter-Two Abstract

OBJECTIVE: To estimate the potential economic benefits associated with hospital-
based postpartum influenza vaccination. METHODS: We constructed a decision
analysis model to estimate the potential cost benefit of this strategy from both a societal
perspective and a third-party perspective. We included a hypothetical cohort of 1.47
million U.S. postpartum women, assuming an influenza season beginning September 1
and ending April 30. Probabilities and costs were derived from published literature,
Centers for Disease Control and Prevention data, and expert recommendations. We
used one-way and two-way sensitivity analyses. All cost estimates were inflated to year
2010 U.S. dollars and discounted at a 3% annual discount rate. RESULTS: From the
societal perceptive, the expected costs per vaccinated and unvaccinated mother were
$328.45 and $341.02 respectively, resulting in an expected net benefit of $12.57 per
vaccinated mother. The overall savings in the cohort were predicted to range from $3.69
to $14.75 million, depending on the vaccination coverage rate. This strategy would be
cost-beneficial, holding all other variables to the base case, if the annual maternal
influenza attack rate is more than 2.8%, influenza vaccine efficacy is more than 47%, or
if vaccine acquisition and administration cost per dose are less than $32.78.

1
The final, definitive version of this paper has been published in Obstet Gynecol journal, Vol.
119:306-14, 2012 by Lippincott Williams & Wilkins, Inc. All rights reserved.
16

The strategy would not generate net savings from the third-party perspective. Sensitivity
analyses were robust, but disease incidence and vaccine efficacy were important
drivers. CONCLUSION: Our model suggests that postpartum influenza vaccination is a
cost-beneficial approach for prevention of maternal and infantile influenza from a
societal perspective.

17

Chapter-Two Background
Influenza is a common and highly contagious respiratory illness. In the U.S.,
influenza epidemics occur every year resulting in approximately 25,000 deaths and
226,000 hospitalizations annually (Izurieta et al., 2000; Thompson et al., 2004).
Hospitalization of infants <6 months of age for influenza occurs at rates greater than
other high-risk groups (CDC, 2011; Poehling et al., 2006), and mortality due to this
infection remains an important problem (Bhat et al., 2005).

The Advisory Committee on Immunization Practices recommends universal
influenza vaccination for all persons ≥6 months of age (CDC, 2011). Infants <6 months
of age, therefore, are unprotected against infection through primary immunization.
Immunization of pregnant women during influenza season is recommended to protect
newborns by conferring transplacentally transferred maternal antibodies to them
(Lindsay et al., 2006; Roberts et al., 2006). From 1995-2008, influenza vaccine
coverage rates during pregnancy had been less than 25% (Schrag et al., 2003), and
national recommendations also encouraged vaccinating women in the postpartum
period if they had not received the vaccine that influenza season.

Influenza vaccination of preschool children has been shown to be cost-saving
(Prosser et al., 2011). To our knowledge only one study has evaluated the economic
impact of vaccinating pregnant women on disease prevention in the newborn, but this
study also considered infection in the neonate as independent of maternal influenza
(Beigi et al., 2009). A recent case-control study documented a clinical benefit of pre-
18

and/or peripartum influenza vaccination in the prevention of influenza-associated
hospitalization in infants <6 months of age (Benowitz et al., 2010). No economic data
have been reported to assess the cost-benefit of a national postpartum vaccination
strategy. We performed this study to estimate the economic burden associated with this
strategy.

Chapter-Two Methods
Data Source
We searched the MEDLINE database with the keywords influenza vaccine and
cost-effectiveness analysis; household transmission of influenza to infants; and/or direct
and indirect costs of influenza to retrieve the related literature published during 1996–
2011. References from these publications were also retrieved, as appropriate. We used
probabilities and costs derived from our review of published literature, Centers for
Disease Control and Prevention (CDC) data, and expert panel recommendations. The
study was exempt from IRB review, as it relied only on publicly available secondary
literature sources.

Economic Model and Study Population
A decision analysis model was constructed to calculate the costs, benefits, and
potential cost-benefit of a postpartum influenza vaccination strategy during the
immediate postpartum period prior to hospital discharge. Decision tree modeling was
used in this study as it is best suited for acute care episodes (e.g., seasonal influenza in
general is an acute illness with a convalescent period lasting around 5-15 days),
19

compared to other types of pharmacoeconomic models (e.g., Markov models or
discrete-event simulation models) (Hay, 2004). The provisional number of live births in
the U.S. for June 2009-June 2010 period was about 4 million (National Center for
Health Statistics, 2011). In this study, we only included postpartum women who have
not received influenza vaccine prior to delivery in the influenza season and these
postpartum women would be vaccinated in the immediate postpartum period. We
assumed the national rate for pre-delivery vaccine (i.e., vaccine coverage rate in the
pregnant population) to be as high as 44.8% (CDC, 2011). National surveillance data
indicate that influenza virus activity usually begins in September, peaks in January or
February, and continues to April (France et al., 2006). We assumed no influenza
vaccine supply availability from May 1 to August 30 (CDC, 2011; Neuzil et al., 2000),
thus the benefit of postpartum vaccination was not applied for women who delivered
during this time period. Therefore, a cohort of 1.47 million (i.e., 4 million multiplied by
55% multiplied by 2/3) healthy postpartum mothers in the U.S. were included in the
model assuming an influenza season beginning September 1 and ending April 30. We
used the cost of trivalent injectable inactivated vaccine (TIV) in this study as a
conservative cost estimate, since this vaccine is more widely stocked in hospitals (CDC,
2011). Figure 2.1.A depicts the overall structure of the model for a postpartum influenza
vaccination strategy, and Figure 2.1.B shows the maternal influenza sub-tree. In this
decision tree model, there were two major arms: vaccine arm (birth mothers were
vaccinated with TIV), and no vaccine arm (birth mothers received no vaccine). It was
modeled that no infant <6 months of age would receive the influenza vaccine, as per
current national recommendations (CDC, 2011). Each pathway in the model was
20

defined by probabilities of an event to occur and the costs of clinical outcomes. We
included four possible outcomes after the influenza virus infection in both birth mothers
and infants: 1) symptomatic or asymptomatic influenza, but not medically attended,
regardless; 2) outpatient visit only; 3) hospitalization; 4) death.

The base-case analysis was conducted from a societal perspective, which
included all estimated direct and indirect costs. We also analyzed the potential cost-
benefit of postpartum vaccination including direct medical costs only, which is
particularly relevant for a healthcare system perspective. The time frame was one year.
Modeling was performed using Microsoft Excel 2007 (Microsoft, Redmond, WA). All
costs and benefits were discounted at a 3% annual discount rate. Results were reported
in 2010 US dollars.

Probabilities
As noted in Figure 2.1.A, vaccination of a birth mother will result in a limited
number of potential outcomes related to protection against clinical influenza and the
potential for adverse events associated with vaccination. Only those mothers that
receive vaccine may experience a vaccine-related adverse event, including local or
systemic reactions, anaphylaxis, and Guillain-Barré Syndrome (GBS). The probabilities
of experiencing such an event are derived from the published literature (Haber et al.,
2004; Prosser et al., 2008; Vellozzi et al., 2009). Local reactions, which most commonly
include soreness at the vaccination site lasting 2-3 days, are rarely medically attended
(CDC, 2011). Systemic reactions most commonly include fever, headache, or myalgias.
21

Overt anaphylaxis is rare and may manifest as respiratory difficulty, urticaria, vomiting,
hypotension, decreased consciousness, and/or shock (CDC, 2011; Vellozzi et al., 2009).
Lastly, influenza vaccine has been previously reported to be associated with
subsequent development of GBS at a rate of approximately one additional case of GBS
per 1 million vaccines in some years (Haber et al., 2004).

Each birth mother has a definable probability of developing clinical influenza,
which will be lessened with receipt of vaccine. These probabilities were generated using
national and regional data of disease incidence and vaccine efficacy, over several years,
for adults 18-49 years of age (Bridges et al., 2000; Luce et al., 2008; Molinari et al.,
2007; Prosser et al., 2008). Mothers infected with influenza had a chance of
hospitalization for severe disease and we assumed only hospitalized mothers were at
risk for death (Prosser et al., 2008). The probability of cases not medically attended was
calculated to be equal to the influenza attack rate minus the sum of the probabilities of
influenza-associated outpatient visits only and hospitalizations (Molinari et al., 2007).

We assumed that the probability of developing influenza in newborns was, in part,
dependent on the infectious status of the mother and others in the household. Previous
work has shown that the infection rate among household contacts of an index case
increases significantly with the number of older siblings, and the attack rate of
respiratory illness among unvaccinated household contacts (aged 0-4 years of age) of
children attending day care is approximately 50%~69% (Glezen et al., 1997; Hurwitz et
al., 2000). Thus, the probability of a newborn developing influenza when exposed to an
22

infected mother was estimated by multiplying the probability of maternal influenza by the
attack rate in infants with maternal influenza. In this study, we assumed the base case
attack rate in infants with maternal influenza to be 59% (the midpoint of a range of 50%-
69%) (CDC, 2011; Glezen et al., 1997; Hurwitz et al., 2000). The probability of
developing influenza in infants born to mothers without disease was estimated by
multiplying the disease probability in infants by the probability of no maternal influenza:
(1- probability of maternal influenza). In this pathway, the disease probability in infants is
equal to the reported annual attack rate of influenza for infants (Molinari et al., 2007).
Rates of outpatient visits and hospitalizations attributable to influenza were based on
average excess seasonal rates of outpatient visits and hospitalizations using CDC
influenza surveillance data and other data from the published literature (CDC, 2011;
Izurieta et al., 2000; Neuzil et al., 2000; Poehling et al., 2006; Thompson et al., 2004).
The probability of cases not medically attended was calculated to be equal to the
influenza attack rate minus the sum of the probabilities of influenza-associated
outpatient visits only and hospitalizations (Molinari et al., 2007).The mortality rate for
infants was obtained from two previous studies (Beigi et al., 2009; Bhat et al., 2005). All
the base case probability inputs and their ranges for sensitivity analysis are listed in
Table 2.1.

Healthcare Resource Costs
Direct and indirect costs for both mothers and infants are noted in Table 2.2. All
costs in the model were adjusted for inflation to 2010 US dollars using the medical care
services component of the consumer price index when necessary.
23

Direct costs
Direct costs included vaccine acquisition and administration costs, vaccine-
associated adverse events costs, and expenses to treat complications of clinical
influenza for both mothers and infants (Molinari et al., 2007; Prosser et al., 2008).
Vaccine costs for birth mothers were based on the CDC vaccine price list for TIV (CDC
Vaccine Price List, 2010). The costs of adverse events with influenza vaccine were
derived from a recently published economic study (Prosser et al., 2008). We estimated
over-the-counter (OTC) medication costs using weighted mean values from published
clinical trial data (Bridges et al., 2000; Nichol et al., 2003), and used OTC drug costs as
the direct costs to treat cases that were not medically attended. For each outpatient
case, direct medical costs included pharmaceutical claims and outpatient claims.
Outpatient claims were the sum of office visits, laboratory tests, consult fees, outpatient
procedures, and prescription medications (Molinari et al., 2007; Prosser et al., 2008).
We estimated direct medical expenses of each inpatient case, including pharmaceutical,
outpatient and hospitalization costs, from published economic studies (Bridges et al.,
2000; Luce et al., 2008; Molinari et al., 2007; Prosser et al., 2008). In this analysis, we
assumed all adults or infants who died received some treatments before death. So the
direct medical costs for a death included all pharmaceutical claims, outpatient, and
inpatient costs based on a CDC study in which the authors estimated such costs using
Medstat Marketscan health insurance claims database capturing cases from 2000-2004
four influenza seasons (Molinari et al., 2007).

24

Indirect costs
Indirect costs included the unpaid opportunity costs from the mothers’ or
caregivers’ time loss (working time and personal time) due to their own illness or their
infants’ disease, and the indirect societal costs of death. For influenza in infants, indirect
costs sustained by caregivers were calculated by using estimates of 2 hours per office
visit (including travel and waiting) and 8 hours per day for hospitalizations based on
length of hospital stay (LOS) published in literature (Molinari et al., 2007; Principi et al.,
2003; Thompson et al., 2004). For birth mothers, indirect costs were estimated to
include costs of time lost from work and personal activity time lost due to adults' own
influenza (Bridges et al., 2000; CDC, 2011; Molinari et al., 2007; Nichol et al., 2003;
Prosser et al., 2008). The average compensation per hour was $29.71 in 2010, as
reported by the U.S. Bureau of Labor Statistics (BLS, 2011). The indirect societal cost
per death was based on value of a statistical life estimates, which included the value of
lost productivity value and the social value placed on human life (Molinari et al., 2007).
We assumed no additional indirect costs incurred for administration of vaccine to birth
mothers, as the vaccine would likely be given during the immediate postpartum period
in the hospital, not during a separately scheduled medical visit.

Chapter-Two Results
Base case results
From the societal perspective, the average cost in the vaccine administration and
no vaccine arms were estimated at $328.45 and $341.02 per mother, respectively. Each
term includes annual expected costs associated with the mother and her infant(s). For
25

example, the expected cost per unvaccinated mother was calculated by summation of
(probability of each potential outcome associated with maternal/infantile influenza
multiplied by its cost respectively). In the base case, our analysis suggests an expected
net societal benefit (ENB) of $12.57 per postpartum vaccinated mother, compared to no
vaccination. In our hypothetical cohort of 1.47 million birth mothers in the U.S., the
overall expected net benefit was calculated by the following equation: Overall ENB in
the cohort = ($12.57)*(1.47 million)*(vaccination coverage rate). Here, we assumed the
coverage rate for postpartum influenza vaccination to vary from 20% to 80% (Yeh SH,
et al. Effectiveness of hospital-based procedures on postpartum vaccination with
tetanus toxoid, reduced diphtheria toxoid and acellular pertussis and seasonal
influenza.49th Annual Meeting of the Infectious Diseases Society of America, October
20-23, 2011), and then the overall annual ENB in the cohort of US birth mothers was
estimated to be $3.69 to $14.75 million.

From the healthcare system perspective (direct medical costs only), the average
cost in vaccine and control arms were predicted to be $197.6 and $183.9 per birth
mother, respectively. The ENB value of -$13.70 per vaccinated mother indicates that
postpartum vaccination would not generate net savings in this case compared to no
vaccination.

Sensitivity analyses
From year to year, the circulating influenza virus strains may change, which
usually significantly alters the disease incidence, clinical morbidity, vaccine efficacy,
26

and/or other variables we have included in our model. As such, we performed one-way
sensitivity analyses on selected variables. Ranges were selected based on the high and
low values from previously reported studies. A tornado diagram in Figure 2.2 shows the
impact of varying model estimates on ENB per vaccinated mother from a societal
perspective. The variables that elicit the greatest impact on the model are the annual
attack rate in birth mothers and vaccine efficacy, both of which affect the direct and
indirect costs. For example, a 70% decrease in attack rate among mothers (adults)
results in a 120% decrease in ENB per mother. If vaccine virus does not match the
circulating strain, vaccine efficacy may in fact approach zero; our base case assumes a
match between the vaccine strain and the strain which circulates in any given year,
resulting in vaccine efficacy of 50% to 86% (Poehling et al., 2006). For example, a
change from the base case of 73% efficacy to 50% efficacy nearly eliminates the
expected net benefit from $12.6 to $1.7 per mother. The third variable that exhibits
significant impact on the model is vaccine acquisition and administration cost per dose,
which affects the direct costs only. Other parameters, such as mortality rate,
nonmedical costs of cases that are not medically attended, indirect societal cost per
death, and incidence of influenza-associated hospitalization among birth mothers also
influence the ENB per vaccine (Figure 2.2). Percent change in outcome (ENB per
mother) divided by percent change in each key parameter, representing elasticity, are
reported in Table 2.3.

As the timing, severity, and length of the influenza epidemic can vary substantially
from season to season, the attack rate in birth mothers could range higher than 10% in
27

some years when influenza viruses mismatch the vaccine strain. Two-way sensitivity
analyses were performed using 2%~20% attack rates in birth mothers and 50% versus
80% efficacies for influenza vaccine, with other parameters remaining unchanged in the
base-case values. The results are presented in Table 2.4. From the societal perspective
in which all direct and indirect costs were included, assuming 50% vaccine efficacy,
then the resulting ENB per vaccinated mother ranged from -$8.7 to $31.6 as the attack
rate ranged from 2% to 20%. If we assumed 80% vaccine efficacy, then the resulting
ENB per vaccinated mother ranged from -$0.6 to $63.9 as the attack rate ranged from 2%
to 20%. From the third-party perspective, the trends were similar but most of the ENB
per mother were negative values indicating that the average cost per vaccinated mother
exceeded the average cost per unvaccinated mother.

We performed threshold sensitivity analyses from the societal perspective (using
ENB per vaccinated mother = $0) to determine the breakeven point for the variables
noted to most influence the results in our one-way sensitivity analysis (annual attack
rate in birth mothers, vaccine efficacy, and vaccine acquisition and administration cost
per dose). Our results suggest that this strategy would be cost-saving (ENB > $0 per
mother), holding all other variables to the base case, if the annual attack rate among
birth mothers exceeds 2.8% , influenza vaccine efficacy is greater than 47%, and
vaccination acquisition and administration costs per dose fall below $32.78.

28

Chapter-Two Discussion
We believe this study is the first to evaluate the cost-benefit of a hospital-based
postpartum influenza vaccination strategy. Our model included data-driven vaccine
coverage estimates, and the concept that influenza virus infection in newborns is at
least partially dependent on the infectious status of their mothers. The results suggest
that from a societal perspective, postpartum vaccination of birth mothers is likely to be
cost-beneficial if key parameters fall within reasonable ranges, which in a typical
influenza year is highly likely. When compared with no influenza vaccination, our base-
case data suggest an ENB of $12.57 per vaccinated mother from the societal
perspective, and an ENB of $-13.70 from the third party perspective. Others have
shown the potential for clinical benefit of this vaccination strategy (Benowitz et al., 2010),
and current CDC recommendations endorse vaccination in the postpartum period if the
preferred strategy of vaccination during pregnancy was not followed (CDC, 2011). We
believe our analysis provides a useful and timely economic assessment for its
implementation as well.

Previous work indicated cost-effectiveness of influenza vaccination in pregnant
women (Beigi et al., 2009; Roberts et al., 2006), when rates of influenza vaccine
coverage were low (about 25%) (Schrag et al., 2003). A CDC survey completed in
November 2010 found that mid-season influenza vaccination coverage among pregnant
women was 44.8 % 14. Although recent CDC data revealed a sustained coverage rate
of 44% in pregnancy for the 2010-11 influenza season, studies have shown that
provider recommendation, public awareness, the intensity of the annual epidemic and
29

concerns about vaccine safety are strong predictors of vaccination in other adult
populations (Chi and Neuzil, 2004). When pre-delivery vaccine rates vary from 25% to
45%, a range of 1.47~2.0 million healthy postpartum women would be eligible for the
postpartum strategy. This would realize an overall annual societal cost savings of $3.69
to $20.11 million, depending on the influenza vaccine coverage rates in pregnant and
postpartum population.

The postpartum period provides an opportunity for targeted vaccination that can
be easily incorporated into postpartum routine care. This avoids the indirect costs
associated with waiting and transportation time incurred for a separate healthcare visit
(Schrag et al., 2003). Previous work indicated that the average recipient time lost to
receive influenza vaccine in the doctor’s office setting was 1.24 hours (Bridges et al.,
2000; Prosser et al., 2008). In our model, adding this cost reveals vaccination to be
cost-incurring with an ENB of -$24.3 per vaccinated mother, using an average
compensation of $29.71 per hour in 2010 (BLS, 2011).

The biggest driver of our model from the societal perspective was the annual
attack rate of influenza in birth mothers. In the base case, we derived the annual attack
rate of influenza of 6.6% among adults from published randomized controlled trial data
(Bridges et al., 2000), with a range of 2%~10% in the one-way sensitivity analyses from
CDC-reported data (CDC, 2011; Molinari et al., 2007). However, circulating influenza
virus strains may change from year to year as do attack rates and clinical morbidity. We
therefore evaluated a wider range for the attack rate (2%~20%) in two-way sensitivity
30

analyses, and performed threshold sensitivity analyses. We found that the postpartum
influenza vaccination strategy would generate net savings with an ENB > $0 per
vaccinee, if the annual attack rate among birth mothers exceeds 2.8% with other
parameters remaining unchanged in the base-case scenario.

We acknowledge that we lack precision with some of the point estimates in our
model. For example, few data exist for estimating maternal disease incidence and
disease complications in the months following a child’s birth; we used available data
from the general population for women of child-bearing age. We may have
underestimated the potential cost savings if postpartum women are more susceptible to
influenza or serious influenza-associated complications than we assumed (Lindsay et
al., 2006). In addition, viral virulence and the duration of the annual epidemics can vary
substantially from year to year. We therefore used combined data over several years,
and assigned broad ranges for the key parameters, such as disease attack rate and
vaccine efficacy in the sensitivity analyses. Except for the annual attack rate in birth
mothers, no reasonable variation in these parameters decreased the ENB per vaccinee
below $0 in one-way sensitivity analyses (Figure 2.2). Future work may further clarify
these issues, but our model was robust to wide ranges of parameter values in the
sensitivity analyses.

Conclusions
The primary approach for prevention of influenza disease in young infants
currently is maternal vaccination during pregnancy. Our analysis of a national in-hospital
31

postpartum influenza vaccination strategy, however, indicates cost benefit which we
believe could be helpful to decision-makers and may encourage use of influenza
vaccine in the postpartum period to complement current pre-partum vaccination efforts.

32

Tables and Figures
Table 2.1: Probability Variables Used in the Model
Probability Base case Range
Vaccine(TIV)-associated adverse events, %
Local reaction (not medically attended) 44 10-64
Systemic reaction 1.1 0-2.2
Anaphylaxis 0.000025 0-0.00005
Guillain-Barré Syndrome (GBS) 0.0001 0-0.0002
Vaccine Efficacy, % 73 50-86
Probability of Influenza (estimated annual attack rate), %
In birth mothers 6.6 2-10
In infants 20.3 12.4-30
In infants whose mothers had influenza 59.3 50-69
Incidence of influenza associated cases in mothers
Outpatient visit 0.021 0.018-0.023
Hospitalization(cases/100,000) 28 14-40
Death (cases/100,000) 0.22 0.16-0.3
Not medically attended 0.045 0.03-0.06
Incidence of influenza associated cases in infants
Outpatient visit 0.10 0.08-0.15
Hospitalization (cases/100,000) 1040 0.0072-0.013
Death (cases/100,000) 0.88 0.52-1.39
Not medically attended 0.09 0.03-0.11
TIV: trivalent injectable inactivated vaccine

33

Table 2.2: Base-case Parameters for Direct and Indirect Costs in 2010 US dollars
Variable Base case input
Direct costs
Vaccine costs
Acquisition cost/dose $9.30
Administration/dose $11.15
Medical costs of vaccine-associated adverse events
Systemic reaction $163
Anaphylaxis $543
Guillain-Barré Syndrome (GBS) $79,069
Medical costs for birth mothers with influenza
Outpatient visit $163
Hospitalization $24,945
Death $100,458
Over-the-counter (OTC) medications* $13.3
Medical costs for infants with influenza
Outpatient visit $220
Hospitalization $14,318
Death $37,925
Over-the-counter (OTC) medications* $13.3
Indirect costs
Average compensation per hour $29.71
Nonmedical costs for vaccine-associated adverse events
Anaphylaxis (days lost) 2
Guillain-Barré Syndrome (GBS) (days lost) 39
Nonmedical costs for birth mothers with influenza
Outpatient visit (days lost) 1.3
Hospitalization (days lost) 6.96
Cases not medically attended (days lost) 0.5
Indirect societal cost per death $8.1 million
Nonmedical costs sustained by caregiver in the care of infants with
influenza

Outpatient visit (days lost) 3
Hospitalization (days lost) 4.9
Cases not medically attended (days lost) 1
Indirect societal cost per death $5.9 million
*OTC medication costs were used as the direct costs to treat cases not medically attended

34

Table 2.3: One-way Sensitivity Analysis of Selected Parameters from a Societal Perspective
Parameter Value % Change in
Parameter
ENB/mother % Change in
ENB/mother
Elasticity*
Upper 86% 18% $18.8 50% 2.78
Influenza vaccine efficacy Base case 73% Baseline $12.6 Baseline Baseline
(%) Lower 50% -32% $1.7 -86% 2.74
Upper 10% 52% $23.7 89% 1.72
Attack rate in birth mothers Base case 6.6% Baseline $12.6 Baseline Baseline
(%) Lower 2% -70% -$2.5 -120% 1.72

Vaccine acquisition and
administration cost/dose
Upper $27.4 36% $5.4 -57%
-1.60
Base case $20.4 Baseline $12.6 Baseline Baseline
Lower $14.1 -30% $18.7 49%
-1.60
Incidence of influenza-
associated hospitalization
among birth mothers
(cases/100,000)
Upper 40 44% $15.0 19% 0.44
Base case 28 Baseline $12.6 Baseline Baseline
Lower 14 -49% $9.9 -21% 0.43
Influenza-associated
mortality rate among birth
mothers
(cases/100,000)
Upper 0.30 36% $17.3 38% 1.04
Base case 0.22 Baseline $12.6 Baseline Baseline
Lower 0.16 -27% $9.0 -28% 1.04
Nonmedical costs of
influenza associated
outpatient visit among birth
mothers (days)
Upper 2.5 92% $16.9 34% 0.37
Base case 1.3 Baseline $12.6 Baseline Baseline
Lower 0.6 -54% $10.1 -20% 0.37
Nonmedical costs of cases
not medically attended
among birth mothers(days)
Upper 1.0 100% $16.5 31% 0.31
Base case 0.5 Baseline $12.6 Baseline Baseline
Lower 0.03 -94% $8.9 -29% 0.31

Indirect societal cost per
death (birth mothers)
Upper $10
million
25% $15.8 26% 1.03
Base case $8
million
Baseline $12.6 Baseline Baseline
Lower $6
million
-25% $9.3 -26% 1.03
Nonmedical costs of cases
not medically attended
disease among infants
(days)
Upper 2.0 100% $17.0 36% 0.36
Base case 1 Baseline $12.6 Baseline Baseline
Lower 0.5 -50% $10.3 -18% 0.35

Attack rate in infants
Upper 30% 48% $13.5 7% 0.15
Base case 20.3% Baseline $12.6 Baseline Baseline
Lower 12.4% -39% $11.4 -9% 0.24

Attack rate in infants with
maternal influenza
Upper 69% 16% $13.8 10% 0.60
Base case 59.3% Baseline $12.6 Baseline Baseline
Lower 50% -16% $11.5 -9% 0.54
ENB: expected net benefit
* % change in outcome (ENB per mother) divided by % change in parameter input
35

Table 2.4: Two-way Sensitivity Analyses of Annual Attack Rate and Vaccine Efficacy
Expected net benefit (ENB)/ vaccinated mother in 2010 US$
All direct and indirect costs
(from a societal perspective)
Direct costs only (from a
healthcare system perspective)
Base case $12.6 -$13.7
50% vaccine efficacy, at an annual
attack rate in birth mothers of

2% -$8.7 -$20.3
5% -$1.9 -$17.7
10% $9.3 -$13.5
15% $20.5 -$9.2
20% $31.6 -$5.0
80% vaccine efficacy, at an annual
attack rate in birth mothers of

2% -$0.6 -$19.2
5% $10.2 -$15.1
10% $28.1 -$8.3
15% $46.0 -$1.5
20% $63.9 $5.3

36

Figure 2.1: Overall structure of a postpartum influenza vaccination strategy

Fig.2.1. A. Overall structure of a postpartum influenza vaccination strategy. B. Maternal
influenza subtree. Four possible outcomes after the influenza infection in both birth mothers and
infants: 1) not medically attended, 2) outpatient only, 3) hospitalization, and 4) death.
Gray shading indicates maternal influenza.

37

Figure 2.2: Tornado diagram of one-way sensitivity analyses

Fig.2.2. Tornado diagram of one-way sensitivity analyses on the effect of range of individual
parameters on the expected net societal benefit per mother from a societal perspective. The
vertical line in the diagram represents the base-case expected net societal benefit of $12.57 per
postpartum vaccinated mother. Each horizontal bar indicates the range of expected net societal
benefit per postpartum vaccinated mother with the lower and upper values of each parameter
shown in Table 3.

38

Chapter-Two References
Beigi, R.H., Wiringa, A.E., Bailey, R.R., Assi, T.M., and Lee, B.Y. (2009). Economic
value of seasonal and pandemic influenza vaccination during pregnancy. Clin Infect Dis
49, 1784-1792.
Benowitz, I., Esposito, D.B., Gracey, K.D., Shapiro, E.D., and Vazquez, M. (2010).
Influenza vaccine given to pregnant women reduces hospitalization due to influenza in
their infants. Clin Infect Dis 51, 1355-1361.
Bhat, N., Wright, J.G., Broder, K.R., Murray, E.L., Greenberg, M.E., Glover, M.J., Likos,
A.M., Posey, D.L., Klimov, A., Lindstrom, S.E., et al. (2005). Influenza-associated
deaths among children in the United States, 2003-2004. N Engl J Med 353, 2559-2567.
Bridges, C.B., Thompson, W.W., Meltzer, M.I., Reeve, G.R., Talamonti, W.J., Cox, N.J.,
Lilac, H.A., Hall, H., Klimov, A., and Fukuda, K. (2000). Effectiveness and cost-benefit of
influenza vaccination of healthy working adults: A randomized controlled trial. JAMA 284,
1655-1663.
Centers for Disease Control and Prevention (CDC). Prevention and Control of Influenza
with Vaccines: Recommendations of the Advisory Committee on Immunization
Practices (ACIP). MMWR 2011; 60(33):1128-1132.
Centers for Disease Control and Prevention (CDC). [cited May 24, 2011]; Available from:
www.cdc.gov/flu/pdf/vaccination/DingInternetPanelSurveyPregnantWomen.pdf
Centers for Disease Control and Prevention (CDC) Vaccine Price List. Prices last
updated: October 6, 2010. [cited March 24, 2011]; Available from:
http://www.cdc.gov/vaccines/programs/vfc/cdc-vac-price-list.htm
Chi, R.C., and Neuzil, K.M. (2004). The association of sociodemographic factors and
patient attitudes on influenza vaccination rates in older persons. Am J Med Sci 327,
113-117.
39

France, E.K., Smith-Ray, R., McClure, D., Hambidge, S., Xu, S., Yamasaki, K., Shay, D.,
Weintraub, E., Fry, A.M., Black, S.B., et al. (2006). Impact of maternal influenza
vaccination during pregnancy on the incidence of acute respiratory illness visits among
infants. Arch Pediatr Adolesc Med 160, 1277-1283.
Glezen, W.P., Taber, L.H., Frank, A.L., Gruber, W.C., and Piedra, P.A. (1997). Influenza
virus infections in infants. Pediatr Infect Dis J 16, 1065-1068.
Haber, P., DeStefano, F., Angulo, F.J., Iskander, J., Shadomy, S.V., Weintraub, E., and
Chen, R.T. (2004). Guillain-Barre syndrome following influenza vaccination. JAMA 292,
2478-2481.
Hay, J.W. (2004). Evaluation and review of pharmacoeconomic models. Expert Opin
Pharmacother 5, 1867-1880.
Hurwitz, E.S., Haber, M., Chang, A., Shope, T., Teo, S., Ginsberg, M., Waecker, N., and
Cox, N.J. (2000). Effectiveness of influenza vaccination of day care children in reducing
influenza-related morbidity among household contacts. JAMA 284, 1677-1682.
Izurieta, H.S., Thompson, W.W., Kramarz, P., Shay, D.K., Davis, R.L., DeStefano, F.,
Black, S., Shinefield, H., and Fukuda, K. (2000). Influenza and the rates of
hospitalization for respiratory disease among infants and young children. N Engl J Med
342, 232-239.
Lindsay, L., Jackson, L.A., Savitz, D.A., Weber, D.J., Koch, G.G., Kong, L., and Guess,
H.A. (2006). Community influenza activity and risk of acute influenza-like illness
episodes among healthy unvaccinated pregnant and postpartum women. Am J
Epidemiol 163, 838-848.
Luce, B.R., Nichol, K.L., Belshe, R.B., Frick, K.D., Li, S.X., Boscoe, A., Rousculp, M.D.,
and Mahadevia, P.J. (2008). Cost-effectiveness of live attenuated influenza vaccine
versus inactivated influenza vaccine among children aged 24-59 months in the United
States. Vaccine 26, 2841-2848.
40

Lyn-Cook, R., Halm, E.A., and Wisnivesky, J.P. (2007). Determinants of adherence to
influenza vaccination among inner-city adults with persistent asthma. Prim Care Respir
J 16 (4), 229–235.
Molinari, N.A., Ortega-Sanchez, I.R., Messonnier, M.L., Thompson, W.W., Wortley,
P.M., Weintraub, E., and Bridges, C.B. (2007). The annual impact of seasonal influenza
in the US: measuring disease burden and costs. Vaccine 25, 5086-5096.
National Center for Health Statistics. Recent Trends in Births and Fertility Rates
Through June 2010. [cited March 22, 2011]; Available from:
http://www.cdc.gov/nchs/data/hestat/births2010/births2010.pdf
Neuzil, K.M., Mellen, B.G., Wright, P.F., Mitchel, E.F., Jr., and Griffin, M.R. (2000). The
effect of influenza on hospitalizations, outpatient visits, and courses of antibiotics in
children. N Engl J Med 342, 225-231.
Nichol, K.L., Mallon, K.P., and Mendelman, P.M. (2003). Cost benefit of influenza
vaccination in healthy, working adults: an economic analysis based on the results of a
clinical trial of trivalent live attenuated influenza virus vaccine. Vaccine 21, 2207-2217.
Poehling, K.A., Edwards, K.M., Weinberg, G.A., Szilagyi, P., Staat, M.A., Iwane, M.K.,
Bridges, C.B., Grijalva, C.G., Zhu, Y., Bernstein, D.I., et al. (2006). The
underrecognized burden of influenza in young children. N Engl J Med 355, 31-40.
Principi, N., Esposito, S., Marchisio, P., Gasparini, R., and Crovari, P. (2003).
Socioeconomic impact of influenza on healthy children and their families. Pediatr Infect
Dis J 22, S207-210.
Prosser, L.A., Meltzer, M.I., Fiore, A., Epperson, S., Bridges, C.B., Hinrichsen, V., and
Lieu, T.A. (2011). Effects of adverse events on the projected population benefits and
cost-effectiveness of using live attenuated influenza vaccine in children aged 6 months
to 4 years. Arch Pediatr Adolesc Med 165, 112-118.
41

Prosser, L.A., O'Brien, M.A., Molinari, N.A., Hohman, K.H., Nichol, K.L., Messonnier,
M.L., and Lieu, T.A. (2008). Non-traditional settings for influenza vaccination of adults:
costs and cost effectiveness. Pharmacoeconomics 26, 163-178.
Roberts, S., Hollier, L.M., Sheffield, J., Laibl, V., and Wendel, G.D., Jr. (2006). Cost-
effectiveness of universal influenza vaccination in a pregnant population. Obstet
Gynecol 107, 1323-1329.
Schrag, S.J., Fiore, A.E., Gonik, B., Malik, T., Reef, S., Singleton, J.A., Schuchat, A.,
and Schulkin, J. (2003). Vaccination and perinatal infection prevention practices among
obstetrician-gynecologists. Obstet Gynecol 101, 704-710.
Thompson, W.W., Shay, D.K., Weintraub, E., Brammer, L., Bridges, C.B., Cox, N.J.,
and Fukuda, K. (2004). Influenza-associated hospitalizations in the United States.
JAMA 292, 1333-1340.
U.S. Department of Labor Bureau of Labor Statistics. [cited March 22, 2011]; Available
from: http://www.bls.gov/news.release/ecec.nr0.htm
Vellozzi, C., Burwen, D.R., Dobardzic, A., Ball, R., Walton, K., and Haber, P. (2009).
Safety of trivalent inactivated influenza vaccines in adults: background for pandemic
influenza vaccine safety monitoring. Vaccine 27, 2114-2120.

42

CHAPTER 3: COST-BENEFIT ANALYSIS OF HOSPITAL BASED
POSTPARTUM VACCINATION WITH COMBINED TETANUS TOXOID,
REDUCED DIPHTHERIA TOXOID, AND ACELLULAR PERTUSSIS
VACCINE (TDAP)
2

Chapter-Three Abstract

Objective: To assess the economic benefits associated with hospital-based postpartum
Tdap (combined tetanus toxoid, reduced diphtheria toxoid, and acellular pertussis)
vaccination. Methods: A decision tree model was constructed to calculate the potential
cost–benefit of this strategy from both a health care system and a societal perspective.
Probabilities and costs were derived from published literature, data reported to Centers
for Disease Control and Prevention, and recommendations from expert panels. The
maternal vaccination protection period for infants was defined as 7 months, and 10
years of waning immunity following Tdap for birth mothers was estimated in the model.
All cost estimates were inflated to year 2012 US dollars and discounted at a 3% annual
discount rate. Results: In the base case from a societal perspective, the expected costs
per vaccinated and unvaccinated mother were estimated at $129.27 and $187.97,
respectively, suggesting an expected net benefit of $58.70 per vaccinated mother. The
overall societal benefits in the cohort of 3.6 million U.S. birth mothers ranged from
$52.8–126.8 million, depending on the vaccination coverage level. If including direct
medical costs only, the strategy would not generate net savings from a health care

2
The final, definitive version of this paper has been published in Vaccine journal, Vol. 31:2558-
2564, 2013 by Elsevier Ltd. All rights reserved.
43

system perspective. Annual incidence of pertussis in birth mothers and Tdap efficacy
exhibited substantial impact on the model as shown in one-way and two-way sensitivity
analyses. Conclusions: Although postpartum Tdap vaccination is not cost-beneficial
from a health care system perspective in the base case, this strategy is likely to
generate net benefits from a societal perspective.

44

Chapter-Three Background
Pertussis, commonly known as whooping cough, is a highly contagious respiratory
illness caused by Bordetella pertussis. Worldwide, 30-50 million cases of pertussis and
approximately 300,000 deaths occur each year. In the U.S., the Centers for Disease
Control and Prevention (CDC) reported nearly 17,000 cases of pertussis in 2009 (CDC,
2012; Coudeville et al., 2009; Crowcroft and Pebody, 2006).

Infants under 6 months of age are particularly vulnerable as they are incompletely
vaccinated until at least 6 months of age (CDC, 2012). Over 90% of pertussis-
associated deaths in the United States during 2000--2004 were among infants aged <6
months (CDC, 2011). Studies showed that household contacts of newborns, including
birth mothers, are responsible for 76%-83% of pertussis transmission among young
infants for whom a source was identified (Bisgard et al., 2004; Deen et al., 1995;
Wendelboe et al., 2007).

Since 2006, the Advisory Committee on Immunization Practices (ACIP) has
recommended Tdap (combined tetanus toxoid, reduced diphtheria toxoid, and acellular
pertussis) vaccination for close contacts of young infants, in order to provide personal
protection and to reduce the risk of transmitting pertussis to their infants (CDC, 2006;
CDC, 2008). In 2011, the ACIP recommended Tdap to be given to pregnant women
after 20 weeks of gestation, and if not given during pregnancy, to be given in the
immediate postpartum period (CDC, 2011). Several studies have shown the cost-
effectiveness of different strategies for pertussis vaccination among adolescents and
45

adults (Coudeville et al., 2009; Lee et al., 2005; Lee et al., 2007; Purdy et al., 2004).
The economic benefits solely from routine postpartum vaccination with Tdap have not
been reported. We sought to estimate the cost-benefit of such a strategy in a hospital
setting.

Chapter-Three Materials and Methods
Data Collection and Economic Model
We searched the MEDLINE database with the keywords pertussis vaccine and
cost-effectiveness analysis; household transmission of pertussis to infants; and/or direct
and indirect costs of pertussis to retrieve the relevant literature published during 1995-
2010. References from these publications were also retrieved, as appropriate.
Probabilities and costs used in this study were derived from a review of published
literature, data reported to CDC and recommendations from expert panels. Estimates of
Tdap vaccination coverage were derived from a recent prospective evaluation and a
published study about pertussis immunization in postpartum population (Healy et al.,
2009; Yeh, 2011). The study was exempt from IRB review, as it relied only on publicly
available secondary literature sources.

We constructed a decision tree model to determine the potential cost-benefit of a
postpartum Tdap strategy during the immediate postpartum period prior to hospital
discharge from a societal perspective (Gold MR, 1996). Decision tree modeling is well-
suited for acute care episodes and pertussis is an acute respiratory illness with a
convalescent period lasting no more than 3-6 months (Hay, 2004). We modeled the
46

potential maternal vaccination protection for the infant to 7 months postpartum, since
the infants do not receive a full series of pertussis immunization before 6 months of age,
with 1 month added after the 3
rd
dose of vaccine for maximal development of immunity
(CDC, 2012). The average length of protection by acellular pertussis vaccine was
estimated to be 10.5 years for adults (Lavine et al., 2010); we therefore assumed 10
years of vaccine protection for birth mothers in the model.

The provisional number of live births in the United States for June 2009-June 2010
period was about 4 million (National Center for Health Statistics, 2011). In this study, we
only included postpartum women who have not received Tdap prior to delivery and
these postpartum women would be vaccinated in the immediate postpartum period. We
assumed the national rate for pre-delivery vaccination (i.e., Tdap coverage rate in the
pregnant population) to be 10%, as 10.6% of household contacts of infants < 1year of
age reported receiving Tdap in the U.S. in 2010 (Yeh, 2011). Therefore, a hypothetical
cohort of 3.6 million (i.e., 4 million multiplied by 90%) birth mothers in the United States
were entered into the model. The overall structure of the model for a postpartum Tdap
strategy is depicted in Figure 3.1A, and the maternal pertussis sub-tree was shown in
Figure 3.1B. In this decision tree model, there were two major arms: vaccine arm (birth
mother received Tdap) and no vaccine arm (no receipt of Tdap). The overall design and
tree structure portrayed in Figure1 are consistent with clinical trials and several
published economic studies (Lee et al., 2005; Lee et al., 2007; Purdy et al., 2004; Van
der Wielen et al., 2000; Ward et al., 2005).

47

Outcome Measure
The main outcome measure in the analysis was the expected net benefit (ENB)
per vaccinated mother, which is expressed as follows: ENB per vaccinated mother =
[expected cost per unvaccinated mother in the control arm] – [expected cost per
vaccinated mother in the intervention arm]. Each term includes the expected costs
associated with the mother and her infant. For example, the expected cost per
vaccinated mother was calculated by the sum of the probability of each potential
outcome associated with maternal/infantile pertussis multiplied by its corresponding cost.
Therefore, the ENB represents the potential expected net benefit of postpartum Tdap
per mother, compared with no Tdap vaccination. The base case analysis was
conducted from a societal perspective, which included all estimated direct costs and
indirect costs. We also analyzed the cost-benefit of postpartum vaccination including
direct medical costs only, which is particularly relevant for the healthcare system
perspective. Modeling was performed with Microsoft Excel 2010 (Microsoft, Redmond,
WA). All costs were discounted at a 3% annual discount rate and results were reported
in 2012 US dollars.

Probabilities
Vaccination of a birth mother with Tdap may result in a potential for adverse
events associated with Tdap. The probabilities of experiencing adverse events,
including local reaction, systemic reaction, and anaphylaxis, were derived from
published literature (Lee et al., 2005; Lee et al., 2007; Purdy et al., 2004). Each birth
mother has a probability of developing pertussis in the following 10 years, which will be
48

reduced with receipt of Tdap. In the no vaccine arm, the probability (i.e., annual
incidence) of maternal pertussis among unvaccinated mothers was derived from clinical
trial data in adults of child-bearing age, as no specific incidence data exist in postpartum
women (Halperin et al., 2000; Van der Wielen et al., 2000; Ward et al., 2005). For
mothers in the vaccine arm, we assumed that the protection afforded by Tdap wanes
over time (Barreto et al., 2007). The 1
st
year initial vaccine efficacy was estimated at
80% and the yearly decreasing efficacies were predicted based on published data (Lee
et al., 2005; Lee et al., 2007). The disease probability for vaccinated mothers was
obtained by multiplying (1- vaccine efficacy) by the disease probability for unvaccinated
mothers for the following 10 year protection period. In the maternal pertussis subtree
(Figure 3.1B), we classified complications associated with pertussis among birth
mothers into four categories: mild cough, severe cough, pneumonia, and hospitalization.
The frequencies of such maternal complications were derived from published literature
(Lee et al., 2007; Moses et al., 2013).

We assumed that the risk of an infant acquiring pertussis was not independent of
maternal pertussis, as household contacts of newborns, including birth mothers, are the
most common identified source of pertussis in this age group (Bisgard et al., 2004;
Deen et al., 1995; Wendelboe et al., 2007). The risk of an infant developing pertussis
was determined from two arms: 1) the risk of infantile pertussis due to having a mother
with pertussis (maternal pertussis), and 2) the risk of pertussis in infants without
maternal pertussis. Previous studies have calculated the annual attack rate from
household contacts to be 69-87% among approximately 50% unimmunized or under-
49

immunized children whose source of pertussis was identified (Deen et al., 1995; Wirsing
von Konig et al., 1998). In our study, we assumed the base-case annual attack rate in
infants with maternal pertussis to be 39% (i.e., mid-point estimate of the range of the
annual attack rate multiplied by 50%). For infants without maternal pertussis, the
disease probability in infants was the product of 7/12 (7-month susceptibility period) by
the annual incidence of pertussis in infants multiplied by the probability of no maternal
pertussis: (1-probability of maternal pertussis). We used the CDC reported annual
incidence of pertussis among infants under 6 months of age in the base case (CDC,
2012). The high case annual incidence of pertussis was assumed to be 385.4 cases per
100,000, based on data from the California Department of Public Health (CDPH)
collected during the 2010 statewide pertussis epidemic (Hay, 2004). Complications of
infantile pertussis include respiratory symptoms/complications (treated outpatient or
inpatient), neurologic complications, and/or death (Lee et al., 2007; Tanaka et al.,
2003). Among infants who develop pertussis-related encephalopathy, a proportion may
result in permanent brain damage (PBD), which can lead to costs associated with long-
term care management (Purdy et al., 2004). The frequencies of these complications and
their ranges for sensitivity analysis were listed in Table 3.1, based on published
literature and CDC reported data.

Costs
Total costs comprised of direct and indirect costs for both mothers and infants.
Generally, ‘direct’ implies changes in resource use associated with the intervention or
treatment procedure; ‘indirect’ refers to productively gains or losses related to illness or
50

death (e.g., opportunity costs associated with caregivers who missed work to care for
their sick children, the productivity losses due to premature mortality, and the indirect
costs from permanent disability) (Gold MR, 1996; Zhou et al., 2005). All costs in the
model were adjusted for inflation to 2010 US dollars using the medical care services
component of the consumer price index (US Department of Labor Statistics, 2011).

Direct costs
Potential direct costs include vaccine acquisition and administration costs,
vaccine-associated adverse events costs, and expenses to treat complications of
pertussis for both mothers and infants. Vaccine costs for birth mothers were based on
the CDC vaccine price list for ADACEL® (Packaging: 10 pack- 1 dose vial) (CDC
Vaccine Price, 2012). The costs of adverse events following Tdap vaccine were derived
from a recently published study (Lee et al., 2007). Data from previous economic studies
were used to estimate the medical costs for respiratory complications and neurologic
complications based on prevailing fee schedules (Lee et al., 2005; Lee et al., 2007;
Purdy et al., 2004). For each case of respiratory complication, direct medical costs
included costs of treating mild disease on an outpatient basis and costs of
hospitalization associated with severe disease. We estimated direct medical expenses
of each neurologic case (e.g., uncontrolled seizures, epilepsy or encephalopathy) from
previous cost studies of pertussis (Lee et al., 2005; Lee et al., 2004). In this analysis, we
assumed all infants who died received some treatment before death. And the direct
medical costs for a death, including pharmaceutical claims, outpatient, and inpatient
costs, were derived from an economic study in which the authors estimated such costs
51

using data from Massachusetts Department of Public Health and Centers for Medicare
& Medicaid Services in 2002 (Lee et al., 2005).

Indirect costs
Indirect costs include the unpaid costs from the mothers’ or the caregivers’ time
loss (working time & personal time) due to their own illness or their infants’ disease, and
the indirect societal costs of death and PBD (permanent brain damage). For pertussis in
infants, caregiver support costs were calculated by using estimates of 2 hours per office
visit and 8 hours per day for hospitalizations (Lee et al., 2004; Lee et al., 2007), which
covers both clinical office waiting time and medical procedure time (e.g., time for
diagnostic, therapeutic, and rehabilitation procedures). Indirect costs for birth mothers
were estimated to include costs of time lost from work and personal activity due to
adults' own pertussis (maternal pertussis) (Lee et al., 2004). The average compensation
per hour was $30.61 in 2012, as indicated by the US Bureau of Labor Statistics (BLS,
2012). The indirect societal cost per infantile death was based on economic estimates
of lifetime production (Purdy et al., 2004). The societal cost per case of PBD was
obtained from a published cost-benefit analysis (Purdy et al., 2004). There were no
additional indirect costs incurred for birth mothers to receive Tdap, as the vaccine was
given during the immediate postpartum period while in the hospital, and not during a
scheduled outpatient visit.

52

Chapter-Three Results
Base case results
From the societal perspective, the expected cost in the Tdap administration and
control arms were estimated at $129.27 and $187.97 per mother, respectively. In the
base case, our analysis suggests an ENB of $58.70 per vaccinated mother (shown in
Table 3.3). In our hypothetical cohort of 3.6 million birth mothers in the U.S., the overall
expected net benefit was calculated by the following equation: Overall ENB in the cohort
= ($58.70)*(3.6 million)*(vaccination coverage rate). Based on recent prospective data
and a published study of Tdap coverage in postpartum population, the coverage rate of
postpartum Tdap vaccination in the hospital was estimated between 25 ~ 60% (Yeh et
al., 2011; Healy et al., 2009). Therefore, the overall ENB in the cohort was estimated to
be $52.8 ~ $126.8 million.

From the health care system perspective (direct medical costs only), the expected
costs in the Tdap administration and control arms were estimated to be $57.30 and
$21.93 per birth mother, respectively. In this case, the ENB value was -$35.37 per
vaccinated mother suggesting that postpartum Tdap vaccination would not generate net
savings compared to no Tdap vaccination.

Sensitivity analysis
One-way sensitivity analyses were performed on key variables (Table S1 and
Figure 3.2). For some of the variables, ranges were selected based on the high and low
values from studies previously identified, +25% and -25% symmetric ranges were
53

estimated for others variables with limited published range information. Figure 3.2 is a
Tornado diagram, which shows the impact of varying model estimates on expected net
benefit per vaccinee. The variables that elicit the greatest impact on the model are
vaccine efficacy and annual pertussis incidence in birth mothers. For example, a 25%
decrease in vaccine efficacy leads to a 46% decrease in ENB, and a 22% increase in
annual pertussis incidence in birth mothers results in a 40% increase in the ENB
measure. The overall cost-benefit of postpartum vaccine strategy is highly sensitive to
the above two parameters. Other parameters, such as indirect societal costs per
infantile death, nonmedical costs attributable to maternal pertussis, percentage of
pertussis-associated severe/mild cough in mothers with pertussis, and vaccine price
may also influence the outcome measure. Percent change in outcome (ENB per mother)
divided by percent change in each key parameter, representing elasticity, are reported
in Table S1 in the appendix.

On the basis of the findings of one-way sensitivity analyses, we also performed
two-way sensitivity analyses by simultaneously varying disease incidence and vaccine
efficacy. Annual incidence of 100 to 500 cases per 100,000 birth mothers and 60%
versus 80% efficacies for Tdap were used with other parameters remaining unchanged
in the base-case values. The results of these two-way sensitivity analyses for the two
variables are presented in Table 3.4. From a societal perspective, all direct and indirect
costs were included. If we assumed 60% vaccine efficacy, then the resulting ENB per
vaccinated mother ranged from -$30.6 to $40.9 as the incidence ranged from 100 cases
per 100,000 to 500 cases per 100,000. If we assumed 80% vaccine efficacy, then the
54

resulting ENB per vaccinated mother ranged from -$24.4 to $70.6 as the incidence
ranged from 100 cases per 100,000 to 500 cases per 100,000. From a health care
system perspective, the trends were similar, but all of the ENB per mother were
negative values, indicating cost-incurring (data not shown).

If we change our assumption of pre-delivery Tdap coverage rate from 10% to 20%,
the ENB for each individual mother will remain the same with our base case
assumptions from a societal perspective. But this change would alter the hypothetical
cohort from 3.6 million birth mothers to 3.2 million postpartum women in the U.S., which
would correspondingly change the total dollar amount saved on a population level. The
overall societal benefits in the new cohort ranged from $47.0 ~ $112.7 million,
depending on the postpartum vaccination coverage level.

Chapter-Three Discussion
This study suggests that in-hospital postpartum vaccination of birth mothers with
Tdap prior to discharge is likely to be cost-beneficial in the base case from a societal
perspective. Compared with no Tdap vaccination, our economic model suggests an
ENB of $58.70 per vaccinated mother, predicting overall societal benefits in the cohort
of 3.6 million U.S. birth mothers at $52.8 ~ $126.8 million, depending on the postpartum
vaccine coverage rate. From a health care system perspective (HCS), the base-case
data suggest a negative value of ENB per vaccinated mother (-$35.37). Therefore, the
vaccination strategy is cost-beneficial from the societal perspective, but does not
generate net benefits from the HCS perspective.
55

In October 2012, ACIP recommended use of Tdap in every pregnancy,
irrespective of the patient's prior history of receiving Tdap (CDC, 2013). Although the
Tdap vaccination rate in pregnancy may alter the total dollar amount saved on a
population level, receiving Tdap for each individual mother in the hospital prior to
discharge is cost-beneficial (ENB=$58.70) with our base case assumptions from a
societal perspective. In addition, a proportion of infants may not receive a full series of
pertussis vaccine until later than 6 months of age (e.g., delay in vaccine timeliness due
to provider delaying vaccination for illness or lack of timely access to a vaccination
appointment). CDC data from 2011 indicated that only about 70% of 7-month old infants
have received at least 3 doses of a pertusiss vaccine (CDC Vaccination Coverage,
2011). For those infants not fully vaccinated at 7 months of age, maternal vaccination
protection period would be more than 7 months postpartum. As a result, the ENB of
postpartum vaccination in this population would be predicted to be higher than that
estimated from the base-case assumptions.

To our knowledge, only two economic studies have evaluated the cocoon strategy,
wherein the mother and at least 1 other household contact (usually the other parent) are
vaccinated with Tdap during the peripartum time period (Coudeville et al., 2009; Lee et
al., 2005). Lee et al. assumed reduction of infant disease by 40% and reported a cost-
effectiveness ratio of $268,000 per life-year saved from the societal perspective (Lee et
al., 2005). Coudeville et al. showed that vaccinating parents of newborns plus a single
booster dose for adults at 40 years of age was cost-effective, using a willingness-to-pay
56

threshold of $100,000 per life years gained (Coudeville et al., 2009). Ours is the only
study to evaluate the cost-benefit of routine postpartum Tdap vaccination of the mother
alone. Developing hospital-based strategies for vaccination of postpartum women prior
to delivery requires a concerted effort among physicians and nurses and support from
hospital administrators. The postpartum period provides a back-up opportunity for
targeted Tdap vaccination and can be easily incorporated into routine care. Healy et al.
2011 concluded that the postpartum Tdap program was well accepted and can be
successfully implemented in 8,334 postpartum women, although practical barriers such
as inaccurate recall of vaccination history and extended vaccination hours may exist for
vaccinating household contacts of infants (Healy et al., 2011).

Our assumptions on attack rate in infants exposed to maternal pertussis were
based on limited published data from studies of under-immunized children exposed to
household contacts. However, the ENB could vary with changing base-case vaccine
efficacy, annual attack rate in infants with maternal pertussis, as well as costs
attributable to pertussis. For example, few data exist for estimating pertussis attack rate
in newborns that are exposed to an infected mother or household contacts. Our model
utilized relatively old data about annual attack rate among under-immunized children
from household contacts. We may have underestimated the potential cost savings as
young infants are more susceptible to pertussis or pertussis-associated complications
than under-immunized children. To address this issue, we assigned wide ranges for the
key parameters in sensitivity analyses. No single variable in the one-way sensitivity
57

analyses decreased the ENB below $30 per vaccinee from a societal perspective
(Figure 3.2).

Conclusions
In conclusion, we found that hospital-based postpartum Tdap vaccination of birth
mothers who have not yet received this vaccine prior to delivery provides economic
benefits to society. Further development of robust data for certain input parameters
would increase precision of our findings. This study could be helpful to decision-makers
and may encourage the implementation of Tdap vaccine in the postpartum period to
complement current pre- and intra-partum vaccination efforts.

58

Tables and Figures
Table 3.1: Probability Variables Used in the Model
Probability Base case Range
Vaccine(Tdap)-associated adverse events, %
Local reaction 2 0-20
Systemic reaction 1 0-15
Anaphylaxis 0.0001 0-1
Vaccine (Tdap) Efficacy, %
Year 1
Year 2
Year 3
Year 4
Year 5
Year 6
Year 7
Year 8
Year 9
Year 10

80
78
77
76
65
55
44
34
23
19

60-90
58-89
55-87
44-76
34-65
23-55
19-44
14-34
10-23
4-19
Annual pertussis incidence (cases/100,000)
In birth mothers 450 370-550
In infants aged less than 6 months 71.6 69.9-385.4
Annual attack rate in infant with maternal pertussis, % 39.0 34.5-43.5
Frequencies of complications with maternal pertussis, %
Mild cough 27 16-38
Severe cough 67 48-78
Pneumonia 3 1-5
Hospitalization 3 1-5
Frequencies of complications with infantile pertussis, %
Respiratory complications (outpatient) 34.3 11.8-38.8
Respiratory complications (hospitalized) 63 34.2- 69
Neurologic complications 2 1.6-3
Death 0.7 0.6-0.8
Permanent brain damage (PBD) due to encephalopathy 0.001 -

59

Table 3.2: Base-case Parameters for Direct and Indirect Costs in 2012 US dollars
Variable Base case input, $
Direct costs
Vaccine costs
Adacel® price/dose 30.41
Administration/dose 15
Costs of vaccine-associated adverse events
Local reaction 54
Systemic reaction 174
Anaphylaxis 2,682
Medical costs for birth mother
Mild cough 325
Severe cough 452
Pneumonia 499
Hospitalization 1,601
Medical costs for infants
Respiratory complications (outpatient only) 127
Respiratory complications (hospitalized) 11,205
Neurologic complications 8,101
Death 18,212
Indirect costs
Average compensation per hour 30.61
Costs of time lost due to maternal pertussis 2,525
Caregiver support costs due to infant disease
Respiratory complications (outpatient only) 54
Respiratory complications (hospitalized) 561
Neurologic complications 858
Indirect societal cost per infantile death 5.9 million
Societal costs per permanent brain damage 3.3 million

60

Table 3.3: Base-case Results from a Societal or Healthcare System Perspective
Control
(unvaccinated)
Tdap
intervention
Incremental
From a Societal Perspective
Expected cost associated with maternal
pertussis ($)
117.74 53.52 64.22
Expected cost associated with infantile
pertussis ($)
70.23 30.35 39.88
Expected cost associated with vaccine ($) 0 45.40 -45.40
Expected cost per mother ($) 187.97 129.27 58.70 (ENB)
From a Healthcare System Perspective
Expected cost associated with maternal
pertussis ($)
14.34 8.62 5.72
Expected cost associated with infantile
pertussis ($)
7.59 3.28 4.31
Expected cost associated with vaccine ($) 0 45.40 -45.40
Expected cost per mother ($) 21.93 57.30 -35.37 (ENB)
ENB: expected net benefit

61

Table 3.4: Two-way Sensitivity Analyses of Annual Incidence and Vaccine Efficacy from a
Societal Perspective
ENB per vaccinated mother in
2012 US$
Base case $58.7
60% vaccine efficacy in the 1
st
year, at an annual incidence
(cases/100,000) in birth mothers of

100 -$30.6
200 -$12.9
300 $4.67
400 $23.3
500 $40.9
80% vaccine efficacy in the 1
st
year , at an annual incidence
(cases/100,000) in birth mothers of

100 -$24.4
200 -0.41
300 $23.1
400 $46.8
500 $70.6
ENB: expected net benefit

62

Figure 3.1: Model structure of a postpartum Tdap strategy

Fig.3.1. (A) Overall structure of a postpartum Tdap strategy; (B) maternal pertussis sub-tree.

Maternal
pertussis
Mild cough
Severe cough
Pneumonia
Hospitalization
Maternal pertussis subtree:
B
Postpartum
vaccination

Not vaccinate
Birth mother
Vaccinate
birth mother
With Tdap
Maternal
pertussis
Pertussis in infant
No maternal
pertussis
Adverse
events with
vaccine
Pertussis in infant
No adverse
events
A
No pertussis in infant
Respiratory complications
(outpatient only)
Respiratory complications
(hospitalized)
Neurologic complications
Death
Respiratory complications
(outpatient only)
Respiratory complications
(hospitalized)
Neurologic complications
Death
Same as node *
Same as node *
*
No pertussis in infant
63

Figure 3.2: Tornado diagram of one-way sensitivity analyses

Fig.3.2. Tornado diagram of one-way sensitivity analyses: shows the impact of range of
individual parameters on the expected net benefit (ENB) per mother. The vertical line in the
diagram represents the base-case ENB of $58.70 per postpartum vaccinated mother. Each
horizontal bar indicates the range of ENB per mother with the lower and upper values of each
parameter described on the right side of the figure.

64

Chapter-Three References
Barreto, L., Guasparini, R., Meekison, W., Noya, F., Young, L., and Mills, E. (2007).
Humoral immunity 5 years after booster immunization with an adolescent and adult
formulation combined tetanus, diphtheria, and 5-component acellular pertussis vaccine.
Vaccine 25(48), 8172-8179.
Bisgard, K.M., Pascual, F.B., Ehresmann, K.R., Miller, C.A., Cianfrini, C., Jennings, C.E.,
Rebmann, C.A., Gabel, J., Schauer, S.L., and Lett, S.M. (2004). Infant pertussis: who
was the source? Pediatr Infect Dis J 23, 985-989.
California Department of Public Health. Number of Cases of Pertussis in California:
Pertussis Summary Reports. Available:
http://www.cdph.ca.gov/programs/immunize/Documents/PertussisReport2010-12-15.pdf
(accessed July 12, 2011).
Centers for Disease Control and Prevention. Pertussis: Outbreaks. Available:
http://www.cdc.gov/pertussis/outbreaks.html (accessed March 24, 2012).
Centers for Disease Control and Prevention. Preventing tetanus, diphtheria, and
pertussis among adults: Use of Tetanus toxoid, reduced diphtheria toxoid and acelluar
pertussis vaccine.Recommendations of the Advisory Committee on Immunization
Practices. MMWR 2006, Vol.55 (RR17): 1-37.
Centers for Disease Control and Prevention. Prevention of Pertussis, Tetanus, and
Diphtheria among pregnant and postpartum women and their infants, recommendations
of the Advisory Committee on Immunization Practices (ACIP). MMWR 2008; 57(No.
RR-4).
Centers for Disease Control and Prevention. Updated Recommendations for Use of
Tdap in Pregnant Women and Persons Who Have or Anticipate Having Close Contact
with an Infant Aged <12 Months-Advisory Committee on Immunization Practices (ACIP).
MMWR 2011; 60(41):1424-1426.
65

Centers for Disease Control and Prevention. Morbidity and Mortality Weekly Report
(MMWR) 2013; 62(07):131-135. Available:
http://www.cdc.gov/mmwr/preview/mmwrhtml/mm6207a4.htm?s_cid=mm6207a4_w
(accessed March 3, 2013).
Centers for Disease Control and Prevention (CDC) Vaccine Price List. Prices last
updated: October 15, 2012. Available:
http://www.cdc.gov/vaccines/programs/vfc/awardees/vaccine-management/price-
list/index.html (accessed October 22, 2012).
Centers for Disease Control and Prevention. Estimated Vaccination Coverage with
Individual Vaccines Before 7 Months of Age by State and Local Area US, National
Immunization Survey, Q1/2011-Q4/2011. Available: http://www.cdc.gov/vaccines/stats-
surv/nis/tables/11/tab06_7mo_iap_2011.pdf (accessed March 4, 2013).
Consumer Price Index, US Department of Labor Statistics, item: medical care. Available:
http://data.bls.gov/labjava/outside.jsp?survey=cu (accessed May 14, 2011).
Coudeville, L., Van Rie, A., Getsios, D., Caro, J.J., Crepey, P., and Nguyen, V.H. (2009).
Adult vaccination strategies for the control of pertussis in the United States: an
economic evaluation including the dynamic population effects. PLoS One 4, e6284.
Crowcroft, N.S., and Pebody, R.G. (2006). Recent developments in pertussis. Lancet
367, 1926-1936.
Deen, J.L., Mink, C.A., Cherry, J.D., Christenson, P.D., Pineda, E.F., Lewis, K.,
Blumberg, D.A., and Ross, L.A. (1995). Household contact study of Bordetella pertussis
infections. Clin Infect Dis 21, 1211-1219.
Gold MR, S.J., Russell, L.B., and Weinstein, M.C. (1996). Cost-Effectiveness in Health
and Medicine (New York, NY: Oxford University Press).
Halperin, S.A., Smith, B., Russell, M., Scheifele, D., Mills, E., Hasselback, P., Pim, C.,
Meekison, W., Parker, R., Lavigne, P., et al. (2000). Adult formulation of a five
component acellular pertussis vaccine combined with diphtheria and tetanus toxoids
66

and inactivated poliovirus vaccine is safe and immunogenic in adolescents and adults.
Pediatr Infect Dis J 19, 276-283.
Hay, J.W. (2004). Evaluation and review of pharmacoeconomic models. Expert Opin
Pharmacother 5, 1867-1880.
Healy, C.M., Rench, M.A., and Baker, C.J. (2011). Implementation of cocooning against
pertussis in a high-risk population. Clin Infect Dis 52, 157-162.
Healy, C.M., Rench, M.A., Castagnini, L.A., and Baker, C.J. (2009). Pertussis
immunization in a high-risk postpartum population. Vaccine 27, 5599-5602.
Lavine, J., Broutin, H., Harvill, E.T., and Bjornstad, O.N. (2010). Imperfect vaccine-
induced immunity and whooping cough transmission to infants. Vaccine 29, 11-16.
Lee, G.M., Lebaron, C., Murphy, T.V., Lett, S., Schauer, S., and Lieu, T.A. (2005).
Pertussis in adolescents and adults: should we vaccinate? Pediatrics 115, 1675-1684.
Lee, G.M., Lett, S., Schauer, S., LeBaron, C., Murphy, T.V., Rusinak, D., and Lieu, T.A.
(2004). Societal costs and morbidity of pertussis in adolescents and adults. Clin Infect
Dis 39, 1572-1580.
Lee, G.M., Murphy, T.V., Lett, S., Cortese, M.M., Kretsinger, K., Schauer, S., and Lieu,
T.A. (2007). Cost effectiveness of pertussis vaccination in adults. Am J Prev Med 32,
186-193.
National Center for Health Statistics. Recent Trends in Births and Fertility Rates
Through June 2010. Available:
http://www.cdc.gov/nchs/data/hestat/births2010/births2010.pdf (accessed June 10,
2011).
Purdy, K.W., Hay, J.W., Botteman, M.F., and Ward, J.I. (2004). Evaluation of strategies
for use of acellular pertussis vaccine in adolescents and adults: a cost-benefit analysis.
Clin Infect Dis 39, 20-28.
67

Tanaka, M., Vitek, C.R., Pascual, F.B., Bisgard, K.M., Tate, J.E., and Murphy, T.V.
(2003). Trends in pertussis among infants in the United States, 1980-1999. JAMA 290,
2968-2975.
Van der Wielen, M., Van Damme, P., Joossens, E., Francois, G., Meurice, F., and
Ramalho, A. (2000). A randomised controlled trial with a diphtheria-tetanus-acellular
pertussis (dTpa) vaccine in adults. Vaccine 18, 2075-2082.
Ward, J.I., Cherry, J.D., Chang, S.J., Partridge, S., Lee, H., Treanor, J., Greenberg,
D.P., Keitel, W., Barenkamp, S., Bernstein, D.I., et al. (2005). Efficacy of an acellular
pertussis vaccine among adolescents and adults. N Engl J Med 353, 1555-1563.
Wendelboe, A.M., Njamkepo, E., Bourillon, A., Floret, D.D., Gaudelus, J., Gerber, M.,
Grimprel, E., Greenberg, D., Halperin, S., Liese, J., et al. (2007). Transmission of
Bordetella pertussis to young infants. Pediatr Infect Dis J 26, 293-299.
Wirsing von Konig, C.H., Postels-Multani, S., Bogaerts, H., Bock, H.L., Laukamp, S.,
Kiederle, S., and Schmitt, H.J. (1998). Factors influencing the spread of pertussis in
households. Eur J Pediatr 157, 391-394.
Yeh (2011). Yeh S, Mink CM, Kim M, Naylor SC, Zangwill KM. Effectiveness of hospital-
based procedures on postpartum vaccination with tetanus toxoid, reduced diphtheria
toxoid and acellular pertussis (Tdap) and seasonal influenza (sTIV). Presentation #632,
Infectious Diseases Society of America (IDSA) 49th Annual Meeting. October 20 - 23,
2011; Boston, Massachusetts.
Zhou, F., Santoli, J., Messonnier, M.L., Yusuf, H.R., Shefer, A., Chu, S.Y., Rodewald, L.,
and Harpaz, R. (2005). Economic evaluation of the 7-vaccine routine childhood
immunization schedule in the United States, 2001. Arch Pediatr Adolesc Med 159,
1136-1144.

68

CHAPTER 4: DIRECT MEDICAL COSTS ASSOCIATED WITH
PEYRONIE'S DISEASE IN THE U.S.

Chapter-Four Abstract

Background: Peyronie’s disease (PD) is a connective tissue disorder of the penile
tunica albuginea with a self-reported prevalence of 0.5% in the U.S. male population.
The standard treatment of PD is conservative therapy. Surgery is generally reserved for
men with severe penile deformities that impede sexual intercourse. Our primary
objective was to estimate the direct medical costs of PD in the U.S. Methods: Male
adults aged ≥ 65 in Medicare Advantage (MA) plans or aged 18-64 in commercial plans
with ≥1 PD diagnosis (ICD-9-CM: 607.85) were selected from a national administrative
claims database (Humana) between 07/2007 and 06/2010. The index date was defined
as the first observed date with a PD diagnosis. Continuous enrollment of ≥ 6 months
before (baseline) and ≥ 1 year after the index date (study period) was required. PD
patients were propensity score matched to controls based on age, race, geographic
region, plan type, diagnosis month and baseline comorbidities. Generalized Linear
Model (GLM) and Fixed Effect (FE) model with a difference-in-difference design were
used to estimate the incremental healthcare costs in PD patients over matched controls.
Sensitivity analyses were conducted to test the appropriateness of the base-case
control cohort by creating alternative comparison groups. Results: GLM model-adjusted
mean incremental healthcare costs per 6 months in PD patients over matched controls
were $2,610 (p<0.001), $189, and $943 (p<0.05) for the post-index 1-6, 7-12 and13-18
months in the MA cohort; $2,746 (p<0.05), $1,058 (p=0.063), $793, and $1,754 (p<0.05)
69

for the post-index 1-6, 7-12, 13-18 and 19-24 months in the commercial cohort,
respectively. FE model shows that MA and commercial PD patients had on average
significantly (p<0.05) higher risk-adjusted total costs of $802 and $1,364 per 6 months
compared to their matched controls, respectively. Outpatient cost was the main driver of
the healthcare cost differences. Sensitivity analyses, after refining the index date in
controls, reduced or obviated our base-case estimates for the incremental cost effect of
PD diagnosis. Implications: Our study may underestimate the direct medical costs
associated with PD, as claims data do not capture the over-the-counter drugs and PD
patients may not seek the treatment because of no effective drug specifically labeled for
PD before 2014. Future studies will focus on collecting the indirect costs associated with
diminished quality of life due to PD.

70

Chapter-Four Background
Peyronie’s disease (PD), also known as “Chronic Inflammation of Tunica
Albuginea”, is a connective tissue disorder of the penile tunica albuginea with unclear
cause (Taylor and Levine, 2007). It begins with the formation of collagen plaque or scar
tissue on the shaft of the penis that leads to penile curvature and painful erection. The
painful erection diminishes typically within 6 to 18 months as the penile deformity
stabilizes, and in some men, the penile curvature improves or resolves spontaneously
(Taylor and Levine, 2007; Dibenedetti et al., 2011; Jalkut et al., 2003). The clinical
course of PD includes a progressive and a quiescent phase. PD can cause erectile
dysfunction (ED) in the quiescent phase with stability of the penile deformity (Langston
and Carson, 2014). The prevalence of PD was estimated to be 1%~ 4% in European
male population (Kuehhas et al., 2011; Schwarzer et al., 2001). A recent large-scale
survey study has reported the prevalence of PD to be 0.5% (diagnosis of PD) based on
self-reported data among 11,420 U.S. male adults aged ≥ 18 years (Dibenedetti et al.,
2011). Old men aged 50-70 years are at the greatest risk of PD (Taylor and Levine,
2007; Kadioglu et al., 2011; Langston and Carson, 2014).

Patients with vascular comorbidities may have an increased risk of PD. Such
vascular comorbidities include diabetes, hypertension, hyperlipidemia, and ischemic
heart disease (Deveci et al., 2006; Mulhall et al., 2006; Kadioglu et al., 2002; Usta et al.,
2004; Kendirci et al. 2007). PD is frequently associated with other comorbid conditions,
including ED (Wessells et al., 2007; Dibenedetti et al., 2011), prostate cancer (Tal et al.,
2010), other male genital organs disease (e.g., inflammatory diseases of prostate,
71

orchitis and hypertrophy of prostate) (Usta et al. 2004), other collagen-based diseases
(e.g., Dupuytren's contracture and Ledderhose disease) (Nugteren et al., 2011; Levine
and Greenfield, 2003) and/or smoking (Deveci et al., 2006; Usta et al., 2004). ED is not
only a potential risk factor associated with PD, but also an adverse event due to PD. A
recent study showed that 58.1% of 1,001 PD patients reported having ED (Kadioglu et
al., 2011). Another study reported the prevalence rate of PD among diabetic patients
with ED to be 20.3% (Arafa et al., 2007). Published literature also suggested that PD
can cause embarrassment, low self-esteem, emotional distress, depression, and
diminished quality of life (QoL) (Bella et al., 2007; Rosen et al., 2008).

The standard treatment of PD is conservative therapy. Previous studies showed
that the disorder improves or resolves spontaneously in 4% to 13% of PD patients
(Jalkut et al., 2003; Mulhall et al., 2006; Kadioglu et al., 2002). Surgery is generally
reserved for men with severe penile deformities that impede sexual intercourse and
should be considered only when the disease is stable for at least 6 months (Taylor and
Levine, 2007; Kendirci and Hellstrom, 2004). The mean time from first diagnosis to first
surgery was estimated to be 6.3± 6.8 months (Dibenedetti et al., 2011). Surgical options
for PD include tunica plication procedures for less severe disease, penile plaque
excision with grafting procedures for more severe patients with a significant penile
curvature greater than 60
o
, and insertion or replacement of penile prosthesis which
most occurs with diagnosis of PD and/or ED (Wessells et al., 2007; Taylor and Levine,
2008). Nonsurgical treatment options, including oral medication (e.g., Vitamin E), topical
therapy (e.g., Verapamil), intralesional injection (e.g., corticosteroid), and traction
72

therapy (e.g., penile traction device) lack adequate clinical data support (Taylor and
Levine, 2007; Kuehhas et al., 2011). In December 2013, the first drug (Xiaflex made by
Auxilium Pharmaceuticals) to treat PD was approved by the U.S. Food and Drug
Administration (FDA).

Clinical burden studies of PD were primarily from patient self-reported survey data
that may capture patients’ psychosocial problems, sexual functioning, or relationship
problems (Levine and Greenfield, 2003; Rosen et al. 2008; Smith et al., 2008; Smith et
al., 2009). Compared to self-reported data, administrative claims data are free from
recall bias and relatively more objective. This study sought to estimate the direct
medical costs associated with PD in the real-world setting, specifically, to compare
healthcare costs and resource utilization between patients with PD and matched
controls without a diagnosis of PD.

Chapter-Four Materials and Methods
Data Source
Humana administrative claims data over a 6-year period (from 1/1/2007 to
6/30/2013) were used in this study. Humana Inc. is one of the largest health plans and
the second largest provider of Medicare Advantage (MA) plans in the U.S. The Humana
database covers 8-14 million enrollees with approximately 60% of them from MA and 40%
from commercial insurance plans. Patient-level information include demographics (e.g.,
age, gender, race, state), monthly enrollment history, dates of service, allowed payment
amounts, primary and up to 8 secondary diagnoses (using the International
73

Classification of Diseases, ninth revision, ICD-9 code) and procedures (using the
Current Procedural Terminology, CPT code). The pharmacy claims include drug name,
days of supply, quantity, date of prescription fill, and the National Drug Code. The
database also contains laboratory results and date of collection for the laboratory tests.
This study was exempt from the institutional review board review as the patient-level
data are de-identified.

Study Population in the Base Case
To select the population of interest for this study, we applied the following inclusion
and exclusion criteria: Patients with at least one PD diagnosis during the study intake
period from 07/01/2007 to 06/30/2010 were identified using a primary or secondary ICD-
9 code of 607.85 in the medical claims. Index date in cases was defined as the first
observed date of PD diagnosis. PD patients were considered “newly diagnosed” if they
had at least 6 months of data prior to the “index date” without a claim for PD diagnosis
or surgical procedure of PD. For example, an individual with 2 ICD-9 codes for PD in
2008 would be considered newly diagnosed if they had no other ICD-9 or CPT codes for
the condition in prior year (i.e. 2007) if continuously enrolled (Joyce et al. 2008).
Continuous enrollment for at least 6 months before (baseline period) and 1 to 3 years
after the index date (study period) was required to ensure sufficient time to observe the
incurred medical costs and adverse events associated with PD. In this study, insurance
plan type could be MA or commercial health maintenance organization (HMO), MA or
commercial preferred provider organization (PPO), MA or commercial point of service
(POS), MA private fee-for-service (PFFS), and individual major medical plan tailored for
74

individuals, families or small businesses (IMV). Patients with Medicare supplemental
plans were excluded, because we cannot observe their complete claims. We then
divided the sample into 3 cohorts: (1) PD patients aged above 65 enrolled in MA plans,
(2) patients aged 18-64 enrolled in commercial insurance plans, and (3) patients aged
below 65 from MA or aged above 65 from commercial plans. We evaluated these
cohorts separately because adults aged below 65 in Medicare plans are people with
disability or end-stage renal diseases, and commercial patients may be dual eligible if
they are above 65. In addition, the race information is unknown for commercial
insurance plans. This is another reason why we separate the analyses for MA and
commercial cohorts.

Controls were identified among enrollees who had no diagnoses of localized
fibrotic conditions, such as PD, Dupuytren’s contracture (ICD-9: 728.6, 718.44),
Ledderhose disease (ICD-9: 728.71), and no PD-related surgical procedures in their
claims histories between 2007 and 2013 (Macaulay et al., 2012). The initial pool of
potential controls without PD was about 1.61 million. Controls were also required to
have a continuous enrollment for at least 6 months before and at least 1 year after a
randomly assigned index date. To take into account the different healthcare-seeking
behavior by season and other unknown time-dependent confounders, the index date in
controls were assigned within a +/- 30-day time window of the index date among the
cases.

75

We conducted a power analysis with alpha at 0.05 (i.e., type I error of 0.05 refers
to a 5% chance that the results are due to chance rather than to the intervention) and
power at 0.8 (i.e., there’s 80% probability that one will reject the null hypothesis when
the null hypothesis is correct) to estimate the sample size required to detect a 5%, 10%,
15% and 20% difference in costs between cases and controls (i.e., effect size of 0.05 to
0.20). Previous literature suggested that a variable ratio (i.e., one treated patient
matched to up to 5 untreated subjects), balanced 1-to-5 nearest neighbor approach
increases the precision over 1-to-1 matching at a small cost in estimation bias (Rassen
et al., 2012; Austin PC, 2010). To ensure adequate power, PD patients were matched at
a 1-to-5 variable rate (i.e., one PD case is matched to up to 5 controls) using a
propensity scoring matching within 0.1 caliper (Rosenbaum and Rubin, 1985; Rassen et
al., 2012). Specifically, the propensity score or the likelihood of being diagnosed with
PD was obtained by a stepwise logistic regression for each subject in the study cohort.
The regressors include age, race, geographic region, plan type, diagnosis month
dummies, and baseline comorbidities. We derived the comorbidities or risk factors of PD
from prior literature (details described above in the Background section). Comorbidities
were coded as binary indicators (i.e., dummy variables) in the equation, as it would be
the best way to avoid estimation bias in retrospective studies (Aiken et al., 2014). The
ICD-9 diagnosis, procedure codes and CPT codes associated with PD are presented in
Appendix Section A. Figure 4.1 demonstrates the base-case study design and sample
selection process following the above inclusion and exclusion criteria.

76

Comparative Analyses and Statistical Tests
Descriptive analyses compared the baseline characteristics, the post-index costs
and the resource utilizations between PD patients and controls. Total healthcare costs
and costs by type of service (e.g., outpatient, inpatient, emergency department and
prescription) at baseline were compared between cases and controls to define if the
matching was good. The baseline costs by type of insurance plan (e.g., HMO, PPO,
POS, etc.) were also examined to evaluate if they were comparable among different
plan types. All costs were adjusted to 2014 US dollars using the medical component of
the consumer price index (U.S. Department of Labor Statistics, 2014). The health
resource utilizations during the 3-year study period were described using frequencies
and proportions (n, %). As the skewed distribution of cost data violates the normality
assumption, Wilcoxon rank-sum tests (non-parametric) were used to compare the costs
between cases and controls. Chi-square tests or Fisher's exact tests were used to
compare categorical variables, including race, geographic region, and plan type.
McNemar’s test was used to check the marginal frequencies of two binary outcomes,
such as the comorbidity indicators. All statistical tests and resulting p-values were two-
sided.

Model Structure
Health care expenditures that reflect the economic burden of illness generally
include the allowed payments made by a third-party payer (e.g., primary and secondary
coverage, net of negotiated discounts) and the enrollee’s liability (e.g., co-pay,
deductible, and coinsurance). Based on review of the literature, Generalized Linear
77

Models (GLMs) are best suited to normalize the skewness of healthcare cost data with
no need to re-transform (Deb et al., 2006; Hardin and Hilbe, 2007). For GLMs, the
distribution of a response could be one of the exponential families of distributions (e.g.,
Poisson, gamma, binomial, inverse Gaussian). A monotonic link function that relates the
mean of the response to a scale on which the model effects combine additively could be
in the form of identity, logarithmic, square root, logistic, or power (Hardin and Hilbe,
2007). It has been suggested that health care expenditure data frequently have a log
link and gamma distribution (Barber and Thompson, 2004). In this study, Box-Cox test
and Park test (i.e., GLM Family test) were performed to determine the particular
exponential distribution and the form of link function (Deb et al., 2006; Manning and
Mullahy, 2001).

The equation for the GLM model is as follows: 𝑦 𝑖𝑡
= 𝛽 0
+ 𝛽 1
𝑥 𝑖𝑡
+ 𝛽 2
(𝐷 𝑖 ∗ 𝛾 𝑡 ) +
𝛽 3
𝐷 𝑖 + 𝛾 𝑡 + 𝜇 𝑖 + 𝜀 𝑖𝑡
, 𝑖 = 1, … , 𝑛 ; 𝑡 = 0,1 … , 6, where 𝑦 𝑖𝑡
denotes healthcare costs per
patient (i) per 6 months (t); 𝑥 𝑖𝑡
represent time-variant variables including age and
comorbidities at (t-1); 𝐷 𝑖 defines a dummy variable indicating PD diagnosis; 𝛾 𝑡 refers to a
time variable indicating panels with a unit of 6 months; 𝜇 𝑖 represent individual specific
effects, including race, region, plan type, diagnosis month; and 𝜀 𝑖𝑡
denotes an error term.
We added an interaction term between the diagnosis indicator 𝐷 𝑖 (i.e., disease, no
disease) and the study periods 𝛾 𝑡 (i.e., baseline, 1-6, 7-12,...,25-30, 31-36 months after
the index date) to allow the impact of disease to vary over time.

78

Alternative Model
We also applied a fixed effects model with a difference-in-difference design (i.e.,
compare costs before and after the index date with a comparison group) to examine the
incremental cost effect of PD diagnosis. Studies showed that a difference-in-differences
approach can be incorporated into a regression framework to control for observed
differences across groups (Kessler and McClellan, 1996; Joyce et al., 2013). The model
structure is 𝑦 𝑖𝑡
= 𝛽 0
+ 𝛽 1
𝑃𝑜𝑠𝑡 𝑖 + 𝛽 2
𝐷 𝑖 + 𝜷 𝟑 (𝐷 𝑖 ∗ 𝑃𝑜𝑠𝑡 𝑖 ) + 𝛽 4
𝑥 𝑖𝑡
+ 𝛾 𝑡 + 𝜇 𝑖 + 𝜀 𝑖𝑡
, 𝑖 =
1, … , 𝑛 ; 𝑡 = 0,1, … , 6, where 𝑃𝑜𝑠𝑡 𝑖 is a dummy variable indicating pre- or post-index
period, 𝜇 𝑖 denote individual fixed effects, other variables such as 𝑥 𝑖𝑡
, 𝐷 𝑖 and 𝛾 𝑡 are the
same as those described above for GLM model. This alternative model includes
individual fixed effects and only two observations per individual (one for pre- and one for
post-index period) to estimate the coefficient on the interaction term(𝐷 𝑖 ∗ 𝑃𝑜𝑠𝑡 𝑡 ).

Tests for Attrition bias
The longitudinal inpatient, outpatient, physician and prescription claim files of all
patients were followed from 1 to 3 years after the index date. Attrition in this sample
resulted in smaller cohorts over time, with more than 20% attrition during the post-index
study period (Figure 4.1). In order to understand if attrition affected the estimates of our
model coefficients, we performed the following statistical tests. (1) First we ran separate
GLM regressions of total healthcare costs on time-invariant characteristics (e.g., the
diagnosis indicator, demographics and clinical characteristics at baseline period) using
a full and a completing sample. A full sample includes all available cases and controls
with at least 1-year post-index continuous enrollment, ignoring attrition; while a
79

completing sample represents those who remained in their insurance plans with a
continuous enrollment of at least 3 years after the index date. We then used seemingly
unrelated estimation (i.e., a Wald Chi-squared test) to evaluate the difference in
coefficients of GLM model between the full versus the completing samples (Fitzgerald
JM, 2011; Clogg et al., 1995). (2) We also performed a Probit test developed by
Fitzgerald Gottschalk and Moffitt to predict the probability of attrition on characteristics
variables and lagged dependent variables (Fitzgerald et al., 1998). (3) We applied the
Becketti, Gould, Lillard and Welch (BGLW) test by regressing the first lag of total costs
on an indicator variable for attrition, time-invariant characteristics, and interaction terms
of attrition dummy with other explanatory variables during the post-index 13-24 months
and 25-36 months. We then performed an F-test for joint significance of coefficients on
the attrition dummy and the interaction terms to determine if the coefficients differ
between patients who remained and patients who dropped out (Becketti et al. 1988)

Sensitivity Analyses with Alternative Control Groups
The validity of the difference-in-difference approach assumes that variation in
healthcare costs between cases and controls before and after their index date only
reflects the medical costs associated with PD. However, it’s possible that there are
unobserved differences in healthcare-seeking behavior across groups that is changing
differentially over time. In order to test the appropriateness of the base-case control
cohort, we re-estimated the models creating an alternative control group with a new
index date satisfying the following two conditions: (1) with a physician office visit for any
other diseases except PD, Dupuytren’s contracture, or Ledderhose disease; and (2)
80

within +/- 30 day window of the index date among PD cases. As such, there would be at
least a physician office visit on the index date of both cases and controls. Multivariable
regressions were used to examine the differences in post-index costs between PD
patients and the alternative controls. We hypothesized that unobserved differences in
healthcare-seeking behavior would obviate our base-case results for the incremental
cost effect of PD diagnosis.

Next, we conducted a series of multivariable regressions to compare the
healthcare costs between 2 disease groups: those diagnosed with PD and those
diagnosed with benign prostatic hypertrophy (ICD-9-CM: 600.01). Benign prostatic
hypertrophy (BPH) is a slowly progressive disorder that affects one third of men older
than 50 years. Approximately 14 million men in the U.S. have lower urinary tract
symptoms caused by BPH (Emberton et al., 2008; McVary et al., 2011). In clinical
practice, BPH patients who are not bothered by their symptoms and are not
experiencing complications of BPH should be managed with watchful waiting. Similar to
PD, surgery is only reserved for severe patients with very large prostates (>75g). Most
BPH patients that do not present with obvious surgical indications are treated with
temporizing medications (McVary et al., 2011). BPH was selected in this scenario in
order to test if the direct economic burden of PD is comparable to that of a similar but
more common medical condition. We hypothesized to find smaller amount of cost
differences between the 2 comparison groups.

81

Chapter-Four Results
Base-case results
Table 4.1 presents demographic characteristics, healthcare costs by service type,
and frequencies of comorbidities at the baseline period. As also shown in Figure 4.1,
879 and 540 PD patients enrolled in MA and commercial plans met the inclusion criteria,
respectively. Most PD patients were between 45-74 years of age (82.9%), and 743
(89.8%) patients were white among those 827 MA enrollees with a known race.
Moreover, most enrollees in Humana were from the southern region (66.0%) and no
significant regional difference was observed before and after matching. As for the
baseline comorbidities, commercial patients were younger and healthier than MA
patients with lower frequencies reported (Table 4.1). The propensity score matching
resulted in 5,999 controls with 3,708 from MA and 2,291 from commercial cohort. All the
baseline characteristics were well-matched except that a significant difference remained
between cases and controls for the presence of ED. There were 210 (23.9%) and 655
(17.7%) patients with ED in the case and matched control groups in MA cohort (p <0.05).
In commercial cohort, the proportions of ED patients among cases and controls were
16.1% and 10.6%, respectively (p <0.05).

Figure 4.1 shows the sample selection process and the sample sizes with 1- to 3-
year continuous enrollment after the index date. The sample size in PD group
decreased from 1,419 to 839 at almost the same attrition rate as that in the control
group (from 5,999 to 3,511) when following up from 1 to 3 years after the index date.
The results for the tests of attrition bias cannot reject the null hypothesis of a random
82

attrition that is primarily because of dropout from the Humana insurance in both PD and
control groups (Appendix Section D). Specifically, the results for the probit tests show
that the pseudo R
2
values, which imply the explanatory power, were similar for attrition
during the post-index 13-24 months and during 25-36 months. Lagged cost outcomes
were not significant predictors of attrition, as their coefficients were not statistically
significant except for the first lag of total cost during the 13-24 months in MA cohort
(Table A.4.6). Table A.4.7 shows the results of the BGLW test (i.e., a reverse of probit
test) for attrition. The coefficients on the attrition dummy variable were not statistically
significant in both MA and commercial cohorts (all p-values >0.05), suggesting that
attrition might be a random process. The F-test for joint significance of coefficients on
the attrition dummy and its interactions with demographic characteristics cannot reject
the null hypothesis that attrition was random in this sample (all p-values >0.05). In
addition, descriptive comparison of the characteristics of patients who remained in the
case and control groups at 3 years (i.e., a completing sample) showed that there was
no significant change within and between groups for the characteristics that were
matched at baseline.

Table 4.2.a reports the resource utilization in Medicare and commercial cohorts
during the 3-year post-index study period. Compared to controls, Medicare PD patients
had significantly higher rates of outpatient hospital visit (81.9% vs. 72.9%), ambulatory
surgical center visit (43.9% vs. 32.7%), off-label use of PD-related prescription (20.5%
vs. 6.2%), ED prescription use (31.3% vs. 18.4%), cardiovascular agents use (34.1% vs.
26.2%), corticosteroids use (47.2% vs. 39.7%), and surgery resource use for PD and/or
83

ED (16.0% vs. 0.4%) (all comparisons p<0.01). On the other hand, commercial PD
patients had significantly higher rates of outpatient visit (100.0% vs. 96.3%), outpatient
hospital visit (66.0% vs. 49.3%), any prescription drug use (84.4% vs. 69.4%), off-label
use of PD-related prescription (26.2% vs. 5.5%), ED prescription use (36.5% vs. 15.4%),
analgesics use (69.7% vs. 55.9%), anti-anxiety medication use (24.2% vs. 13.4%),
cardiovascular agents use (36.9% vs. 17.9%), corticosteroids use (49.6% vs. 32.2%),
and surgery resource use for PD and/or ED (7.0% vs. 0%) compared to matched
controls (all comparisons p<0.01). We also reported the selected drug use at the
baseline period (Table 4.2.b) for those categories of drugs noted to be used more
frequently in cases compared to controls in the study period. The results showed that
the baseline drug usage was comparable expect that Medicare PD patients had a
significantly (p<0.05) higher rate of ED prescription use than that in controls.

Table 4.3.a and 4.3.b demonstrate the comparison of unadjusted cost outcomes
between cases and controls. Distribution of the unadjusted costs, with mean,
interquartile range (difference between the 75th and 25th percentile estimates) and 90th
percentile values were reported in the baseline and the study periods. Healthcare costs
at the baseline were comparable (i.e., no significant difference) between cases and
controls, which suggests a balanced matching. Descriptive comparisons suggested that
PD patients in Medicare cohort had on average significantly higher total and outpatient
costs; while PD patients in commercial cohort had on average significantly higher total,
outpatient, and prescription drug costs (p<0.05) in the study period.

84

As suggested by the Box-Cox test (i.e., coefficient of theta equals to 0.011) and
the Park test (coefficient indicating the relationship between mean and variance is 1.88),
the GLM model with a gamma distribution and a log link was used to compare the risk-
adjusted healthcare costs. Multivariate regressions provided similar results to the
descriptive comparisons of the healthcare costs in Table 4.3. GLM model-adjusted
mean incremental all-cause healthcare costs per 6 months in PD patients over matched
controls were $2,610 (p<0.001), $189, and $943 (p<0.05) for the post-index 1-6, 7-12
and 13-18 months in MA cohort; $2,746 (p<0.05), $1,058 (p<0.1), $793, and $1,754
(p<0.05) for the post-index 1-6, 7-12, 13-18 and 19-24 months in commercial cohort,
respectively. Incremental costs were not statistically significant after two-year follow up.
Figure 4.2 demonstrates the trajectory of GLM model-adjusted incremental total costs
with a fluctuation trend that may imply the treatment pattern of PD.

Figure 4.3.a and 4.3.b demonstrate the GLM model-adjusted outpatient and
prescription drug costs in PD patients over matched controls. The fluctuation trend in
this figure is consistent with that in Figure 4.2 for the total costs, suggesting that
outpatient cost would be the main driver of the total healthcare cost differences between
cases and controls. Prescription medication costs also contributed to the total
healthcare cost differences in PD patients over matched controls, particularly for
patients in the commercial cohort, as shown in Figure 4.3.b.

The present value of the costs incurred beyond one year time horizon was
calculated by the formula PV=FC{1-1/(1+r)^t }/r where, PV = present value, FC = future
85

costs, r = discount rate (3%) and t = time horizon. After applying a discount rate of 3%
for costs incurred beyond the diagnosis year, the mean incremental cumulative costs
(GLM model-adjusted) per patient over matched control in the 3-year study period were
estimated at $6,720 for total costs, $4,927 for outpatient costs, and $1,379 for
prescription drug costs in commercial cohort; $4,319 for total costs, $4,229 for
outpatient costs, and $912 for prescription drug costs in MA cohort, respectively. FE
model showed that Medicare PD patients had on average significantly higher total costs
of $802 (p=0.011), outpatient costs of $738 (p<0.001), and prescription drug costs of
$43 (p=0.248); Commercial PD patients had on average significantly higher total costs
of $1,364 (p=0.025), outpatient costs of $759 (p=0.005), and prescription drug costs of
$51 (p=0.412) per patient per 6 months compared to their matched controls.

Results for Sensitivity Analyses
Sensitivity analyses using an alternative control group that has at least a physician
office visit on the index date show that unadjusted or model-adjusted healthcare costs
were comparable between cases and controls. That is, no significant healthcare cost
difference was observed in PD patients compared to the “physician office visit” controls
(all comparisons p>0.05 shown in Table A.4.2 in the Appendix Section B). Selected
drug uses in the study period were also comparable, except that cases had on average
significantly (p<0.01) higher utilization of PD-related prescription, ED prescription,
cardiovascular agents, and corticosteroids. In order to investigate the reason why there
was no significant incremental cost in PD over alternative controls, frequencies of
physician office visits at the baseline period were checked. Number of physician office
86

visits reported as mean value [median] (standard deviation) were 12 [8] (14) and 16 [11]
(20) for the disease and the alternative control group, respectively. This implies that the
“physician office visit” control group had selected those enrollees who were inclined to
use the healthcare services more frequently, although the baseline healthcare costs
were well matched by the propensity score matching. As the physician office visit on the
index date of the alternative controls could be that for any other diseases except PD,
Dupuytren’s contracture, or Ledderhose disease, this method of generating controls
may result in selecting a sample of the general population with relatively higher costs.

Appendix Table A.4.3 shows no significant cost difference (unadjusted) between
PD patients and the patients diagnosed with BPH. GLM model-adjusted incremental
outpatient costs in PD compared to BPH patients were demonstrated in Figure A.4.1 in
Appendix Section B. Table A.4.4 in Appendix Section C shows the distribution of
unadjusted healthcare costs in the 261 PD patients that were excluded in our base-case
analysis. The significantly higher healthcare costs in this subgroup confirmed our
hypothesis that this population may have higher rates of comorbidities due to disabilities
or end-stage renal diseases.

Chapter-Four Discussion
To our knowledge, only one published study used administrative datasets to
calculate physician office visit and ambulatory surgery visit rates for PD among
traditional Medicare beneficiaries or veterans (Wessells et al., 2007). Ours is the first
study to estimate the healthcare costs and resource utilization associated with PD in the
87

U.S. Our findings of descriptive characteristics are consistent with the literature of PD
(Taylor and Levine, 2007; Kuehhas et al., 2011; Langston and Carson, 2014). Our
results suggest that PD incurs a small amount of direct economic burden with a model-
adjusted mean incremental outpatient cost of $4,000 to $5,000 in the 3-year study
period, compared to their base-case controls. Outpatient costs would be the main driver
of the total cost differences between cases and controls, as outpatient services include
physician’s office visits, injection procedures and surgeries for PD. In addition,
sensitivity analyses using the physician office visit control group and the BPH control
group suggest that the direct economic burden of PD is comparable to that for other
diseases or to that of a similar but more common condition.

Incremental healthcare costs fluctuated during the 3-year study period, which may
imply the treatment pattern of PD (Figures 4.2 and 4.3.a). That is, after being diagnosed
in physicians’ office, some patients may wait-and-see or take oral medications first. If
the condition does not improve within 12 months, they would go back to use outpatient
services for injection therapy, and/or a few of patients may end up receiving surgery.
These figures also show a delay of the second peak value for the incremental costs
from 18 months to 2 years in commercial patients. A likely reason is that the coverage
of Medicare plans might be more generous, so that Medicare patients would seek the
more expense treatments (e.g., surgery) earlier than commercial patients. Another likely
interpretation could be that commercial patients are younger, they may consider this
disease more seriously and are willing to seek treatments more often (e.g., go to
physicians’ office, take oral or injection medications) in the first year after diagnosis. At
88

the same time, it’s also possible that they want to avoid surgeries because of the
associated risks (e.g., post-operative impotency, shortening of penis, diminished
sensation of penis). Our model shows no significant total cost difference between cases
and controls after 2 years follow up (i.e., in the third year after diagnosis) (Figure 4.2). A
likely reason could be that penile curvature improves or resolves in most PD patients by
then. Or patients may stop seeking treatments and get used to the condition with a
stable penile deformity.

We acknowledge that the “index date” in PD patient was defined as the first
observed date of PD diagnosis, which was not necessarily the true initial diagnosis date.
The off-label use of PD-related prescriptions might not necessarily be used to treat PD.
These medications, such as Carnitine, Potaba, Colchicine, Tamoxifen, Soltamox,
Aminobenzoate potassium and Pentoxifylline, could be prescribed to treat other
diseases, such as the peripheral artery disease. This explains why we found that both
Medicare and commercial cohorts had comparable PD-related medication use between
cases and their matched controls at the baseline period (4.9% vs. 4.6% and 2.9% vs.
2.4%) (Table 4.2.b). In addition, PD patients had significantly higher usage of
analgesics (i.e., pain medications), anti-anxiety medications and anti-depressants
compared to their matched controls correspondingly (p<0.05). Our findings are
consistent with the literature of PD on the effects of patients’ health-related quality of life
(HRQOL).

89

The study has several limitations. First, there is no way to directly estimate the
diagnosis-specific costs, particularly with claims. Generally, you can only estimate all-
inclusive or all-cause healthcare costs in the case group compared to that in the control
group. Second, the severity of disease could not be measured using claims data. Third,
although previous research indicated that Humana data is reasonably representative of
the general U.S. population (Shafazand et al., 2010), our findings from this study may
not be generalizable to the national population of patients with PD. Fourth, claims data
cannot capture over-the-counter (OTC) drugs, such as Vitamin E and L-Arginine, which
may underestimate the economic burden associated with PD. In addition, PD patients
may seek the treatment anonymously at community health clinics, resulting in fewer
claims being filed and the direct medical costs of PD being underestimated here.

A more important reason that may cause an underestimation of the economic
burden is that PD is under-treated (Dibenedetti et al., 2011). Surgery is reserved for the
worst cases with severe penile deformities and no effective drug was specifically
labeled for PD before 2014. Some patients will be scared off by the risks associated
with the surgery, such as diminished penile sensation, post-operative shortening of the
penis and impotence (about 30% of men who underwent surgery become impotent from
expert opinion) (kovac et al., 2007). Some patients would not be able to tolerate
anesthesia or the intraoperative or postoperative complications, including injury to the
tendon, digital nerve, skin necrosis, infection and complex regional pain syndrome. In
addition, surgery may require prolonged recovery and rehabilitation periods (Dibenedetti
et al., 2011; Taylor and Levine, 2008). These patients may prefer treatment with the
90

new drug XIAFLEX® (collagenase clostridium histolyticum) approved for injectable use
in non-ventral plaques. We believe that a substantial quantity of patients may try to
avoid surgery and switch to a nonsurgical therapy option due to the limitations of
surgery (Kovac et al., 2007). Currently, clinical studies are conducted to determine the
safety and effectiveness of injection for ventral plaques as well as the outcomes of early
treatment by the new injectable medication (Langston and Carson, 2014).

In conclusion, our study may underestimate the direct medical costs associated
with PD, as claims data do not capture over-the-counter drugs and patients might not
seek treatment because no effective drugs were specifically labeled for PD before 2014.
The burden of illness involves not only the direct economic burden, but also involves the
indirect economic burdens due to lost productivity or unquantifiable psychological
impact imposed on the patient for having to cope with the disease. As PD can cause
psychosocial stress, relationship difficulties as well as sexual dysfunction (suggested by
the literature and shown in our results in Table 4.2.a), future studies will focus on
collecting the indirect costs associated with the diminished quality of life due to PD.

91

Tables and Figures
Table 4.1: Baseline characteristics and comorbidities

Medicare Advantage cohort Commercial cohort
PD (n= 879) No PD
(n=3708)
PD (n=540) No PD
(n=2291)
Age, years, Mean (SD) 70.0 (4.2) 70.1 (4.3) 51.3 (9.7%) 51.4 (9.8%)
Age distribution, N (%)
18-44 N/A N/A 109 (20.2%) 469 (20.5%)
45-54 N/A N/A 170 (31.5%) 735 (32.1%)
55-64 N/A N/A 261 (48.3%) 1087 (47.5%)
65-69 488 (55.6%) 2027 (54.7%) N/A N/A
70-74 256 (29.2%) 1080 (29.1%) N/A N/A
75-87 134 (15.3%) 601 (16.2%) N/A N/A
Race, N (%)
White 743 (84.6%) 3164 (85.3%) N/A N/A
Black 56 (6.4%) 248 (6.7%) N/A N/A
Hispanic 17 (1.9%) 68 (1.8%) N/A N/A
Other race
a
11 (1.3%) 43 (1.2%) N/A N/A
Unknown 51 (5.8%) 185 (5.0%) N/A N/A
Geographic Region, N (%)
Midwest 203 (23.1%) 881 (23.8%) 128 (23.7%) 539 (23.5%)
Northeast 11 (1.3%) 45 (1.2%) 1 (0.2%) 3 (0.1%)
South 578 (65.8%) 2430 (65.5%) 373 (69.1%) 1576 (68.8%)
West 86 (9.8%) 352 (9.5%) 38 (7.0%) 173 (7.6%)
Insurance Plan Type, N (%)
HMO 343 (39.1%) 1458 (39.3%) 98 (18.2%) 438 (19.1%)
PPO 315 (35.9%) 1355 (36.5%) 218 (40.4%) 919 (40.1%)
POS 24 (2.7%) 89 (2.4%) 143 (26.5%) 615 (26.8%)
PFFS 196 (22.3%) 806 (21.7%) N/A N/A
IMV N/A N/A 81 (15.0%) 319 (13.9%)
Baseline comorbidity, N (%)
Erectile dysfunction 210 (23.9%)* 588 (19.1%) 87 (16.1%)* 243 (10.6%)
Other male genital organs
disease
259 (29.5%) 1004 (27.1%) 83 (15.4%) 304 (13.3%)
Prostate cancer 73 (8.3%) 300 (8.1%) 20 (3.7%) 46 (2.1%)
Diabetes 184 (21.0%) 768 (20.7%) 59 (10.9%) 258 (11.3%)
Hypertension 456 (51.9%) 1919 (51.8%) 133 (24.6%) 600 (26.2%)
92

Hyperlipidemia 488 (55.6%) 2042 (55.1%) 166 (30.7%) 713 (31.1%)
Urinary disease 58 (6.6%) 209 (5.6%) 25 (4.6%) 91 (4.0%)
Smoking 50 (5.7%) 202 (5.5%) 20 (3.7%) 94 (4.1%)
Alcoholism 9 (1.0%) 44 (1.2%) 5 (0.9%) 23 (1.0%)
Dupuytren’s contracture
b
15 (1.7%)* 0 (0.0%) 11 (2.0%)* 0 (0.0%)
Ledderhose disease
b
27 (3.1%)* 0 (0.0%) 12 (2.2%)* 0 (0.0%)
Chronic ischemic heart
disease
166 (18.9%) 722 (19.5%) 26 (4.8%) 126 (5.5%)
Peripheral vascular disease 78 (8.9%) 262 (7.1%) 5 (0.9%) 27 (1.2%)
Cerebrovascular disease 49 (5.6%) 231 (6.2%) 6 (1.1%) 35 (1.5%)
Chronic lung disease 110 (12.5%) 443 (12.0%) 30 (5.6%) 110 (4.8%)
Chronic renal disease 67 (7.6%) 324 (8.7%) 8 (1.5%) 32 (1.4%)
Chronic liver disease 16 (1.8%) 64 (1.7%) 9 (1.7%) 37 (1.6%)
Myocardial infarct 30 (3.4%) 179 (4.8%) 1 (0.2%) 14 (0.6%)
Congestive heart failure 31 (3.5%) 172 (4.6%) 4 (0.7%) 21 (0.9%)
Depression 32 (3.6%) 125 (3.4%) 22 (4.1%) 89 (3.9%)
*p<0.05. Chi-square tests or Fisher's exact tests were used for comparisons of categorical
variables; Wilcoxon rank-sum tests were used for comparisons of continuous variables.
a
Other race includes Asian, Native American, etc.
b
Controls were identified among enrollees who had no Dupuytren’s contracture (ICD-9-CM:
728.6, 718.44), Ledderhose disease (ICD-9-CM: 728.71), or PD (ICD-9-CM: 607.85) diagnoses
in their claims histories between 2007 and 2013.

93

Table 4.2.a: Resource use during the 3-year study period
Post-index resource utilization,
N (%), (≥ one event)
Medicare Advantage cohort Commercial cohort
PD (n= 595) No PD
(n=2503)
PD (n=244) No PD
(n=1008)
Inpatient stay 212 (35.6%) 856 (34.2%) 41 (16.8%) 150 (14.9%)
Emergency department visit 278 (46.7%) 1122 (44.8%) 73 (29.9%) 314 (31.2%)
Outpatient visit 595 (100.0%)* 2479 (99.0%) 244 (100.0%)* 971 (96.3%)
Outpatient hospital 487 (81.9%)** 1824 (72.9%) 161 (66.0%)** 497 (49.3%)
Ambulatory surgical center 261 (43.9%)** 818 (32.7%) 62 (25.4%) 216 (21.4%)
Any prescription drug use 539 (90.6%)* 2188 (87.4%) 206 (84.4%)** 699 (69.4%)
Selected drug use
PD-related medications
a
122 (20.5%)** 156 (6.2%) 64 (26.2%)** 55 (5.5%)
ED drugs
b
186 (31.3%)** 461 (18.4%) 89 (36.5%)** 155 (15.4%)
Analgesics 401 (67.4%)* 1570 (62.7%) 179 (69.7%)** 563 (55.9%)
Anti-anxiety medications 113 (19.0%)* 381 (15.2%) 59 (24.2%)** 135 (13.4%)
Anti-depressants 126 (21.2%)* 439 (17.5%) 43 (17.6%) 164 (16.3%)
Anti-diabetics 116 (19.5%) 514 (20.5%) 27 (11.1%) 117 (11.6%)
Anti-hyperlipidemics 361 (60.7%) 1532 (61.2%) 91 (37.3%) 400 (39.7%)
Anti-hypertensives 349 (58.7%) 1512 (60.4%) 79 (32.4%) 349 (34.6%)
Cardiovascular agents 203 (34.1%)** 656 (26.2%) 90 (36.9%)** 180 (17.9%)
Corticosteroids 281 (47.2%)** 994 (39.7%) 121 (49.6%)** 325 (32.2%)
PD surgery resource use
Plication procedures 26 (4.4%)** 0 (0.0%) 6 (2.5%)** 0 (0.0%)
Penile plaque excision 11 (1.9%)** 0 (0.0%) 6 (2.5%)** 0 (0.0%)
Penile prosthesis
implantation or replacement
c

58 (9.8%)** 11 (0.4%) 5 (2.1%)** 0 (0.0%)
a
Off-label use of PD-related prescriptions were retrieved from published literature and expert
opinion.
b
ED drugs were retrieved from MICROMEDEX® 2.0.
c
Penile prosthesis could be used for treating PD and/or ED.
*p<0.05, **p<0.01, Chi-square or Fisher’s exact tests were used for comparisons of categorical
variables.

94

Table 4.2.b: Selected drug use at baseline
Baseline drug use, N (%), (≥ one
event)
Medicare Advantage cohort Commercial cohort
PD (n= 595) No PD
(n=2503)
PD (n=244) No PD
(n=1008)
PD-related medication
a
29 (4.9%) 115 (4.6%) 7 (2.9%) 24 (2.4%)
ED drugs
b
85 (14.3%)* 251 (10.0%) 18 (7.4%) 63 (6.3%)
Analgesics 59 (24.2%) 202 (20.0%) 171 (28.7%) 653 (26.1%)
Anti-anxiety medications 44 (7.4%) 170 (6.8%) 14 (5.7%) 48 (4.8%)
Anti-depressants 62 (10.4%) 239 (9.6%) 21 (8.6%) 87 (8.6%)
Cardiovascular agents 67 (11.3%) 232 (9.3%) 16 (6.6%) 54 (5.4%)
Corticosteroids 76 (12.8%) 281 (11.2%) 24 (9.8%) 94 (9.3%)
a
off-label use of PD-related prescriptions might not necessarily be used to treat PD .
b
ED drugs were retrieved from MICROMEDEX® 2.0.
*p<0.05, Chi-square or Fisher’s exact tests were used for comparisons of categorical variables.

95

Table 4.3.a: Distribution of healthcare costs in MA cohort (unadjusted)
Healthcare costs
($ per 6 months)
N= 879 PD vs. 3708 controls N=711 PD vs. 3034
controls
N=595 PD vs. 2503
controls
Baseline 1-6 7-12 13-18 19-24 25-30 31-36
Total costs in PD
a

Mean 4047 6146** 4791* 4886** 4671 4342 4758
IQR 2730 4937 3179 3332 3208 3009 3139
90
th
percentile 8566 16004 10888 10543 9442 8061 9695
Total costs in control
a

Mean 4130 4188 4627 4348 4775 4018 4591
IQR 2982 2977 2949 2901 3024 2837 2837
90
th
percentile 8681 9201 9591 9236 9937 8604 9065
Outpatient costs in PD

Mean 1877 3237** 2174** 2409** 2175** 2269** 2537**
IQR 1562 2464 1922 1976 1765 1826 1819
90
th
percentile 4100 7703 4935 5273 4828 4434 4989
Outpatient costs in control
Mean 1808 1703 1864 1621 1818 1765 1740
IQR 1444 1453 1412 1371 1421 1401 1400
90
th
percentile 4120 3694 4079 3714 3863 3639 3729
Prescription costs in PD
Mean 927 967** 971* 953 937 821 889
IQR 1158 1171 1190 1067 1093 975 1043
90th percentile 2195 2202 2171 1953 2017 1985 1935
Prescription costs in control
Mean 898 901 890 866 866 800 815
IQR 1141 1110 1087 1046 1039 970 952
90th percentile 2091 2104 2006 2010 2085 1961 1896
a
Total costs include all-cause healthcare costs, such as inpatient, outpatient, emergency
department, home health services and prescription medication costs
*p<0.05, **p<0.01. Wilcoxon rank-sum tests were used for comparisons of cost variables. All
costs were reported in 2014 US dollars
IQR: interquartile range (difference between the 75th and 25th percentile estimates)

96

Table 4.3.b: Distribution of costs in commercial cohort (unadjusted)
Healthcare
costs ($ per 6
months)
N= 540 PD vs. 2291 controls N=359 PD vs. 1493
controls
N=244 PD vs. 1008
controls
Baseline 1-6 7-12 13-18 19-24 25-30 31-36
Total costs in PD
a

Mean 4837 5318** 4154** 3966** 5081** 3677* 3523
IQR 2639 3342 2865 2825 3106 2857 2835
90th percentile 7198 8625 7560 5925 8705 8616 7554
Total costs in control
a

Mean 4045 3707 3412 2933 3691 3400 3949
IQR 2337 2255 2243 2232 2349 2284 2235
90th percentile 7470 6720 6237 6045 6604 5897 6615
Outpatient costs in PD

Mean 1625 2730** 1907** 2669** 2691** 1420** 1432**
IQR 1167 1872 1392 1412 1330 1116 1210
90th percentile 3997 5513 4056 3873 4515 4530 3747
Outpatient costs in control
Mean 1657 1490 1451 1223 1505 1329 1400
IQR 1088 916 849 881 873 899 839
90th percentile 3643 3162 2981 2604 3196 2705 2887
Prescription costs in PD
Mean 736 903** 920** 983** 945* 877 856
IQR 857 1010 1063 1096 1082 992 1022
90th percentile 2058 2458 2330 2299 2171 2329 2287
Prescription costs in control
Mean 706 762 785 821 818 897 871
IQR 784 853 846 881 842 818 766
90th percentile 1911 1925 1944 2077 2066 1951 1956
a
Total costs include all-cause healthcare costs, such as inpatient, outpatient, emergency
department, home health services and prescription medication costs
*p<0.05, **p<0.01. Wilcoxon rank-sum tests were used for comparisons of cost variables. All
costs were reported in 2014 US dollars
IQR: interquartile range (difference between the 75th and 25th percentile estimates)
97

1:5 matching*
100% male population of Humana enrollees 01/01/2007-6/30/2013 (1.61M)
Patients with ≥ 1 PD diagnosis (ICD-9-CM: 607.85) during study intake period from
07/01/2007-06/30/2010 (n=3115)
1419 from MA aged ≥ 65 (n=879) or commercial
plans aged 18-64 (n= 540)

Continuous enrollment of ≥ 6 months before (baseline) and ≥ 1 year after the index date
(study period) (n=1702), a tt ri tio n = 4 5 %
261 from MA aged <65 (n=233)
or commercial plans aged ≥ 65
(n=28)
Disease group (n=1419) with 879 from MA and 540 from commercial plans
Control group (n=5999) with 3708 from MA and 2291 from commercial plans
Medicare Advantage (MA) plan type: HMO, PPO, POS, PFFS
‡
or commercial plan type:
HMO, PPO, POS, IMV
§
(n=1680), excluding 22 patients from Medicare supplemental plan
Disease group (n=1070) with 711 from MA and 359 from commercial plans
Control group (n=4527) with 3034 from MA and 1493 from commercial plans
Disease group (n=839) with 595 from MA and 244 from commercial plans
Control group (n=3511) with 2503 from MA and 1008 from commercial plans
Continuous enrollment of ≥ 2 years after the index date, a tt ri tio n = 2 5 %
Continuous enrollment of ≥ 3 years after the index date, a tt ri tio n = 2 2 %
Figure 4.1: Study Scheme at Base Case

‡PFFS: Medicare advantage private fee-for-service plan
§ IMV: Individual Major Medical plan tailored for individuals, families or small businesses
*Each PD patient was propensity score matched to controls at a 1-to-5 variable ratio
Attrition in this sample resulted in smaller cohorts over time and attrition rates are
comparable between disease and control groups
98

Figure 4.2: Adjusted Incremental Total Costs per 6 months

*p<0.05 and **p<0.01
y axis: adjusted mean incremental cost (+Standard Error) per patient per 6 months in PD over base-
case controls, estimated by using the “margins” command in STATA
$2,746*, $1,058, $793, and $1,754* in post-index 1-6, 7-12, 13-18, 19-24 months in commercial cohort
$2,610**, $189, and $943* for the post-index 1-6, 7-12 and 13-18 months in MA cohort
99

Figure 4.3.a: Adjusted Incremental OUTPATIENT Costs

Figure 4.3.b: Adjusted Incremental Prescription DRUG Costs
y axis: adjusted incremental outpatient or prescription costs per 6 months (+Standard Error) in
PD patients over matched controls in the base-case scenario
*p<0.05 and **p<0.01
100

Chapter-Four Appendix
Appendix Section A.
Table A.4.1: List of ICD-9 codes and CPT procedure codes
Diagnosis or procedure codes Description
ICD-9 diagnosis codes 607.85 Other specified disorders of penis: Peyronie's
disease
ICD-9 procedure codes 64.11 Biopsy of penis
64.19 Other diagnostic procedures on penis
64.2 Local excision or destruction of lesion of penis
64.4^ Repair and plastic operation on penis
64.92 Incision of penis
64.94* Fitting of external prosthesis of penis
64.95* Insertion or replacement of non-inflatable penile
prosthesis
64.96* Removal of internal prosthesis of penis
64.97* Insertion or replacement of inflatable penile
prosthesis
64.98 Other operations on penis (Corpora cavernosa-
corpus spongiosum shunt; Corpora-saphenous
shunt; Irrigation of corpus cavernosum)
64.99 Other Operations on Male Genital Organs, Other
CPT procedure codes
54110 Excision of penile plaque (Peyronie's disease)
54111 Excision of penile plaque (Peyronie's disease); with
graft to 5 cm in length
54112 Excision of penile plaque (Peyronie's disease); with
graft greater than 5 cm in length
54200 Injection procedure for Peyronie's disease
54205 Injection procedure for Peyronie's disease; with
surgical exposure of plaque
54360 Plastic operation on penis to correct angulation
54400* Insertion of penile prosthesis; non-inflatable (semi-
rigid)
54401* Insertion of penile prosthesis; inflatable (self-
contained)
54402* Removal or replacement of non-inflatable or
inflatable penile prosthesis
* most occur with diagnosis of both PD and ED (ICD 9-CM: 302.70, 607.82, 607.89)
^ most occur with diagnosis of impotence of organic origin (ICD-9-CM: 607.84)
101

Appendix Section B. Sensitivity Analyses with Alternative Control Groups
Table A.4.2: Cost comparison between PD and physician office visit controls
Healthcare costs ($ per
6 months)
Total costs Outpatient costs
Baseline 1-6 7-12 Baseline 1-6 7-12
PD in Medicare cohort (N=879)
Mean 4047 6146 4791 1877 3237 2174
IQR 2730 4937 3179 1562 2464 1922
90th percentile 8566 16004 10888 4100 7703 4935
Alternative control in Medicare cohort (N=2723)
Mean 3996 6093 5216 1844 2977 1886
IQR 3020 4365 3466 1724 2579 1784
90th percentile 8874 13831 11870 4421 6345 4488
PD in commercial cohort (N=540)
Mean 4837 5318 4154 1625 2730 1907
IQR 2639 3342 2865 1167 1872 1392
90th percentile 7198 8625 7560 3997 5513 4056
Alternative control in commercial cohort (N=1560)
Mean 4942 5444 4598 1731 3189 2259
IQR 2976 3414 3143 1711 2089 1445
90 percentile 8109 8771 9883 4664 6083 4357
IQR: interquartile range (difference between the 75th and 25th percentile estimates)
All comparisons p>0.05: no significant cost difference was observed in PD patients compared to
alternative controls that have at least a physician office visit on their index date

102

Table A.4.3. Cost comparison between PD and benign prostatic hypertrophy patients
Healthcare costs ($ per
6 months)
Total costs Outpatient costs
Baseline 1-6 7-12 Baseline 1-6 7-12
PD in Medicare cohort (N=879)
Mean 4047 6146 4791 1877 3237 2174
IQR 2730 4937 3179 1562 2464 1922
90th percentile 8566 16004 10888 4100 7703 4935
Benign prostatic hypertrophy (BPH) in Medicare cohort (N=3690)
Mean 4146 6534 5128 1757 2990 2187
IQR 3108 5196 3435 1625 2780 1819
90th percentile 8574 16204 11475 3998 6279 4750
PD in commercial cohort (N=540)
Mean 4837 5318 4154 1625 2730 1907
IQR 2639 3342 2865 1167 1872 1392
90th percentile 7198 8625 7560 3997 5513 4056
BPH in commercial cohort (N=2225)
Mean 4530 5854 4555 1675 3028 2250
IQR 3133 3993 2998 1314 2017 1505
90 percentile 8041 10931 8013 4150 5964 4501
IQR: interquartile range (difference between the 75th and 25th percentile estimates)
All comparisons p>0.05: no significant cost difference was observed in PD patients compared to
patients diagnosed with benign prostatic hypertrophy

103

Figure A.4.1. GLM-adjusted Incremental Outpatient Costs in PD patients over Benign
Prostatic Hypertrophy patients
y axis: adjusted incremental outpatient costs per 6 months (+Standard Error) in PD patients
over patients diagnosed with benign prostatic hypertrophy (BPH)
*p<0.05 and **p<0.01

104

Appendix Section C.
Table A.4.4. Subgroup analysis of the base-case excluded PD patients
Costs ($ per 6 months) Total costs Outpatient costs
Baseline 1-6 7-12 Baseline 1-6 7-12
PD aged < 65 in Medicare cohort (n=233): HMO(n=70), PPO(n=88), PFFS(n=73), POS (n=2)
Mean 6798 8531 8855 3013 4214 2660
IQR 5470 7248 6653 2602 2795 2568
90th percentile 12642 22875 15398 7132 11534 7236
PD aged ≥ 65 in commercial cohort (n=28): HMO (n=8), PPO (n=10), POS (n=10)
Mean 4718 8369 5347 2686 5724 2198
IQR 6287 8011 3718 3534 3732 3125
90 percentile 10934 33482 15677 8060 22536 7075
IQR: interquartile range (difference between the 75th and 25th percentile estimates)

105

Appendix Section D. Test for Attrition Bias in the base-case scenario
Table A.4.5. Seemingly unrelated estimation results evaluating difference in coefficients
of GLM model between full and completing samples
Dependent variable =
total costs
Medicare Advantage cohort Commercial cohort
Joint test for equality of
coefficients between full
and completing samples
Chi-squared statistic 79.96
p-value <0.001
Chi-squared statistic 46.41
p-value = 0.0017
Characteristics Full Completing p >|chi2| Full Completing p >|chi2|
Constant 6.609 6.951 0.0856 6.455 6.613 0.4197
Age 0.024 0.018 0.0599 0.037 0.035 0.3595
Diagnosis indicator 0.062 0.117 <0.05 0.350 0.234 0.0534
White 0.204 0.170 0.2497 N/A N/A N/A
Black 0.231 0.157 0.3059 N/A N/A N/A
Hispanic 0.134 0.049 <0.05 N/A N/A N/A
Other or unknown race 0.118 0.389 0.4591 0.329 0.358 0.8888
Northeast -0.425 -0.414 0.9034 0.496 0.826 0.1558
South 0.102 0.075 0.2801 -0.083 -0.057 0.6878
West -0.060 -0.001 0.1128 -0.011 -0.290 0.1049
Plan type
HMO -0.002 -0.131 <0.01 -0.546 -0.378 <0.05
PFFS
‡
0.090 -0.094 <0.01 N/A N/A N/A
POS 0.031 -0.066 0.1768 -0.081 -0.154 0.3569
PPO -0.004 -0.098 <0.05 -0.024 -0.188 <0.05
Study period (vs.
baseline)

1-6 months 0.083 0.065 0.6213 -0.135 -0.133 0.9862
7-12 months 0.092 0.035 0.1585 -0.265 -0.288 0.8578
13-18 months 0.104 0.077 0.4777 -0.363 -0.483 0.3092
19-24 months 0.083 0.076 0.8418 -0.192 -0.206 0.9113
25-30 months 0.001 0.019 0.6864 -0.202 -0.258 0.6506
31-36 months 0.044 0.132 <0.01 -0.162 -0.111 0.6286
‡
PFFS: Medicare advantage private fee-for-service plan

106

Table A.4.6. Cross-sectional probit regressions to predict the probability of attrition on
characteristics variables and lagged outcomes
Dependent variable
P(Attrit=1)
Medicare Advantage cohort Commercial cohort
13-24 months 25-36 months 13-24 months 25-36 months
N 3,745 3,098 1,852 1,252
Pseudo R-sq 0.139 0.104 0.020 0.027
Covariates Coeff.[SE] Coeff.[SE] Coeff.[SE] Coeff.[SE]
Auxiliary variables
1

Ln(total costs)
(t-1)
0.05* 0.04 -0.00002 -0.01
[0.03] [0.02] [0.020] [0.02]
Ln(total costs)
(t-2)
0.02 0.01 -0.009 0.04
[0.02] [0.02] [0.019] [0.03]
Ln(total costs)
(t-3)
N/A 0.02 N/A 0.01
[0.02] [0.02]
Characteristics Variables
Age at baseline period 0.007 0.006 -0.003 0.006
[0.005] [0.005] [0.003] [0.004]
Diagnosis indicator 0.001 -0.04 -0.08 -0.06
(PD vs. control) [0.06] [0.06] [0.07] [0.08]
White -1.88** -2.40** N/A N/A
[0.11] [0.28]
Black -1.76** -2.28** N/A N/A
[0.14] [0.29]
Hispanic -1.75** -2.34** N/A N/A
[0.20] [0.32]
Other or unknown race -1.78** -2.29** 0.14 0.48*
[0.25] [0.36] [0.17] [0.20]
Northeast 0.13 0.24 0.32 -0.23
[0.21] [0.20] [0.64] [0.93]
South 0.08 0.03 0.07 -0.03
[0.06] [0.06] [0.07] [0.08]
West 0.14 0.22* 0.11 0.15
[0.09] [0.09] [0.12] [0.14]
HMO 0.15 0.02 -0.07 -0.03
[0.25] [0.21] [0.10] [0.12]
PFFS 0.65** 0.59** N/A N/A
(private fee-for-service) [0.25] [0.22]
POS 0.31 -0.16 0.16* 0.24*
[0.32] [0.29] [0.08] [0.09]
PPO 0.33 0.14 0.23** 0.29**
[0.25] [0.21] [0.08] [0.09]
1
Fitzgerald, J., Gottschalk, P., and Moffitt, R. (1998). Journal of Human Resources 33 (2).
SE: standard errors in brackets; *p<0.05; **p<0.01.
107

Table A.4.7. Becketti, Gould, Lilard and Welch (BGLW) test for random attrition
Dependent variable = Ln
(total costs)
(t-1)

Medicare Advantage cohort
Commercial cohort
13-24 months 25-36 months 13-24 months 25-36 months
N
3,745 3,098 1,852 1,252
F-statistic for joint
significance of Att and IattX*
1.25 1.54 1.12 1.55
|p-value| > F
0.2397 0.0899 0.3357 0.1003
Covariates Coeff.[SE] Coeff.[SE] Coeff.[SE] Coeff.[SE]
Att (1 = attrition) -0.29 0.77 0.17 -0.31
[0.69] [1.03] [0.24] [0.26]
Att*(Age at baseline period) 0.05 0.05 -0.05 0.23
[0.17] [0.17] [0.19] [0.20]
Att*(Diagnosis indicator) -0.03 -0.26 -0.24 0.10
[0.14] [0.14] [0.19] [0.18]
Att*White 0.34 -0.66 N/A N/A
[0.26] [0.88]
Att*Black 0.26 -0.78 N/A N/A
[0.34] [0.90]
Att*Hispanic 0.37 -0.79 N/A N/A
[0.50] [0.97]
Att*(Other or unknown race) -0.02 -0.95 0.37 -0.40
[0.63] [1.03] [0.41] [0.46]
Att*Northeast 0.18 0.39 0.06 -2.18
[0.46] [0.46] [1.59] [2.24]
Att*South -0.002 0.008 -0.21 -0.39
[0.14] [0.14] [0.16] [0.17]
Att*West -0.12 -0.34 -0.27 0.09
[0.20] [0.20] [0.28] [0.31]
Att*HMO 0.48 0.14 0.06 0.06
[0.63] 0.53 [0.25] [0.26]
Att*PFFS 0.11 0.11 N/A N/A
(private fee-for-service) [0.63] [0.53]
Att*POS -0.24 -0.71 0.14 0.67*
[0.82] [0.74] [0.20] [0.22]
Att*PPO 0.14 0.12 0.24 0.43*
[0.63] [0.53] [0.19] [0.20]
Constant 8.44** 7.74** 7.06** 7.15**
[0.32] [0.90] [0.14] [0.17]
R-sq 0.031 0.021 0.094 0.091
Att = attrition dummy (0 = no attrition / 1 = attrition during each period);
IattX* = interaction of attrition dummy with characteristics covariates;
SE: standard errors in brackets; *p<0.05; **p<0.01.
108

Chapter-Four References
Aiken, L.H., Sloane, D.M., Bruyneel, L., Van den Heede, K., Griffiths, P., Busse, R., et al.
(2014). Nurse staffing and education and hospital mortality in nine European countries:
a retrospective observational study. Lancet 383(9931), 1824-1830.
Arafa, M., Eid, H., El-Badry, A., Ezz-Eldine, K., Shamloul, R. (2007). The prevalence of
Peyronie's disease in diabetic patients with erectile dysfunction. Int J Impot Res 19(2),
213-217.
Austin, P.C. (2010). Statistical criteria for selecting the optimal number of untreated
subjects matched to each treated subject when using many-to-one matching on the
propensity score. Am J Epidemiol 172(9), 1092-1097.
Barber, J. and Thompson, S. (2004). Multiple regression of cost data: use of
generalised linear models. J Health Serv Res Policy 9(4), 197-204.
Becketti, S., Gould, W., Lillard, L., Welch, F. (1988). The panel study of income
dynamics after fourteen years: an evaluation. Journal of Labor Economics 6(4), 472-492
Bella, A.J., Perelman, M.A., Brant, W.O., Lue, T.F. (2007). Peyronie's disease (CME). J
Sex Med 4(6), 1527-1538.
Clogg, C.C., Petkova, E., Haritou, A. (1995). Statistical methods for comparing
regression coefficients between models. American Journal of Sociology 100(5), 1261-
1293
Deb, P., Manning, W., Norton, E. ( 2006). Modeling health care costs and counts.
ASHE-Madison Conference. Available at:
http://www.unc.edu/~enorton/DebManningNortonPresentation.pdf
Deveci, S., Palese, M., Parker, M., Guhring, P., Mulhall, J. P. (2006). Erectile function
profiles in men with Peyronie's disease. J Urol 175(5), 1807-1811.
109

Dibenedetti, D.B., Nguyen, D., Zografos, L., Ziemiecki, R., Zhou, X. (2011). A
population-based study of Peyronie's disease: prevalence and treatment patterns in the
United States. Adv Urol, 282503.
Emberton, M., Cornel, E.B., Bassi, P.F., Fourcade, R.O., Gó mez, J.M., Castro, R.
(2008). Benign prostatic hyperplasia as a progressive disease: a guide to the risk
factors and options for medical management. Int J Clin Pract 62(7), 1076-1086
Fitzgerald, J.M., Gottschalk, P., and Moffitt, R. (1998). An analysis of sample attrition in
panel data: the michigan panel study of income dynamics. Journal of Human Resources
33(2), 251-299
Fitzgerald, J.M. (2011). Attrition in models of intergenerational links using the psid with
extensions to health and to sibling models. B E J Econom Anal Policy 11(3)
Hardin, J.W. and Hilbe, J.M. Generalized linear models and extensions, second edition,
2007, publisher: Stata Press.
Jalkut, M., Gonzalez-Cadavid, N., Rajfer, J. (2003). Peyronie's disease: a review. Rev
Urol 5(3), 142-148.
Joyce, G.F., Zissimopoulos, J., Goldman, D.P. (2013). Digesting the doughnut hole. J
Health Econ 32(6), 1345-1355.
Joyce, G.F., Goldman, D.P., Karaca-Mandic, P., Lawless, G.D. (2008). Impact of
specialty drugs on the use of other medical services. Am J Manag Care 14(12), 821-828.
Kadioglu, A., Sanli, O., Akman, T., Canguven, O., Aydin, M., Akbulut, F., Kucukdurmaz,
F. (2011). Factors affecting the degree of penile deformity in Peyronie disease: an
analysis of 1001 patients. J Androl 32(5), 502-508.
Kadioglu, A., Tefekli, A., Erol, B., Oktar, T., Tunc, M., Tellaloglu, S. (2002). A
retrospective review of 307 men with Peyronie's disease. J Urol 168(3), 1075-1079.
110

Kendirci, M., Trost, L., Sikka, S. C., Hellstrom, W. J. (2007). Diabetes mellitus is
associated with severe Peyronie's disease. BJU Int 99(2), 383-386.
Kessler, D. and McClellan, M. (1996). Do doctors practice defensive medicine? The
Quarterly Journal of Economics 111, 353-390.
Kuehhas, F. E., Weibl, P., Georgi, T., Djakovic, N., Herwig, R. (2011). Peyronie's
disease: nonsurgical therapy options. Rev Urol 13(3), 139-146.
Langston, J.P. and Carson, C.C. (2014). Peyronie's disease: review and recent
advances. Maturitas 78(4), 341-343
Levine, L.A. and Greenfield, J.M. (2003). Establishing a standardized evaluation of the
man with Peyronie's disease. Int J Impot Res 15, S103-112.
Macaulay, D., Ivanova, J., Birnbaum, H., Sorg, R., Skodny, P. (2012). Direct and
indirect costs associated with Dupuytren's contracture. J Med Econ 15(4), 664-671.
Manning, W.G. and Mullahy, J. (2001). Estimating log models: to transform or not to
transform? J Health Econ 20, 461-494.
McVary, K.T., Roehrborn, C.G., Avins, A.L., Barry, M.J., Bruskewitz, R.C., Donnell, R.F.,
et al. (2011). Update on AUA guideline on the management of benign prostatic
hyperplasia. J Urol 185(5), 1793-1803.
Mulhall, J. P., Schiff, J., Guhring, P. (2006). An analysis of the natural history of
Peyronie's disease. J Urol 175(6), 2115-2118.
Nugteren, H.M., Nijman, J.M., de Jong, I.J., van Driel, M.F. (2011). The association
between Peyronie's and Dupuytren's disease. Int J Impot Res 23(4), 142-145.
Rassen, J.A., Shelat, A.A., Myers, J., Glynn, R.J., Rothman, K.J., Schneeweiss, S.
(2012) One-to-many propensity score matching in cohort studies. Pharmacoepidemiol
Drug Saf 21, 69-80.
111

Rosen, R., Catania, J., Lue, T., Althof, S., Henne, J., Hellstrom, W., et al. (2008). Impact
of Peyronie's disease on sexual and psychosocial functioning: qualitative findings in
patients and controls. J Sex Med 5(8), 1977-1984.
Rosenbaum, P.R. and Rubin, D. B. (1985). Constructing a control group using
multivariate matched sampling methods that incorporate the propensity score. American
Statistician 39(1), 33-38.
Schwarzer, U., Sommer, F., Klotz, T., Braun, M., Reifenrath, B., Engelmann, U. (2001).
The prevalence of Peyronie's disease: results of a large survey. BJU Int 88(7), 727-730.
Shafazand, S., Yang, Y., Amore, E., O'Neal, W., Brixner, D. (2010). A retrospective,
observational cohort analysis of a nationwide database to compare heart failure
prescriptions and related health care utilization before and after publication of updated
treatment guidelines in the United States. Clin Ther 32(9), 1642-1650.
Smith, J.F., Brant, W.O., Fradet, V., Shindel, A.W., Vittinghoff, E., Chi, T., et al. (2009).
Penile sonographic and clinical characteristics in men with Peyronie's disease. J Sex
Med 6(10), 2858-2867.
Smith, J.F., Walsh, T.J., Conti, S.L., Turek, P., Lue, T. (2008). Risk factors for emotional
and relationship problems in Peyronie's disease. J Sex Med 5(9), 2179-2184.
Tal, R., Heck, M., Teloken, P., Siegrist, T., Nelson, C.J., Mulhall, J.P. (2010). Peyronie's
disease following radical prostatectomy: incidence and predictors. J Sex Med 7(3),
1254-1261.
Taylor, F.L. and Levine L.A. (2007). Peyronie's Disease. Urol Clin North Am 34(4), 517-
534.
Taylor, F.L. and Levine, L. A. (2008). Surgical correction of Peyronie's disease via
tunica albuginea plication or partial plaque excision with pericardial graft: long-term
follow up. J Sex Med 5(9), 2221-2228.
112

US Department of Labor Statistics, Consumer Price Index. item: medical care. Available:
http://data.bls.gov/labjava/outside.jsp?survey=cu (accessed November 11, 2014).
Usta, M.F., Bivalacqua, T.J., Jabren, G.W., Myers, L., Sanabria, J., Sikka, S.C., et al.
(2004). Relationship between the severity of penile curvature and the presence of
comorbidities in men with Peyronie's disease. J Urol 171, 775-779.
Usta, M.F., Bivalacqua, T.J., Jabren, G.W., Myers, L., Sanabria, J., Sikka, S.C., et al.
(2004). Stratification of penile vascular pathologies in patients with Peyronie's disease
and in men with erectile dysfunction according to age: a comparative study. J Urol
172(1), 259-262.
Wessells, H., Joyce, G. F., Wise, M., Wilt, T. J. (2007). Chapter 15: Erectile dysfunction
and Peyronie’s DISEASE. Urologic Diseases in America. US Department of Health and
Human Services, Public Health Service, National Institutes of Health, National Institute
of Diabetes and Digestive and Kidney Diseases. Washington, DC: US Government
Printing Office, NIH Publication No. 07-5512.

113

CHAPTER 5: CONCLUSIONS
This dissertation focused on the applications of different economic evaluation
methods to provide health economic and outcomes-based evidence for health
technology assessment (HTA). Through these studies, we showed that economic
evaluation analyses can be conducted from a societal or a population perspective. For
example, costs reported in chapter 2 and 3 were estimated from the societal and the
healthcare system perspective; while costs in chapter 4 were restricted to medical
claims from a third party payer’s perspective. Using a rare disease (i.e., Peyronie’s
disease without an effective drug specifically labeled for it before 2014) as an example
in chapter 4, we also showed that the cost of illness (COI) study provides a partial
economic evaluation, as it aims to estimate the preventable economic burden of a
disease that may not have an effective intervention.

On the other hand, cost-benefit analysis (CBA) is a comprehensive economic
evaluation technique, which requires a sequence of partial analyses to collect cost data
from COI studies and health outcomes data, such as clinical validity of treatment, from
systematic evidence review or clinical trials. Therefore, it is useful in guiding decision
makers to determine the best prevention intervention or treatment with respect to the
disease studied, and assisting policy makers to fund or to evaluate the interventions,
such as the postpartum vaccination program studied in chapter 2 and 3.

This dissertation also introduces new approaches for conducting sensitivity
analyses and advances in making economic assumptions for decision analysis
114

modeling. All the three studies performed sensitivity analyses to address model
validation and precision. For example, study 1 and 2 performed the one-way, two-way
and threshold sensitivity analyses for key input variables. Study 3 in chapter 4 provides
an alternative way in conducting sensitivity analyses by generating alternative
comparison groups in addition to the base-case control cohort to estimate the
incremental healthcare costs between cases and controls. Furthermore, the diagram of
the decision analysis modeling developed in chapter 2 (Fig.2.1) is consistent with the
clinical pathway of postpartum vaccination process, which can be directly applied to pre-
and intra-partum vaccination strategies. As it takes into account that the risk of an infant
acquiring an infectious disease was not independent of the infectious status of their
caregivers or household contacts, this model also can be adapted to other populations,
such as children and elderly, for the infectious disease prevention.

In summary, the findings of these studies provide insight for stakeholders,
including patients, healthcare providers and payers, into the applicability of economic
evaluation analyses to aid decision making. At the same time, the economic or
econometric approaches employed in these studies, such as decision analysis modeling,
generalized linear model, fixed effects model with a difference-in-difference design, and
the application of sensitivity analyses to control uncertainty by using different
econometric models or by creating alternative comparison groups, may find wider
acceptance in HTA research for other diseases.

Abstract (if available)

Abstract Health technology assessment (HTA) is a multidisciplinary process that assesses and evaluates the economic, clinical, social, and ethical impacts of health care technologies. HTA incorporates number of areas of expertise including clinical research, epidemiology, health services research, economics, and psychometrics. The field has rapidly expanded in the last decade and played a crucial role in improvement of the quality of healthcare. ❧ This three‐paper dissertation demonstrates two common approaches of economic evaluation in HTA: Paper 1 and 2 are cost‐benefit analyses (CBAs), with cost data obtained from cost‐of‐illness (COI) studies and health outcomes data from systematic evidence review or clinical trials, to compare between different healthcare programs or interventions. Paper 3 is a COI study conducted from a third‐party payer’s perspective with cost data directly collected from administrative claims databases used for health care payment. The cost data from one or more such sources often are combined with data from primary clinical studies, epidemiological studies, and other sources to conduct the CBAs, cost‐effectiveness analyses and other analyses that involve weighing health and economic impacts of a health technology. Compared to the COI analysis, CBA provides additional evidence that can be used to determine the best prevention intervention or treatment with respect to the disease studied, which can assist policy makers to fund the interventions or to evaluate completing public health programs (e.g., immunization, newborn screening, and water purification).

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Evaluating approaches to reduce inappropriate antibiotic use in the United States

PDF

Economic aspects of obesity

PDF

Selected papers on the evaluation of healthcare costs of prematurity and necrotizing enterocolitis using large retrospective databases

PDF

New approaches using probabilistic graphical models in health economics and outcomes research

PDF

Value in health in the era of vertical integration

PDF

Essays in health economics

PDF

Essays on the economics of infectious diseases

PDF

Essays in pharmaceutical and health economics

PDF

Discriminating changes in health using patient-reported outcomes

PDF

Selected papers on methods for evaluating the impact of marijuana use on BMI and other risk factors for metabolic syndrome

PDF

Meta-analysis methods in the cost-effectiveness analysis of mammography screening

PDF

Health economics and outcomes research for informed decision making in rapidly evolving therapeutic areas

PDF

Three essays on emerging issues in hemophilia care

PDF

Three essays on comparative economic and clinical outcomes of antipsychotics using retrospective claims data

PDF

Impact of pharmacy-based transitional care on healthcare utilization and costs

PDF

Evaluating treatment options for metastatic, castration-resistant prostate cancer: a comprehensive value assessment

PDF

Essays on the use of microsimulation for health and economic policy analysis

PDF

Evaluating cancer treatments with health economics methods

PDF

Hepatitis C in the post-interferon era: selected essays in health economics

PDF

The impact of Patient-Centered Medical Home on a managed Medicaid plan

Asset Metadata

Creator Ding, Yao (author)

Core Title Advances and applications for economic evaluation methods in health technology assessment (HTA)

School School of Pharmacy

Degree Doctor of Philosophy

Degree Program Pharmaceutical Economics and Policy

Publication Date 07/16/2015

Defense Date 05/01/2015

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag cost‐benefit analysis,cost‐effectiveness analysis,cost‐of‐illness study,economic evaluation,health technology assessment,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Hay, Joel W. (committee chair), Sood, Neeraj (committee member), Zangwill, Kenneth M. (committee member)

Creator Email yaoding@usc.edu,ydhelena12@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-594397

Unique identifier UC11299798

Identifier etd-DingYao-3606.pdf (filename),usctheses-c3-594397 (legacy record id)

Legacy Identifier etd-DingYao-3606.pdf

Dmrecord 594397

Document Type Dissertation

Format application/pdf (imt)

Rights Ding, Yao

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