Citation
Concerns, attitudes, beliefs, and intentions of parents about vaccines for their child

Material Information

Title:
Concerns, attitudes, beliefs, and intentions of parents about vaccines for their child
Alternate title:
Development and evaluation of a survey instrument in an integrated health care system in Colorado
Creator:
Shoup, Jo Ann ( author )
Language:
English
Physical Description:
1 electronic file (174 pages) : ;

Subjects

Subjects / Keywords:
Vaccination of children ( lcsh )
Public opinion ( lcsh )
Vaccines ( lcsh )
Vaccination of children -- Public opinion ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
Routine childhood vaccination has led to the eradication, elimination or control of previously common infectious diseases. Through mandates, vaccination rates have remained high. However, parents have concerns about the safety of vaccines. ( ,,, )
Review:
The purpose of this dissertation was to develop, implement and evaluate a theory-driven survey instrument to measure parents’ concerns, attitudes, beliefs, and intentions about vaccines for their child. There were three phases to develop and evaluate the survey instrument. First, the instrument was developed through extensive literature review, use of special matter experts, and cognitive interviews with the target population to establish face and content validity.
Review:
Second, the pilot survey instrument was administered to 120 pregnant mothers. After analyses, further revisions were made. The revised survey was then administered to 320 pregnant mothers and parents of children under twelve months of age. The baseline responses were assessed using factor analysis (FA) and internal consistency (IC) methods to establish construct validity and reliability. FA yielded four factors: Beliefs about Vaccinating; Evaluation of Vaccine Preventable Disease (VPD) and Vaccine Adverse Events (VAE); Subjective Norms about Vaccinating; Perceived Control over Vaccinating Decision. Using repeatability, temporal stability of the instrument was very high, r = .930.
Review:
Finally, association of four survey constructs with undervaccinated status was examined using logistic regression (LR). Univariate analyses demonstrated strong relationships between survey constructs and undervaccination status. In separate models controlling for covariates, significant associations between three survey constructs and undervaccination were found: Beliefs about Vaccinating, (adjusted odds ratio (AOR), 10.01; 95% Confidence Interval (CI) 2.11, 18.00); Evaluation of VPD/VAE (AOR, 24.92, CI 5.22, 119.01); and Subjective Norms about Vaccinating (AOR, 13.65, CI 4.71, 39.53). Additional analyses assessed the survey items for trends over time, and different measures of hesitancy about vaccines were compared across survey items. Overall, parent’s hesitancy about vaccines decreased after the birth of their child. Concerns that continued from pregnancy through six months of age of the child were: concerns about vaccine side effects, ingredients in vaccines, and concerns that vaccines cause autism. Different hesitancy measures were consistent in differentiating survey item responses of those who were hesitant versus nonhesitant.
Thesis:
Thesis (Ph.D.) - University of Colorado Denver.
Bibliography:
Includes bibliographic references
System Details:
System requirements: Adobe Reader.
General Note:
School of Public Affairs
Statement of Responsibility:
by Jo Ann Shoup.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
945931867 ( OCLC )
ocn945931867
Classification:
LD1193.P86 S56 ( lcc )

Downloads

This item has the following downloads:


Full Text
CONCERNS, ATTITUDES, BELIEFS, AND INTENTIONS OF PARENTS ABOUT
VACCINES FOR THEIR CHILD:
DEVELOPMENT AND EVALUATION OF A SURVEY INSTRUMENT IN AN
INTEGRATED HEALTH CARE SYSTEM IN COLORADO
By
JOANNSHOUP
B.A., Edinboro University of Pennsylvania
M.A., Edinboro University of Pennsylvania
M.S.W., University of Pittsburgh
M.S., Carnegie Mellon University
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado Denver in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Public Affairs
2015


This thesis for the Doctor of Philosophy degree by
Jo Ann Shoup
has been approved for the
Public Affairs Program
by
Mary Guy, Advisor
Jessica Sowa, Chair
Danielle Varda
Jason M. Glanz
November 18, 2015


Shoup, Jo Ann (Ph.D., Public Affairs)
Concerns, Attitudes, Beliefs and Intentions of Parents about Vaccines for their Child:
Development and Evaluation of a Survey Instrument in an Integrated Health Care
System in Colorado
Thesis directed by Professor Mary Guy
ABSTRACT
Routine childhood vaccination has led to the eradication, elimination or
control of previously common infectious diseases. Through mandates, vaccination
rates have remained high. However, parents have concerns about the safety of
vaccines.
The purpose of this dissertation was to develop, implement and evaluate a
theory-driven survey instrument to measure parents concerns, attitudes, beliefs,
and intentions about vaccines for their child. There were three phases to develop
and evaluate the survey instrument. First, the instrument was developed through
extensive literature review, use of special matter experts, and cognitive interviews
with the target population to establish face and content validity.
Second, the pilot survey instrument was administered to 120 pregnant
mothers. After analyses, further revisions were made. The revised survey was then
administered to 320 pregnant mothers and parents of children under twelve
months of age. The baseline responses were assessed using factor analysis (FA) and
internal consistency (IC) methods to establish construct validity and reliability. FA
yielded four factors: Beliefs about Vaccinating; Evaluation of Vaccine Preventable
Disease (VPD) and Vaccine Adverse Events (VAE); Subjective Norms about
m


Vaccinating; Perceived Control over Vaccinating Decision. Using repeatability,
temporal stability of the instrument was very high, r = .930.
Finally, association of four survey constructs with undervaccinated status
was examined using logistic regression (LR). Univariate analyses demonstrated
strong relationships between survey constructs and undervaccination status. In
separate models controlling for covariates, significant associations between three
survey constructs and undervaccination were found: Beliefs about Vaccinating,
(adjusted odds ratio (AOR), 10.01; 95% Confidence Interval (Cl) 2.11,18.00);
Evaluation ofVPD/VAE (AOR, 24.92, Cl 5.22,119.01); and Subjective Norms about
Vaccinating (AOR, 13.65, Cl 4.71, 39.53). Additional analyses assessed the survey
items for trends over time, and different measures of hesitancy about vaccines were
compared across survey items. Overall, parents hesitancy about vaccines decreased
after the birth of their child. Concerns that continued from pregnancy through six
months of age of the child were: concerns about vaccine side effects, ingredients in
vaccines, and concerns that vaccines cause autism. Different hesitancy measures
were consistent in differentiating survey item responses of those who were hesitant
versus nonhesitant.
The form and content of this abstract are approved. I recommend its publication.
Approved: Mary Guy
IV


ACKNOWLEDGMENTS
I would like to acknowledge the financial support of the George Bennett
Dissertation Grant through the Informed Medical Decisions Foundation, and the
scientific support of the Institute for Health Research, Kaiser Permanente Colorado.
Without these supports, I would not have been able to pursue the cohort
recruitment and additional methods necessary for reliability and validity.
Additionally, I gratefully thank the University Scholars Committee at the University
of Colorado for their ongoing financial commitment to my academic pursuits.
I would like to thank a number of people who have been central to the
completion of this dissertation. First, to my dissertation committee of Dr. Mary Guy
(chair), Dr. Danielle Varda, Dr. Jessica Sowa and Dr. Jason Glanz (mentor), thank you
for your sage wisdom and advice through this process. A multidisciplinary
dissertation committee has proven a very valuable asset in my development.
Second, I would especially like to thank those who support my endeavors at work
and school Komal Narwaney, Sophia Newcomer, and Nikki Wagner.
v


TABLE OF CONTENTS
CHAPTER
I. REVIEW OF THE LITEATURE................................................1
Introduction to the Problem...............................................1
Purpose of the Dissertation...............................................3
Background on Vaccine Decision Making.....................................4
Role of Public Policy in Vaccination....................................5
Vaccine Hesitancy........................................................13
Decision Making about Vaccines...........................................16
Interventions for Concerns and Hesitancy about Vaccines.................20
Research Questions.......................................................23
II. THEORIES AND FRAMEWORKS FOR HEALTH DECISION MAKING..................24
Individual and Interpersonal Health Behavior Models.....................24
Concept Model for Vaccine Decision Making...............................30
Hypotheses and Aims......................................................32
III. DEVELOPMENT AND EVALUATION OF A MEASUREMENT INSTRUMENT: THE
CONCERNS, ATTITUDES, BELIEFS AND INTENTIONS OF PARENTS ABOUT
VACCINES FOR THEIR CHILD SURVEY..........................................35
Reliability and Validity of Survey Instruments..........................35
Methods for Phase One. Survey Development................................38
Results from Phase One. Survey Development...............................43
Methods for Phase Two. Survey Evaluation.................................47
Results from Phase Two. Survey Evaluation................................52
Methods for Phase Three. Predictive Validity.............................63
Results from Phase Three. Predictive Validity............................66
Discussion...............................................................77
vi


IV. DESCRIPTION OF PREGNANT MOTHERSVACCINE DECISION MAKING PROCESS
OVERTIME..............................................................84
Introduction..........................................................84
Methods for Longitudinal Data Collection..............................86
Results of the Longituinal Cohort.....................................95
Summary of Longitudinal Findings.....................................109
V. COMPARING THREE MEASURES OF VACCINE HESITANCY.....................114
Vaccine Hesitancy....................................................114
Methods..............................................................116
Results..............................................................121
Discussion...........................................................128
VI. CONCLUSIONS......................................................131
REFERENCES...........................................................135
APPENDIX
A....................................................................151
B....................................................................153
C....................................................................155
D....................................................................161
Vll


LIST OF TABLES
TABLE
2.1. Constructs, Definitions, and Application Examples of the Health Belief Model. 27
2.2. Constructs and Definitions of the Theory of Planned Behavior, 2008.......28
2.3. Concepts and Definitions of Social Network Theory........................29
2.4. Theoretical Constructs, Indirect Measures, Sub-constructs, Definitions and
Examples..................................................................32
3.1. Terms and Definitions Related to Reliability and Validity of Survey Instruments
..........................................................................35
3.2. Methods Used to Determine Reliability and Validity........................39
3.3. Expert Panelist Expertise and Institutional Affiliation...................41
3.4. Survey items (n=23), Response Scale, Theoretical Construct, and Source of
Survey Item...............................................................45
3.5. Internal Consistency of Survey Items within Constructs....................55
3.6. Survey Items Removed after Assessment of Internal Consistency.............55
3.7. Survey Items Removed after Evaluation of Factor Structure.................59
3.8. Factor Loadings From Principal Axis Factor Analsis With Direct Oblim
Orthogonal Rotation For A Four Factor Solution For Baseline Survey Items
(N=320)...............................................................60
3.9. Internal Consistency of Survey Items within Constructs....................61
3.10. Results from Baseline and Second Survey Administered Two Weeks Later....62
3.11. Characteristics of Pregnant Women and Parent Cohort by Vaccine Behavior,
N=239.....................................................................68
3.12. Univariate Analysis of Categorical Covariates from Pregnant Mother and
Parent Cohort, N=239......................................................70
3.13. Logistic Regression Analysis of Independent Modeling of Survey Constructs
and Factors Associated with Undervaccination with Adjusted Odds Ratios...72
3.14. Characteristics of Validation Study Cohort by Vaccine Behavior, N=374...73
viii


3.15. Univariate Analysis of Categorical Covariates from Validation Study Cohort,
N=374....................................................................75
3.16. Logistic Regression Analysis Survey Constructs and Factors Associated with
Undervaccination with Adjusted Odds Ratios, N=374........................76
4.1. Characteristics of the Cohort, N=120.....................................98
4.2. Means and Standard Deviations of Pilot Cohort, N=120....................102
4.3. Trends overtime of Pilot Cohort Using Cochran-Armitage, N=120...........108
4.4. Trends overtime of Pilot Cohort Using Cochran-Armitage, N=120...........109
5.1. Characteristics of the Pilot Cohort Using PAC-V Screener Grouping Variable,
N=120....................................................................122
5.2. Characteristics of the Pilot Cohort Using Self-Reported Vaccine Intention
Grouping Variable, N=120.................................................122
5.3. Characteristics of the Cohort Using Average Days Undervaccinated Grouping
Variable, N=80...........................................................123
5.4. Measures of Vaccine Hesitancy by Survey Items, N=120....................124
5.5. Measures of Vaccine Hesitancy by Survey Items, N=239....................127
IX


LIST OF FIGURES
FIGURE
2.1. Conceptual Model of Factors Contributing to Vaccine Decision Making.31
2.2. Phases and Steps to Evaluating Validity and Reliability of Survey Instrument. 34
3.1. Study Flow Diagram for Pilot Cohort.................................53
3.2. Study Flow Diagram for Primary Cohort...............................58
3.3. Study Flow Diagram of Primary Cohort Linking Mother to Child........53
4.1. Design of Prospective Survey Data Collection, N=120.................87
4.2. Pregnancy Cohort Flow Diagram.......................................97
x


CHAPTER I
REVIEW OF THE LITERATURE
Introduction to the Problem
Vaccines are one of the most significant public health achievements in the
past century (Centers for Disease Control and Prevention (CDC), 1999; 2011). As a
result, routine childhood vaccination has led to the eradication, elimination or
control of common infectious diseases such as small pox and polio (Plotkin &
Plotkin, 2004). In the United States, mandates requiring childhood vaccines prior to
entering public school has contributed to the control of infectious disease, and
therefore decreased morbidity and mortality (Roush & Murphy, 2007). Despite the
success of vaccines and extensive research demonstrating the safety of vaccines
(Baggs, et al., 2011; McNeil, et al., 2014), parents have numerous concerns. These
concerns include the ingredients in vaccines, the number and timing of vaccines, and
short-and-long-term health complications resulting from vaccination (Freed, Clark,
Butchart, Singer, & Davis, 2010).
Although most parents follow the recommended childhood vaccine schedule
when vaccinating their children, concerns about vaccines have generated caution
and hesitancy in parents. As a result, this hesitancy has caused some parents to
delay or decline certain or all vaccines for their child (Freed, et al., 2010; Dempsey,
Schaffer, Singer, Butchart, Davis, & Freed, 2011). There has been a recent surge of
interest and activity in the research literature and mainstream media surrounding
parents decision making about childhood vaccines. These decision making factors
1


include interpretation of scientific evidence about the safety and effectiveness of
vaccines, the role of trust, influence of vaccine policies and mandates, influence of
media, social groups, cultural norms and belief systems (Brown, et al., 2012;
Brunson, 2013; Glanz, et ah, 2013b; MacDonald, Smith, & Appleton, 2012).
The publics trust in vaccine governance is weakening (Feudtner & Marcus,
2000; Larson, Cooper, Eskola, Katz, & Ratzan, 2011). In part, this is attributable to
the very success of vaccine policies and programs designed to protect the publics
safety from infectious disease. Due to the success of vaccines, parents are unlikely to
experience vaccine-preventable diseases in their lifetime, and the perceived risk and
seriousness of these diseases diminish (Chen & Hibbs, 1998; Kennedy, Brown, &
Gust, 2005). Subsequently, public concerns have changed from spread of disease to
the safety of vaccines. As vaccine coverage declines in communities, the risk for
infectious disease outbreaks increases (Fine, 1993). There have been recent
outbreaks in the United States that have contributed to the rapid spread of disease
(Halsey & Salmon, 2015). Consequently, infants who cannot vaccinate due to age,
children who cannot receive vaccination due to medical contraindications, and older
adults whose immunity to disease has waned are at-risk for contracting a vaccine-
preventable disease.
Interventions are needed to facilitate change in health beliefs, attitudes, and
concerns, and ultimately increase vaccination rates. These interventions can be
delivered through web-based resources and decision aids, parent-physician
communication approaches, in-person educational outreach, or other traditional
behavioral intervention techniques. However, such interventions need to be
2


rigorously evaluated for cost-effectiveness and the ability to induce behavior
change. In order to measure change in attitudes and beliefs towards vaccines, well-
developed, theory-driven survey measures that can be implemented longitudinally
are needed (Kaufman, et al., 2013; Sadaf, Richards, Glanz, Salmon, & Omer, 2013).
Purpose of the Dissertation
The purpose of this dissertation was to develop, implement and evaluate a
survey instrument to measure parents concerns, attitudes, beliefs, and intentions
about vaccines for their child. Using Kaiser Permanente Colorados (KPCO) robust
electronic health record (EHR) and self-reported information from KPCO members,
an investigation was conducted that expanded on survey measurements in several
ways. First, the instrument was carefully constructed using theory, assessment from
experts, and input from the target population to guide its development. Second, the
survey was designed to measure change in attitudes and beliefs about vaccines over
time. Third, it was designed to be used in multiple interventions rather than for use
in a particular study. Finally, the survey underwent systematic reliability and
validity evaluation.
This chapter provides background information on vaccine policy, vaccine
hesitancy, decision-making, risk perception and sources of influence and
informationin the context of vaccines. Next, interventions for vaccine concerns
and hesitancy are described. Finally, there is a brief discussion about why
prospective measures of concerns, attitudes, beliefs and intentions about childhood
vaccines are needed. The research question is discussed at the end of the chapter.
3


Background on Vaccine Decision Making
There have been increasing vaccination rates overall in the United States.
Conversely, there has been increasing exemption rates from vaccination (Omer, et
al., 2006; Thompson, etal., 2007). An emerging public health concern is the
deviation away from the recommended childhood vaccine schedule. Parents are
increasing their use of delayed or alternative vaccination schedules, where parents
selectively decide when and which vaccines will be administered to their child. It is
estimated that more than one in ten parents of young children use delayed vaccine
schedules (Dempsey, et al., 2011). These schedules are varied and do not follow
consistent patterns (Glanz, etal., 2013a), which poses challenges for pediatricians as
they communicate the safety of the childhood vaccine schedule to parents
(Maglione, et al., 2014). These challenges include lack of scientific research on the
safety of alternative schedules, concerns about susceptibility to disease due to
delaying or omitting vaccines, and unknown outcomes when deviations from the
recommended schedule occur.
For parents, the process of decision making about vaccination for their child
has several important factors. These factors include perceptions of risk of
vaccination and risk of infectious disease, available vaccine information, and
influence of sources of vaccine information such as family, internet, and media that
act as personal networks. While it is well known that parents have concerns about
the risks of vaccination, little is known of the prospective decision making process of
pregnant women about vaccines for their child. This prospective decision-making
process is important to the timing of when to implement practice-based
4


interventions that strive to improve vaccination rates. By describing the decision
making process before and after the birth of the child, information can be obtained
to assist in optimal planning, timing, and information content in future research and
practice.
Role of Public Health Policy in Vaccination
The health of the nation is essential to the social, economic and
organizational growth of its people (Shi & Singh, 2011). Social determinants of
health such as socioeconomic status, access to preventive health care (i.e. vaccines),
and availability to nutritional foods shape the health outcomes of individuals and
communities (Marmot & Wilkinson, 2005).
Public health policy strives to achieve improvement of individual, group, and
population health status. Examples of public health policies include tobacco control
policies, school nutrition policies, drug and alcohol laws, and vaccine mandates.
Implementation of health policies can be complex, time consuming, and have
variable outcomes (Longest, 2002). Stakeholder buy-in and ongoing support are
necessary to the success of policy processes. Additionally, health policies that are
flexible enough to accommodate shifts in societal perceptions about health issues
are more successful in achieving long-term sustainability (Pluye & Denis, 2004).
Vaccine Policy and Protection of Community Health
Vaccine policy is designed to mitigate disparities in health (Shonkoff, Boyce,
& McEwen, 2009). This policy design strives to assure population-based protection
against infectious disease in individuals and communities, regardless of social
5


factors that may impede individual health outcomes. However, vaccine policy does
not guarantee coverage against disease for all individuals and communities.
Defining the problem within the context of the policy agenda is essential to
understanding the causes of vaccine delay, how severe the problem is, consequences
associated with delaying vaccines, and the incidence or scope of the problem
(Rochefort & Cobb, 1994). Decision making about vaccination happens at the
individual level. Individuals can choose to comply with vaccine mandates that
provide protection to the individual and the larger community or opt out from
vaccines, thus depending on others to protect them. Vaccine mandates were
designed to ensure high coverage against vaccine-preventable diseases in the
population (Cooper, Larson & Katz, 2008).
History of Vaccine Mandates
The origins of the current vaccine mandates in the United States are
patterned from the laws related to smallpox disease control. Smallpox is an
infectious disease that results in fluid filled pustules on the skin. The disease can
have severe complications such as scarring, blindness and encephalitis. The disease
also had a 35% mortality rate (Fenner, Henderson, Arita, Jezek, & Ladnyi, 1988).
Initially, vaccination against smallpox was required when the disease was
widespread.
In the United Kingdom in the mid-nineteenth century, smallpox vaccination
of infants was publically mandated, with a penalty of imprisonment of the parent if
this mandate was not followed. This resulted in a change in the relationship
between the state and its citizens. Public mandates for smallpox vaccination caused
6


protests against the states authority over individual citizens decisions about
vaccination. Eventually, law was enacted that allowed for exemption from
vaccination through a conscientious objection statement (Durbach, 2000; Wolfe &
Sharp, 2002).
Shift from Federal to State Authority in the United States
In the nineteenth century, compulsory vaccination was generally accepted.
However, some believed the requirement interfered with human autonomy and
liberty, invited unwarranted governmental interference, infringed on personal and
religious beliefs and induced medical safety concerns (Hodge & Gostin, 2002).
Congress passed the Vaccine Act of 1813 to encourage vaccination against smallpox.
It was the first federal program designed to improve the health of the public. The
goals of the program were to ensure a reliable source of smallpox vaccination,
authorize distribution to citizens, and distribute the vaccine. With this ambitious
public policy in place, the struggle to implement the program proved to be daunting.
After an incident of contaminated vaccine supply, the act was repealed and vaccine
policy shifted from federal to states authority (Griffin, 2009).
Compulsory School Vaccine Mandates
Beginning in 1818, compulsory childhood vaccination was introduced in
England. Noncompliance with vaccination included annual financial penalties and
not permitting an unvaccinated child entry into public school (Hodge & Gostin,
2002). In the United States, the first school-entry vaccine mandate was enacted in
1827. However, these mandates were not upheld until the smallpox epidemics that
occurred in the mid 1890s (Colgrove, 2006; Hodge & Gostin, 2002).
7


Individual Autonomy versus Societal Obligations
Vaccine policy represents a study in challenging tensions between societal
interest and individual interest, much like Hardins "The Tragedy of the Commons
(Hardin, 1968; 1998). The "commons in this policy example is a community with
high vaccination rates, resulting in reduction in risk from infectious disease and a
community relatively free from vaccine preventable disease. Herd immunity (when
enough people are vaccinated in the community to stop transmission of the disease)
of the community also transfers to those who do not vaccinate due to medical,
religious or other reasons. Other reasons for not vaccinating include avoidance of
the risk of adverse events from vaccines or the assertion of individual choice. In
effect, the individual who does not vaccinate "free rides on the risk of vaccinating
taken by others in the community. Those who do not vaccinate are afforded the
same herd protection from disease as those who vaccinate. However, there are
community level immunity threshold levels to maintain in order to prevent
significant outbreaks of infectious disease (Offit, 2011).
At the individual level, the decision not to vaccinate increases the risk of
infectious disease minimally due to community immunity protection provided to the
individual. However, this individual decision not to vaccinate weakens the overall
herd protection for the entire community. If too many individuals choose to do what
is in their best individual interest, the "common is at risk of depletion and will fail
to protect the community. Thus, infectious disease outbreaks will emerge. A
decrease in community protection also puts vulnerable individualswho cannot
receive vaccines and were previously protected by the community protectionat
8


greater risk for infectious diseases (Malone & Hinman, 2003). This is a "tragedy of
the commons where herd immunity is depleted when individual interests prevail in
order to avoid risk of adverse events following vaccination or to assert individual
choice.
To avoid a tragedy of the commons, legal mandates have been imposed at the
state level for certain vaccines (Orenstein & Hinman, 1999). There is a risk of an
"unmanaged commons and potentially social disaster (Hardin, 1998) as more
states allow for exemptions from vaccine requirement mandates. If this scenario
occurs, the common is used up by self-interested parents, or more appropriately
described as concerned parents. As Hardin pointed out in his Tragedy of the
Commons revision manuscript in 1998:
Individualism is cherished because it produces freedom, but the gift is
conditional: The more the population exceeds the carrying capacity of the
environment, the more freedoms must be given up (p. 683).
Dietz and colleagues (2003) propose a governance framework for managing
the commons in modern society. This framework stresses the importance of critical
and ongoing informed dialogue among interested parties. Additionally, the
framework encourages engagement as a way to deal with conflicts, discussion of
social norms and compliance, and adaptation and change to maximize the commons.
Although this governance framework has valuable approaches to dealing with the
9


tragedy of the commons, it is much more difficult to put into practice with a highly
charged topic such as vaccines.
Vaccination of children by their parents has been likened to a social contract
(Hodge & Gostin, 2002; Rousseau, 1920), where individual decision making about
vaccines benefits the social order of the community. The act of vaccination has
duality of benefits as it protects the individual child against infectious disease and
contributes to the protection of the publics health from infectious disease
outbreaks. Vaccination by most individuals in the community protect the medically
vulnerable who cannot vaccinate, those who do not vaccinate by choice, and those
who are unaware they are unprotectedwhen the vaccine is not effective on the
individual level. However, at the individual level, parents are concerned about the
safety of vaccines and perceived risks associated with vaccination (Freed, et al.,
2010; Kennedy, LaVail, Nowak, Basket, & Landry, 2011). These tensions frame the
challenges of implementation and sustainability of vaccine policy today.
Current Vaccine Mandates
In the United States, death from common childhood infectious disease has
been essentially eliminated, including polio and rubella (Roush & Murphy, 2007).
For example, there has been greater than 98% decline in the incidence of vaccine
preventable diseases such as polio, mumps, Haemophilus influenzae type b, and
measles since vaccines for these diseases became available (Orenstein, Douglas,
Rodewald, & Hinman, 2005; Roush & Murphy, 2007).
This has been successful, in part, due to the implementation of vaccine policy
in the United States (Orenstein & Hinman, 1999; Orenstein, et al., 2005; Pickering &
10


Orenstein, 2002). In 1977, the American Academy of Pediatrics (AAP) made vaccine
policy a top priority, calling for universal vaccination (American Academy of
Pediatrics (AAP), 1977; Wood, 2003). AAP strongly supported enactment of public
policy mandates to require up-to-date vaccinations of all children entering public
school (National Vaccine Advisory Committee (NVAC), 1999; NVAC, 2013).
Exemption from Vaccines
States determine the requirements for school mandated vaccination. These
requirements include the documentation necessary to enter public school, the
documentation required to show evidence of vaccination, and conditions under
which parents can exempt their children from vaccination (Diekema, 2014). There
are three main categories of exemption from vaccination: medical exemption,
religious exemption and personal or philosophical exemption. Not all states allow
religious or personal exemptions. Currently, three states (California, Mississippi,
and West Virginia) do not recognize religious or personal exemptions. Nineteen
states permit personal and religious exemptions. All states permit medical
exemption from vaccination (Sandstrom, accessed 9-22-2015). States rely on the
recommended immunization schedule determined by the Advisory Committee on
Immunization Practices (ACIP) to determine individual state vaccination
requirements. ACIP is a group of fifteen voting medical and public health experts
who determine the recommended doses and use of vaccines to control infectious
disease in the United States (Smith, 2010).
Mandates that permit personal exemption (also known as philosophical
exemption) from school-required vaccination vary in complexity from state to state.
11


However, the general format includes parents acknowledgement of a personal
objection to vaccination (Rota, Salmon, Rodewald, Chen, Hibbs, & Gangarosa, 2001;
Blank, Caplan, & Constable, 2013).
There has been an overall increase in state-level rates of personal
exemptions. Ease of exemption from school-required vaccination has been shown to
influence rates of undervaccination. Personal exemptions rose from an average of
0.99 percent in 1999 to 2.54 percent in 2004 in states that have this policy option
(Omer, et al., 2006; Thompson, et al., 2007). For example, in the state of Arkansas,
the number of personal exemptions increased from 529 in 2001 to 1,145 in 2004,
following a vaccine policy change that permitted personal exemption from required
school entry vaccination (Salmon, et al., 2006; Thompson, et al., 2007). Omer, et al.
(2012) recently updated their landmark research in vaccine policies and rates of
exemption from vaccines. They found that in 2011, nonmedical exemption rates in
states that had relatively easy exemption policies were higher than in states with
difficult exemption policies.
Geographic Location and Vaccine Exemptions Rates
Geographic clustering of exemptions from vaccines occurs within states,
which may reach upwards of one-quarter to one-third of the community. In the state
of Washington in 2007, county-level vaccine exemption rates ranged from 1.2 to
26.9% (Omer, et al., 2008).
There are wide variations in exemption from vaccination rates across states.
In the 2014-2015 school year, Louisiana had less than 0.6% personal exemption
rate from vaccination among kindergartners while Vermont had an exemption rate
12


of 5.9% (Seither, et al., 2015). Furthermore, these rates are based on ages of
children at least five years old, and does not account for rates of undervaccination of
children younger than five years of age. These children may have been on
alternative vaccination schedules and have subsequently "caught up with the age
appropriate recommended childhood vaccine schedule.
There is growing concern among public health officials, policymakers, school
officials, parent groups, and medical providers that the upward trend in personal
exemptions from vaccination may contribute to the re-emergence and rise of
otherwise population-controlled infectious diseases (Chen & Hibbs, 1998; Lantos, et
al., 2010). In fact, there have been recent measles and pertussis outbreaks linked to
children of parents who did not vaccinate their child (Atwell, et al., 2013; Halsey &
Salmon, 2015).
Vaccine Hesitancy
The International Vaccine Hesitancy Working Group defines vaccine
hesitancy as "delay in acceptance or refusal of vaccines despite availability of
vaccine services. Vaccine hesitancy is complex and context specific, varying across
time, place and vaccines. It is influenced by factors such as complacency,
convenience and confidence (WHO SAGE Working Group, 2014, p. 7).
Vaccine Refusal and Delay
Some studies have found that parents who choose not to vaccinate their
children differ demographically from parents who vaccinate their children. They
tend to be older and have higher levels of education (Gust, et al., 2008). Conversely,
the national rates of vaccination for children under the age of three are lower for
13


those living in poverty. Vaccination rates for most series are lower for children who
are black (Elam-Evans, et al., 2014). It is important to clarify that there are distinct
differences between parents who chose not to vaccinate their child and those who
experience logistic barriers such as transportation. This current investigation
focuses on undervaccination by parental choice.
Studies using individual-level data have shown that children of parents who
declined vaccination were approximately twenty-three times more likely to acquire
pertussis (Glanz, et al., 2009) and nine times more likely to contract varicella than
vaccinated children (Glanz, et al., 2010). When compared to parents who accept
vaccines, parents who decline vaccines are more likely to believe their children are
not at risk for vaccine-preventable diseases and that vaccine-preventable diseases
are not dangerous (Salmon, et al., 2005; Smith, et al., 2011).
One in ten parents delay one or more vaccines for their child (Dempsey, et al.,
2011; Offit & Moser, 2009). There is sparse scientific information available on the
safety of delaying vaccines. There is some scientific indication that delaying one
childhood vaccine poses greater risks. Hambidge and colleagues (2014) found that
delaying the first dose of measles, mumps and rubella (MMR) vaccine past fifteen
months of age resulted in higher risk of seizures. While this provides some evidence
that deviation from the recommended childhood vaccine schedule poses increased
adverse events to the individual, at least for one vaccine, it also poses an interesting
dilemma. If a parent refuses any MMR vaccine, their child has a decreased risk of
seizure and an increased risk of the disease of measles.
14


Vaccine Concerns and Trust
Although scientific evidence has refuted many of the misconceptions
regarding vaccine safety (see DeStefano, 2007; Thompson, et al., 2007; Price,
Thompson, Goodson, & Weintraub, et al., 2010), there is continuing mistrust in
mainstream media sources as well as misunderstanding of scientific information.
General concerns of parents regarding vaccination include a perceived causal
linkage between receipt of the measles, mumps, and rubella vaccine and
neurological disorder (Smith, et al., 2008); autism (Thompson, et al., 2010; Freed,
2010); parents worries that the human papillomavirus vaccine may lead to sexual
promiscuity (Dempsey, Zimet, Davis, & Koutsky, 2006); fears regarding the
manufacturing processes and ingredients in vaccines (Benin, Wisler-Scher, Colson,
Shapiro, & Holmboe, 2006); and concerns that too many vaccines overload or
weaken young childrens immune systems (Benin, et al., 2006). Vaccine safety
concerns may reduce parents willingness to vaccinate their children. This is
particularly true for new vaccines that are added to the childhood vaccine schedule
(Freed, et al., 2010).
Several studies have examined the role of trust in vaccine decision making.
Benin and colleagues (2006) conducted qualitative, open-ended interviews with
new mothers (one to three days postpartum) and then again in three to six months
time. The theme of trust was an important factor in parents decision-making about
vaccination of their infants. Those who vaccinated their children had more trust in
their physician, wanted to follow social norms to vaccinate, and wanted to protect
their child and others from infectious disease through vaccination. Mothers who did
15


not vaccinate their child verbalized distrust in their physician, had a trusting
relationship with a complementary medicine provider or others who endorsed their
beliefs about not vaccinating their child, were worried about vaccine adverse events,
and believed that vaccine preventable diseases are not serious. The parents also
considered that others vaccinate and therefore their child is not at risk for a vaccine
preventable disease (Benin, etal., 2006).
Other studies have found similar findings in relation to trust and their childs
provider. Parents who refused and delay vaccines are more likely to have low levels
of trust in their childs pediatrician than parents who accepted vaccines (Glanz, et
al., 2013b). Information from health care professions, health departments and
governmental agencies are viewed as less trustworthy by parents who do not
vaccinate their child (Kennedy, et al., 2005; Salmon, et al., 2005). Studies have
shown that trust in their childs pediatrician influence their decision to vaccinate
their child in those who previously deviated or considered deviating from the
recommended childhood vaccine schedule (McCauley, Kennedy, Basket, & Sheedy,
2012). From these published studies, it appears that their childs pediatrician can
mitigate parents vaccine concerns, to some extent.
Decision Making about Vaccines
With the increasing trend of parents who delay some or all vaccines for their
children, research has focused on how parents approach this decision and what
factors are associated with delay (Omer, et al., 2006; Robison, Groom, & Young,
2012; Omer, et al., 2012). These parents are often concerned that children receive
too many vaccines over a short period of time (Salmon, et al., 2005). Consequently,
16


parents request vaccination schedules that both increase the time between
vaccinations and reduce the number of vaccine doses in a single medical visit.
Studies suggest that children with delayed vaccines may have different
healthcare utilization patterns than vaccinated children. Glanz and colleagues
(2013a) found that unvaccinated children of parents who refused vaccines were
50% less likely to have a clinic visit for an upper respiratory infection compared to
fully vaccinated children (Glanz, et al., 2009).
Furthermore, in a matched cohort analysis, it was found that
undervaccinated children had lower outpatient visit rates and higher inpatient
admission rates compared to age-appropriately vaccinated children (Glanz, et al.,
2013a). Other research has shown that children who do not receive some or all
vaccines for personal or medical reasons had 12% fewer clinic visits than children
who were up-to-date (Wei, Mullooly, Goodman, etal., 2009).
Risk Perceptions
A shift occurred from concerns of infectious disease to concerns about
adverse events associated with receipt of vaccines. This was fueled by the
proliferation of media outlets recounting personal, emotional, and affective stories
of parents whose children were reported as harmed by vaccines (See Gonzalez,
1982; Offit, 2011).
Kennedy and colleagues (2011) used the 2010 HealthStyles data which
surveyed parents of children six years of age or younger. Parents endorsed concerns
about the pain of vaccines during a doctors visit most frequently (38%), followed
by concerns that their child received too many vaccines in one doctors visit (36%).
17


Additional concerns endorsed by at least 25% of the parents included the number of
vaccines for children, symptoms of fever after vaccination, concerns that vaccines
may cause autism, and concerns that vaccine ingredients are unsafe.
In a study of 1500 parents of young infants, parents knowledge, attitudes,
and beliefs were stratified by vaccine decision. Parents stated that vaccine side
effects were of highest concern (McCauley, et al., 2012). These findings confirm
previous studies (Freed, etal., 2010). Weinstein, etal. (2007) prospectively
measured beliefs about risk as a predictor to influenza vaccination in college
students. Anticipated regret of getting influenza was the strongest predictor of
getting the influenza vaccine. Additionally, it was found that risk perceptions framed
as feelings predicted the relationship between risk and vaccine better than rational
statements (Weinstein, etal., 2007). Finally, Brewer, etal. (2007) conducted a meta-
analysis of vaccine risk perceptions in thirty-four studies with 15,988 data points,
finding consistent relationships between risk perceptions and behavior. They found
linear relationships in risk likelihood (r=.26), susceptibility (r=.24), and severity
(r=.16) as significantly predicting vaccination behavior (Brewer, et al., 2007).
Research in the area of risk perception and health behaviors is relatively
undecided. A majority of empirical research does find positive associations between
perception of risk and behaviors (See Brewer, Weinstein, Cuite, & Herrington, 2004;
Weinstein, et al., 2007; Wroe, et al., 2005; Wroe, et al., 2004 as examples). Yet,
individual studies report a range of outcomes, relationships and effect sizes.
18


Sources of Influence
Parents access information about vaccines through a wide variety of ways
including internet, books, social media, personal contacts (neighbors, family, and
groups), physician or complimentary health provider and television as some
common examples. If the attitudes towards vaccination are similar between
pediatrician and parent, the pediatrician can potentially contribute to parents
decisions about vaccination (Mergler, et al., 2013). Furthermore, Opel and
colleagues (2013a) found that when physicians use a presumptive narrative that
assumes parents will vaccinate, the compliance with vaccination increased
compared to a participatory narrative that encouraged discussion about the
decision. This new area of research investigates provider influences on parent
decision making.
Often, media sources and personal contacts contradict one another, which
can lead to conflicting and erroneous information. Media can have both positive and
negative effects on parents decision-making about vaccines for their children
(Smith, et ah, 2008; Nyhan, Reifler, Richey, & Freed, 2014).
An increasing number of adults use the internet to obtain health care
information. Recent Pew survey data show that nearly 81% of adults in the United
States are regular internet users, 72% of those regular internet users looked online
for health information in the past year (Fox & Duggan, 2013). Half of online health
and medical information searches are related to someone elses health (Fox &
Duggan, 2013). Although most parents identify physicians as their most trusted
source for health information regarding their child (Freed, et ah, 2010), parents use
19


the internet as a source of information about vaccine over their childs physician to
guide their decision (Benin, 2006; Downs, Bruine de Bruin, & Fischoff, 2008).
Additionally, Downs (2008) and others have found that parents lack basic
knowledge about how vaccines work. They do not find the vaccine information
provided by their childs pediatrician very helpful or to be enough (see also Gellin,
Mailbach, & Marcuse, 2000; Gust, et al., 2005; Benin, et al., 2006). These parents
were less confident in the safety of vaccines and endorsed communication barriers
with their childs pediatrician. These sources of influence and deficits in knowledge
add to the complexity of decision-making.
Interventions for Concerns and Hesitancy about Vaccines
Given the sweeping changes in the ways in which people receive and retrieve
health information (e.g. internet, social media), it is imperative to design
interventions that meet the needs of individuals and provide flexible frameworks
from which evidence-based health strategies can be implemented. A promising
approach to disseminating health information to individuals is the use of interactive
health information technologies. Health technology offers a scalable platform that
can help improve the quality, cost effectiveness, capacity and efficiency in the health
system (Bennett & Glasgow, 2009), and specifically vaccine decision making.
However, interventions using technology need to be rigorously evaluated for
effectiveness (Shoup, etal., 2015).
Several interventions designed to improve vaccination rates have been
conducted. In a study by Wroe, parents who received a decision aid for vaccination
had significant increases in on-time vaccination of their child, decreased perceptions
20


of risk of vaccines, increased perception of risk of vaccine preventable diseases,
reduced anxiety, and increased satisfaction ten weeks after the birth of their child as
compared to those receiving usual care (Wroe, et al., 2005).
In another study, researchers implemented a web-based MMR decision aid
that significantly increased parents knowledge over time and decreased conflict
over the MMR vaccine decision. Most parents decided to vaccinate their child for
MMR, however the sample size was small (Jackson, Cheater, Peacock, Leask, &
Trevena, 2010). From the same study team, a randomized control trial of 142 study
participants where the intervention arm was an informational pamphlet on MMR
and a parent-led group meeting, and the control group was pamphlet only. Those in
the intervention arm had higher vaccination rates for MMR; however, decisional
conflict remained the same (Jackson, et al., 2011).
There is a need to know when the optimal timing is to implement
interventions aimed to increase vaccines and reduce vaccine-related concerns.
There is sparse evidence as to how and when to intervene (Sadaf, et al., 2013).
Further knowledge in the area of decision making and risk perception regarding
vaccination has strong implications for public policy, as vast amounts of time and
resources are invested in our public infrastructure in planning for disease
pandemics (Miller, Viboud, Blalinska, & Simonsen, 2009). Clearly, there is a need for
interventions designed for parents to increase vaccination rates and decrease
vaccine concerns. However, there is a parallel need for well-designed, theory-driven
and reliable measurement instruments to assess the subjective outcomes of
interventions.
21


Need for Prospective Measures
Although there are many survey instruments in the literature that measure
vaccine concerns, most have not been developed for longitudinal use or evaluated
for association with vaccine behavior. Opel (2011a'b; 2013b) has designed a
validated survey that is a vaccine hesitancy screener administered to parents in the
clinical setting (Opel, et al., 2013b). It measures parents intentions to vaccinate and
quantifies a score of hesitancy about vaccinating. This research holds promise in
assisting medical personnel to tailor information and messages in near real-time to
parents who are concerned about vaccines or are hesitant about vaccinating their
child. It also provides a measure to predict future vaccine decisions and potentially
mediate the potential outcome of vaccine hesitancy through information, education,
and emotional element of vaccine information and decision making that parents
experience.
In summary, there are three major gaps in the literature when considering
the measurement of parent-associated vaccine concerns and behaviors: validated
measures specific to vaccine concerns, attitudes, behaviors and intentions;
measurement of these domains prospectively and measure change over time in
these domains; and use of objective measures to quantify the association with the
survey measures.
The literature reviewed in this chapter included vaccine mandates, rates of
exemption from mandates, and parents hesitancy about vaccines for their child.
Specifically, the chapter identified factors that influence parents decision-making
about vaccines. First, parents have numerous concerns about the safety of vaccines.
22


These include concerns that vaccines cause negative short-and-longterm health
conditions such as autism. Second, sources of information that parents access about
vaccines are influential. However, these sources differ for parents who have
concerns about vaccines. Physicians are the most influential sources for parents
who decide to vaccinate their child, but to a lesser extent for parents who decide to
delay or not vaccinate their child. Finally, perceptions of risk of disease and risk of
vaccination were reviewed. Vaccine behavior is associated with perceived risk of
getting a disease or experiencing an adverse event from vaccination. Vaccine
behavior is also associated with how severe the disease or adverse event might be.
Next, the investigation research question is presented.
Research Questions
The research question focuses the dissertation on the development and
evaluation of a survey instrument to measure parents changes in concerns,
attitudes, beliefs and intentions about vaccines for their child.
Research Question 1: How can changes in attitudes and beliefs about childhood
vaccines be measured?
Research Question 2: What vaccine decision-making factors, as measured
subjectively (by survey items) in a cohort of parents, are associated with
undervaccination behavior?
Next, in chapter two, the theories that guided the design and the
development of a survey instrument that measures the factors identified in the
literature review are presented. For this investigation, the three major theories
23


applied were Health Belief Model (HBM), Theory of Planned Behavior (TPB) and
Social Network Theory (SNT). HBM and TPB guided the development of the survey
constructs, while SNT guided the addition of descriptive questions.
24


CHAPTER II
THEORIES AND FRAMEWORKS FOR HEALTH DECISION MAKING
Traditional theories of decision-making have focused on the rational actor,
bounded rationality, public choice analysis, deliberative democracy and discourse
theory, organizational influence, and policy networks, to name a few (Simon, 1959;
Kingdon & Thurber, 1984; Koppenjan & Kliji, 2004; March, 1994; Miller, 1982).
These theories range in unit of analysis from the individual actor to broad social
groups. More recently, theories and models of health behavior have emerged,
drawing upon several disciplines such as sociology, psychology, epidemiology and
biological sciences. These theories help researchers guide the examination of
reasons why people engage in particular health behaviors or not, assist in planning
and development of public health programs, and contribute strategies to health
communication planning. Most significantly, health behavior theories have focused
on effective health education and health communication strategies that lead to
recommended behavior change (Glanz, Rimer, & Viswanath, 2008).
Policy makers, health care institutions, parents, and schools have identified
vaccine decision-making as an important public health issue over the last decade. In
order for vaccine behavior change to occur in an effective manner, methods should
be designed with emphasis on the individual and their social characteristics, beliefs,
norms and their environment.
Individual and Interpersonal Health Behavior Models
The current study uses several theoretical perspectives to create a
conceptual model. The model integrates individual and interpersonal factors that
25


are theorized to contribute to parents decision making regarding vaccines for their
child. Below, the three health behavior models used for this investigation are
described.
Health Belief Model (HBM)
The Health Belief Model (HBM) is one of the most widely applied health
behavior conceptual framework used to explain both behavior change and
maintenance of health related behaviors, and to guide health behavior interventions
(Champion & Skinner, 2008). It was developed in the 1950s by social psychologists
in the United States Public Health Service to explain poor participation by the public
in health prevention programs. Rosenstock and colleagues (1959) refined the
factors by conducting a systematic review of factors on why people did or did not
accept polio vaccination.
The main factors of the HBM emerged through this research (Rosenstock,
Derryberry, & Carriger, 1959). The model was expanded upon to include the study
of symptoms and behaviors related to illness (Conner & Norman, 2005). HBMs
central focus is health motivation, which makes it amenable to addressing behaviors
that induce health concerns. HBM assumes that the individual can make their own
decisions and to act upon those decisions. The key aspect of the model is that an
individual will take a health-related action (such as vaccination) if there is the belief
and expectation that a negative health condition can be avoided and the
recommended health action can be acted upon (Champion & Skinner, 2008).
Six constructs, as described in Table 2.1 contribute to HBM as key elements
in predicting why people act to prevent, screen for, or control illness. These include
26


susceptibility, seriousness, benefits and barriers to a behavior, cues to action, and
self-efficacy.
Table 2.1. Constructs, Definitions, and Application Examples of the Health Belief
Model3.
Construct Definition Application
Perceived susceptibility Belief about the chances of experiencing a risk or getting a condition or disease Define populations at risk Individual risk based on behaviors
Perceived severity Belief about how serious a condition and its sequelae are Specify consequences of risks and conditions
Perceived benefits Belief in efficacy of the advised action to reduce risk or seriousness of impact Define the action to take and positive effects that are expected
Perceived barriers Belief about the tangible and psychological costs of the advised action Identify and reduce perceived barriers through correction of misinformation or incentives
Cues to action Strategies to activate "readiness Provide how-to information, promote awareness; reminders to health behaviors
Self-efficacy Confidence in ones ability to take action Provide training in recommended action, reinforcement, reduce anxiety
aGlanz, et at, 2008
A significant limitation of HBM is its exclusive focus on the individual without
taking into consideration the environmental context in which the individual resides.
Without a broader context of the individual, there is failure to consider other
influences such as social networks. In addition, HBM is a cognitive-based model and
does not consider the affective or emotional components to behavior. Given the
recent advances in understanding the importance of emotion and risk perception
(Slovic, 2010), this represents a gap in the model that should be added.
27


Theory of Planned Behavior (TPB)
The Theory of Planned Behavior (TPB) is another individual-level health
model, which focuses on individual motivating factors that act as determinants to
performing an intended behavior. TPB is an extension of the Theory of Reasoned
Action (TRA), adding perceived control to the models constructs. The underlying
tenet of the theory is its assumption that behavioral intention (motivation)
determines behavior. Therefore, it is also important to determine the extent of
control the person has over the behavior (Montano & Kasprzyk, 2008).
Three main constructs contribute to the TPB as key elements in explaining
how behavioral intention determines behavior. These constructs are attitude
towards behavior, subjective norm, and perceived behavioral control as influencing
behavioral intention. Table 2.2 provides the definitions of the constructs.
Table 2.2 Constructs and Definitions of the Theory of Planned Behavior, 2008a.
Construct Definition
Behavioral intention Perceived likelihood of performing the behavior
Attitude Overall affective evaluation of the behavior
Subjective norm Belief about whether most people approve or disapprove of the behaviorand an individuals motivation to comply with expectations of others
Perceived control Overall measure of perceived control over behaviorlikelihood and effect of a facilitating or constraining condition
aGlanz, et at, 2008
However, there is conflicting views in the literature regarding TPBs
intentions as a predictor of behavior (see Sheeran, Trafimow, Finaly, & Norman,
2002; Webb & Sherran, 2004). Some studies demonstrate a relationship while
others fail to find a relationship.
28


Social Network Theory (SNT)
The Social Network Theory (SNT), an interpersonal conceptual framework,
has been described as a powerful influence on health. The concept of SNT emerged
in the 1950s, through the work of Barnes (1954; 1969). From early work in the
field, it was found that close-knit or homogenous networks exchange more affective,
emotional support and provide more social influence on members to conform to
social norms (Heaney & Israel, 2008). SNT as a conceptual framework illustrates the
influence of social relationships on health status, health behaviors, and health
decision making.
There are three concepts of SNT; each concept has sub-concepts and
definitions (Heaney & Israel, 2008). Table 2.3 provides the concepts and definitions
of SN.
Table 2.3. Concepts and Definitions of Social Network Theory3.
Concepts
Definitions
Structural characteristics of social networks:
Reciprocity
Extent to which resources and support are both given and received in a
relationship
Intensity or strength Extent to which social relationships offer emotional closeness
Complexity
Extent to which social relationships serve many functions
Formality
Extent to which social relationships exist in the context of organizational
or institutional roles
Density
Extent to which network members know and interact with each other
Homogeneity
Extent to which network members are demographically similar
Geographic dispersion Extent to which network members live in close proximity to focal person
Directionality
Extent to which members of the dyad share equal power and influence
Functions of social networks:
Social capital Resources characterized by norms of reciprocity and social trust
29


Definitions
Concepts
Social influence
Social undermining
Companionship
Social support
Process by which thoughts and actions are changed by actions of others
Process by which others express negative affect or criticism or hinder
ones attainment of goals
Sharing leisure of other activities with network members
Aid and assistance exchanged through social relationships and
interpersonal transactions
Types of social support:
Emotional support
Instrumental support
Informational support
Appraisal support
aGlanz, et al., 2008
Expressions of empathy, love, trust and caring
Tangible aid and services
Advice, suggestions and information
Information that is useful for self-evaluation
SNT lacks empirical evidence on how to enhance social networks and
communication exchange within a network. This culminates in the question of "who
should provide what to whom and when (Heaney & Israel, p. 207, 2008). However,
SNT offers both the broader environment and affective components in relation to
behavior changes.
Concept Model for Vaccine Decision Making
The current investigation uses components of three theories (Health Belief
Model, Theory of Planned Behavior, and Social Network Theory) to create a
conceptual model about vaccine decision-making that integrates individual and
social factors.
Figure 2.1 depicts the conceptual model designed for this study, based on the
literature review. For each model construct, Table 2.4 provides a definition and
example. In this model, the linear pathway of the concept model implies that
30


attitudes and beliefs about vaccines lead to formation of intentions about
vaccinating. These intentions influence vaccine behavior (vaccinating or not
vaccinating on time).
In this study, the constructs from HBM and TPB are used to develop and
evaluate a survey instrument. In the concept model, the following conceptual
definitions are used: beliefs about perceived susceptibility and severity contributes
to Beliefs about Vaccinating and beliefs about effectiveness and costs of taking
action contributes to Evaluation of VPD/VAE. For TPB, the following constructs are
used: Subjective Norms about Vaccinating, Perceived Control of Vaccination
Decision, and Intentions about Vaccinating. Social Network Theory concepts are
used for descriptive purposes. These concepts are Social Influence and
Informational Support. The SNT constructs are used for descriptive questions. These
are: Functions of Social Networks and Types of Social Networks.
Latent (Indirect) Measures
Intention
Outcome
SOCIAL/POLITICAL CONTEXT
Vaccine Policy
Societal Norms
INDIVIDUALATTITUDES
Beliefs about Vaccinating
Evaluation of VPD and VAE
Subjective Norms about
_______Vaccinating________
Perceived Control of Vaccinating
Decisions
Intentions about Vaccinating
Vaccinating Behavior
(i.e. vaccination plan for yl Accept
child) Delay
Decline
Figure 2.1. Conceptual Model of Factors Contributing to Vaccine Decision Making
31


Table 2. 4. Theoretical Constructs, Indirect Measures, Sub-constructs, Definitions
Model Construct/ Predictors of Intention Theoretical Model from sub-construct fHBMa/TPBb) Definition
Beliefs about Vaccinating HBM/TPB This measures the parents' perception of effectiveness of vaccines or perception of physical and psychological costs of vaccination.
Evaluation of VPDc/VAEd HBM/TPB This measures parents' beliefs about whether or not their children are at risk for or seriousness ofVPD/VAE
Perceived Control of Vaccinating Decision TPB This measures the parents' belief in their ability to overcome barriers to vaccination and perceptions of difficulty or ease in vaccination.
Subjective Norms about Vaccinating TPB Beliefs or motivation of parent about complying with social norms
aHBM = Health Belief Model
bTPB = Theory of Planned Behavior
CVPD = Vaccine preventable disease
dVAE = Vaccine adverse event
Hypotheses and Aims
To answer the questions, how can changes in attitudes and beliefs about
childhood vaccines be measured and what vaccine decision-making factors, as
measured subjectively (by survey items) in a cohort of parents, are associated with
undervaccination behavior, a study was conducted with pregnant women and
parents of children under the age of twelve months. The hypotheses and aims are
presented below.
Hypotheses
1) Survey constructs developed through this investigation (Beliefs about
Vaccinating, Evaluation ofVPD/VAE, Perceived Control of the Vaccine
Decision, and Subjective Norms about Vaccination) are associated
with objective vaccine behavior (vaccinated or undervaccinated).
32


2) Concerns, attitudes, beliefs and intentions of parents about vaccines
for their child change over time from pregnancy to six months after
birth of the child.
Aims
1) Develop a longitudinal survey instrument that measures parents
concerns, attitudes, beliefs and intentions about vaccines for their
child.
2) Administer the survey instrument on two cohorts of parents.
a. Cohort of pregnant women over time
b. Cohort of pregnant mothers and parents of children under
twelve months of age from an integrated health system KPCO.
3) Evaluate the survey instrument for reliability and validity.
a. Conduct exploratory factor analysis (EFA) to determine the
survey items that measure latent variables and develop factors
b. Calculate Cronbachs alpha on survey items identified in EFA to
evaluate internal consistency of the related group of items
c. Conduct test-retest reliability to measure stability of the
instrument over time
4) Estimate the association between subjective survey constructs and
objective vaccine behavior (vaccinated or undervaccinated).
5) Compare survey variables at four different time points to evaluate
changes over time.
6) Compare different measures of vaccine hesitancy using cohort data.
33


There are three phases to the investigation. First, the methods and results of
the development of the survey instrument are presented in phase one. This is
followed by phase two, the methods and results of the evaluation of the survey
instrument. Finally, in phase three, the methods and results of testing the
association between the survey constructs with undervaccination are presented.
These phases are discussed in Chapter III. Figure 2.2 shows the layout of each of the
subsequent chapters. Chapter III presents phases one, two and three. Chapter IV
presents the longitudinal analyses of each survey item from the pilot survey.
Chapter V presents comparison of different measures of hesitancy.
PHASE 1 Develop the instrument PHASE 2 Evaluate the instrument PHASE 3 Assess association of survey constructs with undervaccination Analyses of survey items over 4 time points Assessment of survey items using subjective intentions and objective undervaccination
o
Steps 1. Literature review (face validity) 2. Expert panel (content validity) 3. Cognitive interviews (face validity) Steps 1. Identify the cohort 2. Administerthe pilot survey 3. Exploratory factor analyses (construct validity) 4. Internal consistency (reliability) 5. Test retest reliability (repeatability Steps 1. Linkparentto child data 2. Calculate days undervaccinated 3. Screen the data 4. Model building Steps 1. Survey items over 4 data points 2. Repeated measures analyses Steps 1. Use subjective followed by objective outcome 2. Assess each survey item using binary outcomes
Figure 2.2. Phases and Steps to Evaluating Validity and Reliability of Survey
Instrument
34


CHAPTER III
DEVELOPMENT AND EVALUATION OF A MEASUREMENT INSTRUMENT: THE
CONCERNS, ATTITUDES, BELIEFS AND INTENTIONS OF PARENTS ABOUT
VACCINES FOR THEIR CHILD SURVEY
In this chapter, the process of developing, evaluating, and validating a theory-
driven survey instrument that can be used to measure the effectiveness of
interventions targeting vaccine concerns of parents is described. First, the process
of developing and piloting the survey instrument is described. Next, the results of
the process are presented. Then, the methods and results from evaluating the
validity and reliability of the instrument on a cohort of parents are described.
Finally, methods and analytic results of how the survey constructs are associated
with vaccine behavior are shown.
Reliability and Validity of Survey Instrument
Comprehensive development of a survey instrument requires assessment of
reliability and validity from piloting through finalization of the instrument. In order
to establish common definitions and provide a clear understanding of the methods
and results, a table of basic definitions was created (Table 3.1).
Table 3.1. Terms and Definitions Related to Reliability and Validity of Survey
Instruments3
Term Definition
Reliability Survey instrument performs in consistent, predictable ways; scores represent a true meaning of the survey item.
Internal consistency reliability Measures the strength of relationship between survey items in a scale. Items should be highly correlated. Cronbachs Alpha is a measure of reliability.
35


Term Definition
Test re-test reliability Test-retest reliability is a measure of consistency and stability of the survey instrument over time. It is assessed by administering the same test to the same sample on two different occasions. There should be no substantial difference in responses between the two time points.
Validity Assesses whether the survey items predict specific events or its relationship to measures of other constructs.
Face validity Assesses whether, on surface, the survey items appear to represent the survey construct.
Content validity Assessed by recognized subject matter experts to evaluate whether survey items define content.
Criterion validity Assesses how well one instrument stacks up against another instrument or predictor.
Predictive validity Assesses the survey instruments ability to predict an outcome based on survey scores
Concurrent validity Assesses the new instrument against a "gold standard
Construct validity Assesses whether the survey items truly reflects and measures the construct15 when it is put into real world practical use.
Convergent validity Assesses different methods for obtaining the same information about a trait or concept and the results are similar
aLitwin, 1995; Fowler, 2013
bA construct is an abstract idea, theme or subject matter measured by survey items. It is also called a
latent variable in some literature (DeVellis, 2012],
Objectives of the Study
The primary objective of this study was to design a theory-driven survey
instrument to measure parents concerns, attitudes, beliefs and intentions about
childhood vaccines. Additional aims include administration of the survey to two
cohorts. The Pilot cohort was pregnant mothers exclusively and the Primary cohort
was pregnant mothers and parents of children under twelve months of age. Factor
analyses and reliability testing were used to evaluate the survey.
Theory Used to Develop the Survey
The survey instrument was developed by primarily using the theoretical
constructs and definitions from Theory of Planned Behavior (TPB) and Health Belief
Model (HBM), two health behavior theories that have been previously used in health
services research to explain vaccine decision making and behavior. Social Network
36


Theory (SNT) was used to guide the development of descriptive survey items that
were not used in determining the survey constructs.
Two constructs used conceptual definitions from the HBM (Champion &
Skinner, 2008) for development of the survey instrument. These constructs were
Beliefs about Vaccinating, and Evaluation of Vaccine Preventable Diseases (VPD)
and Vaccine Adverse Events (VAE). Beliefs about Vaccinating explain the benefits
and barriers of vaccinating. It measures the parents perceived psychological and
physical costs associated with vaccinating their child and parents perceptions of the
effectiveness of vaccination. Evaluation of VPD and VAE explain a parents belief
about how serious a parent considers the disease or potential adverse events after
vaccinating. It measures parents beliefs about how likely it is to contract a disease
or experience a vaccine adverse event (Montano & Kasprzyk, 2008).
An additional two constructs from TPB (Glanz, 2008) were included. These
constructs were Perceived Control of Vaccinating Decision and Subjective Norms
about Vaccinating. Perceived Control of Vaccinating Decision explains the parents
belief in their ability to overcome barriers to vaccination. It measures parents
perceptions of difficulty or ease in vaccination. Subjective Norms about Vaccinating
explains parents beliefs or motivation about vaccines. It measures parents
compliance with social norms about vaccinations.
Together, these four constructs provided the theoretical infrastructure to
develop a measurement instrument to pilot on pregnant mothers about their
concerns, attitudes, beliefs and intentions about vaccines for their child.
37


Methods for Phase One. Survey Development
To address the need for a survey instrument that can be used to measure
changes over time in parents vaccine concerns, an evaluation of reliability and
validity of a novel survey instrument was conducted. There were two cohorts used
in this phase of the investigation. The Pilot cohort consisted of 120 pregnant
women. After the instrument was revised, it was administered to a different cohort,
the Primary cohort of 320 pregnant women and parents of children under the age of
twelve months. A third cohort, the Validation cohort, was added as a validation
check. It consisted of 374 pregnant mothers and parents of children under the age of
twelve months. Below is a description of each cohort that will guide the reader
throughout the dissertation.
Pilot Cohort
This was the first cohort assembled to test the pilot survey. The Pilot cohort
consisted of 120 pregnant women at the time they were recruited. Participants were
administered a baseline survey when pregnant;
Primary Cohort
This was a cohort assembled after the Pilot cohort baseline survey was
analyzed and revisions were made to the survey instrument. It consisted of 320
pregnant women and parents of children less than twelve months of age. The
Primary cohort was used to finalize the survey instrument.
Validation Cohort
This cohort of 374 pregnant mothers and parents of children under the age
of twelve months included the Primary cohort described above and the "no
38


treatment arm of a randomized control trial for behavior change about childhood
vaccines. The purpose of this Validation cohort was to analyze a larger sample of
children in the undervaccinated strata and test the association between the outcome
of undervaccination and the survey constructs.
Table 3.2 shows the types reliability and validity and the methods used to
assess reliability and validity.
Table 3.2. Methods Used to Determine Reliability and Validity of the CABI-V Survey
Type of Reliability or Validity Method Used
Reliability
Internal consistency reliability Cronbachs Alpha
Test re-test reliability Pearson r
Validity
Face validity Literature review; review by investigator and study team; cognitive interviews with target population
Content validity Expert panel
Construct validity Exploratory and Confirmatory Factor Analysis
Predictive validity Logistic Regression
Criterion validity Multiple measures of hesitancy
Procedures
All regulatory compliance (Research Review Committee approval,
Institutional Review Board (IRB] approval] were obtained for the study conducted
at Kaiser Permanente Colorado, Institute for Health Research. The University of
Colorado Denver, Colorado Multiple Institutional Review Board approval (COMIRB)
ceded regulatory oversight to Kaiser Permanente Colorado IRB. All data was stored
at Kaiser Permanente Colorado Institute for Health Research, within a secured,
locked facility. Data was stored in a password protected Access database. The
limited dataset contained a unique study identification number (ID] which was used
to link to EHR data.
39


Using a theory-driven approach, a twenty-three item survey instrument was
developed to measure participants beliefs, concerns, intentions, control, and
subjective norms about childhood vaccines. The first phase established face and
content validity. This involved the following steps: conduct a literature review,
assemble an expert panel to review the survey items, and conduct cognitive
interviews with the target population. The survey instrument comprised previously
established items that were modified as well as novel items. The steps are described
below.
The following three steps were conducted to establish face and content
validity when developing the measurement instrument.
Step One: Literature Review
First, a broad-based literature review was conducted. Survey items that have
been previously published were categorized by year, author, theoretical model and
behavioral constructs and entered into a database. This item inventory provided
easy organization of the existing survey items, linked the theoretical constructs to
survey items, and revealed gaps in measurement of theoretical constructs that
required creation of novel survey items.
Step Two: Expert Panel
Next, a panel of six national subject matter experts (SME) was convened. The
SMEs had expertise in the areas of vaccine policy, research, epidemiology, and
clinical practice. Their areas of profession included pediatricians, behavioral
scientists, public health epidemiologists, and policy academicians (Table 3.3). The
SMEs received a link to an online rating system and received no incentive for their
40


expert participation. The SMEs independently rated each survey item on its strength
of contribution to the identified theoretical construct using a five-point scale
(strongly unfavorable to the concept to strongly favorable to the concept). SMEs also
had the opportunity to add qualitative comments about the wording, structure and
measure of the survey item. Mean scores from the six SMEs for each survey item
were calculated. Those items scoring less than three out of a five-point scale were
removed from the eligible pool of survey items. At the conclusion of these steps, a
pilot survey instrument was developed and formatted.
Table 3.3. Expert Panelist Expertise and Institutional Affiliation
Expert Panelist Area of Expertise Institutional Affiliation
Douglas Opel, MD Vaccine safely, provider communication strategies, organizational issues underlying ethical conflict ScholarTreuman Katz Center for Pediatric Bioethics and Research Institute at Seattle Childrens Hospital; Professor in the Division of Bioethics, Department of Pediatrics at the University of Washington School of Medicine
Saad Omer, PhD Vaccine safely, infectious disease epidemiology, policy analysis and pandemic disease InvestigatorAtlanta Clinical and Translational Science Institute; Emory Center for AIDS Research; Associate ProfessorEmory Vaccine Center; Global Health and Epidemiology, Rollins School of Public Health, Emory University
Hank Jenkins-Smith, PhD Political science, social research, risk perception and vaccines ProfessorCenter for Applied Social Research and Department of Political Science, University of Oklahoma
Julie Downs, PhD Social influences and decision making, vaccine decision making DirectorCenter for Risk Perception and Communication and Center for Behavioral Decision Research in the Department of Social and Decision Sciences, Carnegie Mellon University
Simon Hambidge, MD, PhD Vaccine safety and delivery, immunology, surveillance, provider communication Chief Ambulatory OfficerDenver Health; ProfessorColorado School of Public Health, University of Colorado DenverHealth
systems, management and policy and
Department of Pediatrics, School of Medicine,
University of Colorado Denver; Researcher
Institute for Health Research, Kaiser
Permanente Colorado
41


Expert Panelist Area of Expertise Institutional Affiliation
Matthew Daley, MD Vaccine safety and delivery, survey design, provider communication Senior ResearcherInstitute for Health Research, Kaiser Permanente Colorado; ProfessorDepartment of Pediatrics, School of Medicine, University of Colorado Denver
Step Three: Cognitive Interviews
Then, the survey instrument was piloted on a sample of ten pregnant women
and parents of children under the age of twelve months, who were identified using
the EHR. Parents participated in-person for a one-to-one thirty-minute interview.
Parents provided written consent to participate in the study and were given a ten-
dollar gift card to a retail store for their participation.
Cognitive interviews were conducted to evaluate potential sources of
response error in survey instruments. A protocol for administration of the interview
was developed so that each session was conducted in a similar way. Techniques
were used to assess participants understanding of the intent, comprehension, and
usability of the survey instrument (Willis, 2004). These techniques asked the
participant to "think out loud when answering the survey questions. The
interviewer took careful notes as to what processes were taken to arrive at the
answer. This was followed by verbal probing, where participants were asked probes
such as "What does this mean to you? and "Tell me what this question is asking
using your own words. Based on parents feedback, survey items were
subsequently modified to improve clarity and readability. The qualitative data was
organized by survey question and analyzed for themes (Patton, 2005). Modifications
to the survey were made after the qualitative and quantitative expert panel and
42


parent cognitive interview feedback was analyzed. These steps resulted in a
finalized pilot survey.
Results from Phase One. Survey Development
Results from Step One. Literature Review, Expert Panel, and Cognitive
Interviews
The results from the development of the survey instrument are described
below. These results from the literature review, expert panel and cognitive
interviews establish face and content validity of the survey instrument.
The initial survey item pool generated from the literature review was quite
large. See Appendices A, C and D for detailed survey development steps and survey
items used at each study phase. There were over 150 potential survey items
contributing to the constructs, not including the descriptive and demographic
questions. The investigator eliminated one hundred and nine survey items because
they were duplicate items or irrelevant to the theoretical constructs. There were
now forty-three survey items. Through extensive literature review followed by
assessment by the investigator and collaborative study team, the item pool was
reduced. Using this method, the survey item pool was reduced to thirty items
associated with the constructs in addition to other descriptive and demographic
questions.
The item pool was further reduced to twenty-three items after the panel of
six subject matter experts rated the questions and provided qualitative feedback.
For survey items with a mean rating of three or under, the questions were
disqualified from the item pool. The qualitative feedback from the SMEs had two
43


main themes: reorganization of survey items under construct headings and provide
additional description and clarification by using qualifiers within the survey items
to improve comprehension of the question. Qualitative feedback from the SMEs
included the following quotes:
Some questions are perfectly good questions, but don't fit perfectly
well into any exact construct. That is OK
As you can tell by my comments, sometimes it may make more
sense to rethink what the construct is, and then go back to the
questions (to see if they capture the construct).
Great work; this is a really excellent integration of a number of
important sources and concepts.
Additional examples of qualitative feedback included suggestions to improve
the content, context and understandability of the question. Several SMEs suggested
a preference of one question over another. For example, one SME commented, "I
wonder if the use of the term "disease" is too general and too vague here. It may be
possible to replace with "several infectious diseases" or "infectious diseases such as
diphtheria, polio, and measles."
Finally, the results from the cognitive interviews were incorporated into the
survey instrument. Overall, there were no significant comprehension difficulties
44


across the sample. Parents provided specific word changes to the survey questions
such as "I think asking if the illnesses are serious may yield a better idea of whether
or not someone agrees that all/none of the illnesses are serious"many gets into
the grey area to begin with and "The question didnt sound right. It should say
serious illnesses I intend to protect my child from. These and other suggestions
were deliberated in a collaborative team discussion and some were incorporated
into the final version of the survey instrument. Table 3.4 shows each survey item,
survey response scale, theoretical construct thought to be associated with the
survey item, and source of the survey item. See Appendix A for the survey items
included prior to expert panel and cognitive interviews.
Table 3.4. Survey Items Response Scale, Theoretical Construct, and Source of Survey
Item, (n=23)
Survey item Response scale Theoretical construct Source of item
Generally I intend to do what my childs Strongly disagree Subjective Freed, et al.,
doctor recommends about vaccines for my child to strongly agree Norms 2010
Most of the parents I know vaccinate their Strongly disagree Subjective New item
children to strongly agree Norms developed
As a parent, I have given a lot of thought Strongly disagree Subjective New item
about vaccinations for my child3 to strongly agree Norm developed
How confident are you that you have the Very confident to Perceived New item
necessary information to make decisions about vaccination for your child not confident at all Control developed
How confident are you that you will be Very confident to Perceived New item
able to protect your child from some types of infectious disease by vaccinating him or her not confident at all Control developed
Parents should have the right to refuse Strongly disagree Perceived Freed, et al.,
vaccines that are required for school for to strongly agree Control 2010
any reason
45


Survey item Response scale Theoretical construct Source of item
I believe many of the illnesses vaccines prevent are serious3 Strongly disagree to strongly agree Evaluation of VPD/VAE Opel, et al., 20113-b
Getting vaccinated will greatly reduce the chance of getting infectious diseases like polio or measles, but will not eliminate the chance of getting these diseases completely Strongly disagree to strongly agree Evaluation of VPD/VAE Song, etal., 2014
I believe vaccines are generally safe Strongly disagree to strongly agree Evaluation of VPD/VAE Modified from Gellin, et al., 2000
I believe my child could get a serious disease if he or she were not vaccinated Strongly disagree to strongly agree Evaluation of VPD/VAE Kennedy, et al., 2011
Children get more vaccines than are good for them3 Strongly disagree to strongly agree Evaluation of VPD/VAE Gellin, et al., 2000
My child will not need vaccines for diseases that are not common anymore, like polio3 Strongly disagree to strongly agree Evaluation of VPD/VAE Adapted from Freed, et al., 2010
I am concerned about serious infectious diseases like whooping cough or measles Strongly disagree to strongly agree Evaluation of VPD/VAE Kennedy, et al., 2011
I am concerned that the ingredients in vaccines are unsafe3 Strongly disagree to strongly agree Evaluation of VPD/VAE Gust, et al., 2005
I am concerned that some vaccines cause autism in healthy children3 Strongly disagree to strongly agree Evaluation of VPD/VAE Freed, et al., 2010
I am concerned that there are serious side effects of vaccines3 Strongly disagree to strongly agree Evaluation of VPD/VAE Freed, et al., 2010
I believe it is better for my child to develop immunity by getting sick than to get a shot3 Strongly disagree to strongly agree Beliefs about vaccines Salmon, et al., 2005
I believe there has not been enough research on the safety of vaccines3 Strongly disagree to strongly agree Beliefs about vaccines Freed, et al., 2010
I believe it is better for my child to get the natural disease than to get a vaccine3 Strongly disagree to strongly agree Beliefs about vaccines Modified from Salmon, et al., 2005
46


Survey item Response scale Theoretical construct Source of item
Vaccines strengthen the immune system Strongly disagree to strongly agree Beliefs about vaccines Salmon, et al., 2005
Getting vaccines is a good way to protect my child from infectious diseases Strongly disagree to strongly agree Beliefs about vaccines Freed, et al., 2010
I am concerned that my childs immune system could be weakened by too many vaccines3 Strongly disagree to strongly agree Beliefs about vaccines Gellin, et al., 2000
I am concerned that it would be painful for my child to receive so many shots during one doctors visit3 Strongly disagree to strongly agree Beliefs about vaccines Gust, et al., 2005
aReverse scored for EFA and Cronbachs Alpha
After the survey was developed through a theory-driven systematic process,
face and content validity was established through assessment by cognitive
interviews with parents and SMEs, respectively (Litwin, 1995). Fleisch-Kincaide
readability statistics were performed on the pilot survey, indicating a 5.1 grade
level. Although caution in interpreting the grade level statistic is suggested by
Streiner and Norman (2014), this metric indicated an instrument that could be used
in populations with lower literacy. The pilot survey was now ready for
administration on a cohort of pregnant women. The piloting of the survey was
essential to knowing what to refine within the survey.
Methods for Phase Two. Survey Evaluation
In this second phase, the survey instrument was administered to a sample of
the target population and then evaluated for construct validity and reliability
(Thompson & Daniel, 1996).
The following five steps were taken to evaluate the survey instrument. Step
one identified the cohort and administered the survey instrument. Step two
explored the factors underlying the survey items. Step three assessed internal
47


consistency of the theoretical factors. Step four assesses the relationship between
the observed survey items and the underlying constructs, as a model fit. Finally, in
step five, two identical survey instruments were administered within fourteen days
of one another to the same participant to establish test re-test reliability.
Step One: Identify the Cohort and Administer the Survey Instrument
By using KPCO membership data, the Pilot cohort of pregnant women was
identified. From these women, a random sample was drawn. They were then
contacted to participate in the survey study. The identification of the Pilot cohort
and the process of administering the survey instrument are described in detail
below.
Study Setting
The study was conducted between April 2012 and December 2014 at Kaiser
Permanente Colorado (KPCO). KPCO is a non-profit integrated health care delivery
system serving the healthcare needs of approximately 600,000 members in twenty
primary care clinics across the metropolitan Denver area. KPCO uses an electronic
health record (EHR), which captures demographic data, health plan enrollment
information, encounter data including diagnosis codes, and vaccination
administration information. It is usual care for parents of young infants to receive
information about childhood development that includes standard handouts about
vaccines, the diseases they prevent, and common adverse events of vaccines, such as
swelling at the site of injection. Pregnant women receive verbal education and
informational handouts about vaccines recommended during pregnancy and are
offered these vaccines at their regularly scheduled obstetric appointments.
48


At KPCO, most vaccines are delivered at routinely scheduled well-child visits
and recorded in real time into the EHR. The EHR also allows providers to track
which vaccinations a child has received and which ones are due at the well-child
visit. Finally, the EHR allows pediatricians to code parents decisions about
vaccinations directly into the medical record, i.e. selecting some vaccines and not
others, electing for a different vaccine schedule from the recommended schedule, or
declining vaccination for their child not due to a medical contraindication. The EHR
also allows researchers to obtain accurate information regarding vaccination status
such as the number of vaccines received compared to the recommended schedule
and any reasons for delay of vaccination such as an illness, logistical barrier, or
parental concerns.
Study Population
Using the EHR, an eligible cohort of pregnant women between twenty and
thirty-two weeks gestation for the pilot survey and an eligible cohort of pregnant
women between twenty and thirty-two weeks gestation and parents of children
under the age of twelve months for the administration of the revised survey were
identified for the study. From each of these cohorts, a random sample was drawn.
An electronic data pull using International Classification of Diseases, 9th Revision,
Clinical Modification (ICD-9-CM) codes for medical exclusions and manual medical
record review was undertaken to confirm pregnancy and ascertain any medical
conditions of the fetus (i.e. medical or elective abortion, fetal deaths, fetal
abnormalities with high probability of not carrying child to full term). Exclusions
included the parent being under age eighteen, disenrollment in the health plan prior
49


to planned contact about the study, or first language of the parent was other than
English. Statistical comparisons to see if there were differences between those who
agreed to participate and those who did not participate (declined participation,
ineligible or did not respond) were conducted using chi-square test of significance
and t-test test of significance. Race, ethnicity, age, insurance type, and number of
days undervaccinated were compared.
For the Pilot cohort, participants were contacted by telephone and after
consenting, and then sent a survey through postal mail. Parents were provided a
ten-dollar gift card to a national retail store for completion of the survey.
For the Primary cohort, participants were contacted up to two times by
postal mail and asked to complete two surveys: a baseline survey followed by an
identical survey fourteen days later. At each contact, a cover letter explaining the
purpose of the study, a paper survey, and a postage-paid return envelope were
included in the postal mailing. For completion of both surveys, participants received
a fifteen-dollar gift card to a national retail store. The local human subjects review
board approved the study.
At this point, the survey instrument was developed and tested on the study
population. Using the baseline survey data from the study population who
participated in the Pilot survey and the Primary survey, responses were entered
into an Access database. Separate analyses were performed on these baseline data
to assess factor structure, internal consistency, and goodness of fit using IBM SPSS
Statistics for Windows, version 22.0 (Armonk, NY: IBM Corp.) and SAS 9.2 (Caiy, NC:
SAS Institute Inc.). The investigator hypothesized that four factors would emerge
50


from the data, based on the conceptual model developed for this investigation. The
concept model incorporates the Theory of Planned Behavior and the Health Belief
Model. It was anticipated that the factors would be highly correlated as the survey
items were measuring attitudes that are consistent across theoretical domains.
Step Two. Exploratory Factor Analysis
First, exploratory factor analyses were performed to determine the structure
of the data. Because health behavior data is correlated, principal axis factoring (PAF)
with direct oblim rotation were used. Correlated survey data was expected, as most
responses about behaviors and attitudes are similar, regardless of the construct.
Normality of the data, independence of the observations, and linearity were tested.
The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was used to
determine if there were sufficient items for each factor (>.70) and Bartletts test of
sphericity was used to determine if correlations in the data exist to allow factor
analysis (<.05) (Leech, Barrett, & Morgan, 2008).
Step Three. Internal Consistency of Survey Items
After factor analysis, Cronbachs Alpha was performed on the constructs.
This method measured how well the items measure the latent construct. High
correlations with other items in the same construct indicate the items should
provide similar scores. A reliability value of .70 and above is considered acceptable
(Cronbach, 1951).
Step Four. Confirmatory Factor Analysis
Then, to determine if the Primary cohort data fit the hypothesized
measurement model determined by EFA and Cronbachs Alpha, confirmatory factor
51


analysis (CFA) was used (Thompson, 2004). To evaluate the model fit, root mean
square error of approximation (RMSEA) and Bentler comparative fit index (CFI)
were assessed. RMSEA ranges from 0 to 1, with a smaller value indicating a better
model (<.08). CFI ranges from 0 to 1, with a higher value indicating a better model
(> .90) (Thompson, 2004).
Step Five. Test Re-test Reliability
Finally, in the Primary cohort, repeatability was measured, using the method
of test re-test. For those participants who completed the baseline survey, the same
instrument was mailed seven days later. This allowed for a fourteen-day period of
time from the baseline survey to the receipt and return of the second survey.
Pearsons reliability statistic was used to determine intra-individual variability. A
coefficient of .70 or higher is considered acceptable reliability.
Results from Phase Two. Survey Evaluation
The results from the evaluation of the measurement instrument are
described below. These results included identification of the Pilot cohort and
Primary cohort, factor analyses, assessment of internal consistency, and test re-test
reliability to establish construct validity and reliability of the survey instrument.
Results from Step One: Identify the Pilot Cohort and Administer the Survey
Instrument
For the Pilot cohort, the study population consisted of 120 pregnant women.
Of the 194 eligible, 120 participated. Participants and nonparticipants were
significantly different with regard to race (X2=10.38, df=5,N=229, p<.05). White
mothers were more likely than expected to participate than other race categories.
52


Also, participants were significantly different from nonparticipants regarding age,
t(209.06)=2.94, pc.Ol. The average age for participants (28.90) was significantly
higher than the age (27.22) of nonparticipants. There were no statistically
significant differences between those who participated and those who did not
participant in ethnicity, days enrolled in Kaiser Permanente health insurance, and
childs average days undervaccinated. Figure 3.1 shows the specific exclusions.
Figure 3.1. Study Flow Diagram for Pilot Cohort
Results from Step Two. Exploratory Factor Analysis of Pilot Cohort
Exploratory Factor Analysis (EFA) is robust against skewed data. One
variable was slightly skewed (many of the illnesses vaccines prevent are serious at -
1.39). This variable was not transformed, as interpretation of transformed variables
53


in factors is not possible. The initial EFA used Principal Axis Factor (PAF) analysis
with varimax rotation to explore the dataset. KMO was .888, indicating the sample
could yield factors from the data. Bartletts test of sphericity was significant,
indicating there was adequate correlation between variables. Two extracted
communalities were under .30. Using the criteria of eigenvalues at one or greater,
there were six factors extracted. However, the first factor had twelve items, and the
last two factors had only two items per factor. As this exploratory approach
produced too many survey items in one factor and too few survey items in other
factors, the next step was to restrict the number of factors to fit the theoretical
model.
Therefore, the next iteration of factor analyses was set to extract four
variables, leaving all other criteria the same as previously described. Since the
theoretical model defined four constructs, this provided a justification for setting
extraction at four. This rotation produced four factors, but with unfavorable results.
Again, the first factor had thirteen survey items and the last two factors had two
survey items within each factor.
The next iteration used PAF with direct oblim. Similar results were obtained.
When extracted communalities were removed, further solution was not possible
within the twenty-five rotation cycle. Therefore, Cronbachs Alpha was used to
explore the theoretical constructs.
54


Results from Step Three. Internal Consistency of Survey Items of the Pilot
Cohort
Cronbachs Alpha was used to see how well an item in a construct correlated
with the other items in the construct. High inter-item correlations imply robust
associations between items and the latent variable (Litwin, 1996). Using Cronbachs
Alpha as a measure of reliability, Table 3.5 below indicates reliable constructs in
two of the four constructs (Beliefs about Vaccinating and Evaluation of VPD and
VAE). Two constructs performed poorly (Subjective Norms about Vaccinating and
Perceived Control of Vaccinating Decisions). These two constructs had only two
questions in each construct.
Table 3.5. Internal Consistency of Survey Items within Constructs
Scale Alpha # questions
Beliefs about vaccinating .769 5
Evaluation of VPD and VAE .839 8
Subjective norms about vaccinating .352 2
Perceived control of vaccinating decisions .671 2
Items were removed if the total score correlation improved without the item.
These items are listed in Table 3.6 below.
55


Table 3.6. Survey Items Removed After Assessment of Internal Consistency, n = 8
Survey Item Latent Construct Notes
Parents should have the right to refuse vaccines that are required for school for any reason Perceived Control Deleted after pregnant cohort analysis (change in a: from .214 to .671 after deletion!
As a parent, I have given a lot of thought about vaccinations for my child Subjective Norm Deleted after pregnant cohort analysis (change in a: from .106 to .352 after deletion-)
Generally I intend to do what my childs doctor recommends about vaccines for my child Subjective Norm Deleted after pregnant cohort analysis (deleted due to low a after previous item deletions')
Most of the parents I know vaccinate their child Subjective Norm Deleted after pregnant cohort analysis (deleted due to low a after previous item deletions')
Getting vaccinated will greatly reduce the chance of getting infectious diseases like polio or measles, but will not eliminate the chance of getting these diseases completely Evaluation of VPD/VAE Deleted after pregnant cohort analysis (change in a: from .784 to .832 after deletion)
I am concerned about serious infectious diseases like whooping cough or measles Evaluation of VPD/VAE Deleted after pregnant cohort analysis (change in a: from .832 to .839 after deletion!
That it would be painful for my child to receive so many shots during one doctors visit Beliefs about vaccines Deleted after pregnant cohort analysis (change in a: from .700 to .769 after deletion!
It is better for my child to develop immunity by getting sick than to get a shot Beliefs about vaccines Deleted due to duplicity with other survey item
These exploratory analyses provided insights as to additional direction the
survey instrument needed in order to assess reliability. Additional survey items
were needed for the two constructs that performed poorly (Perceived Control of
Vaccinating Decision and Subjective Norms about Vaccinating). The survey
instrument was revised by re-assessing the literature on subjective norms and
perceived control. Additional questions were added to the survey instrument in
56


these constructs. Revisions to the instrument also included re-wording of the middle
response from "Neither agree nor disagree to "Not sure. Twenty-nine items were
on the second iteration of the survey.
Results from Step One: Identify the Primary Cohort and Administer the Survey
Instrument
For the Primary cohort, the study population consisted of 320 pregnant
women and parents of children under the age of twelve months. Participants and
nonparticipants were significantly different with regard to race (X2=13.92,
df=5,N=520, p<.05). White mothers were more likely to participate than other race
categories. Also, participants were significantly different from nonparticipants
regarding age, t(341.01)=2.93, pc.01. The average age for participants (30.39) was
significantly higher than the age (29.00) of nonparticipants. There were no
significant differences between the two groups in ethnicity, days enrolled in Kaiser
Permanente health insurance, and childs average days undervaccinated.
In Figure 3.2, of the Primary cohort of 520, there were 320 participants
recruited and who completed a baseline survey for a response rate of 62%. Two
hundred and twenty-two participants returned a second survey within the fourteen-
day period (69% response rate).
57


Figure 3.2. Study Flow Diagram for Primary Cohort
Results from Step Two. Exploratory Factor Analysis of Primary Cohort
Exploratory Factor Analysis (EFA) was conducted. One variable was
moderately skewed (vaccines protect my child at -1.65). This variable was not
transformed because interpretation of transformed variables in factors is not
possible and EFA is robust against skewed data. The EFA used Principal Axis Factor
(PAF) analysis with direct oblim to explore the dataset. KMO was .933, indicating
58


the sample could yield factors from the data. Bartletts test of sphericity was
significant, so there was enough correlation between variables to yield results. Eight
extracted communalities were under .30 and one question was duplicative of
another. The nine survey items in Table 3.7 were deleted after EFA. Using the
criteria of eigenvalues at one or greater, there were four factors extracted.
Table 3.7. Survey Items Removed after Evaluation of Factor Structure, n = 6
Survey Item Latent Construct Notes
Information on the internet about vaccines helps me to make decisions about vaccinating my child Subjective Norm Deleted after secondary cohort analysis
I think that there are other parents, like me, struggling with the decision about vaccines for their child Subjective Norm Deleted after secondary cohort analysis
School laws requiring that children have up-to-date vaccines to enter daycare or public school influence my decisions about vaccinating my child Subjective Norm Deleted after secondary cohort analysis
Allowing parents to delay vaccine doses or skip some vaccines lets parents be more in charge of their childrens health care Subjective Norm Deleted after secondary cohort analysis
Parents who skip or delay certain vaccines are relying on other people in the community being vaccinated to protect their unvaccinated children from getting sick Subjective Norm Deleted after secondary cohort analysis
The risks from getting a vaccine outweigh the risks from getting a disease Evaluation of VPD/VAE Deleted after secondary cohort analysis
Table 3.8 shows the factor structure using EFA. For all of the items that load
onto a particular construct, the other loading values are low. This factor solution
resulted in six items in the Beliefs about Vaccinating construct, four in the Perceived
Control about Vaccination Decision construct, eight in the Evaluation ofVPD/VAE
construct, and five in the Subjective Norms about Vaccinating construct for twenty-
three items. The EFA explained 56.63% of the variance, which is acceptable in
health behavior research (Hair, Black, Babin, Anderson, & Tatham, 2005).
59


Table 3.8. Factor Loadings From Principal Axis Factor Analysis With Direct Oblim
Orthogonal Rotation For A Four Factor Solution For Baseline Survey Items (N=320).
Item Factor
Beliefs Perceived Control Evaluation Subjective Norm
Many of the illnesses vaccines prevent are serious .785 -.006 .059 .058
My child could get a serious disease if not vaccinated .732 .088 .095 -.015
My child does not need vaccines for diseases that are not common -.521 -.005 .244 .010
Getting vaccines is a good way to protect my child from infectious diseases .511 -.115 -.173 .149
Vaccines strengthen the immune system .416 .070 -.232 .002
Confident I can protect my child from disease by vaccinating .406 .260 -.187 .180
Confident about your knowledge about infectious diseases .009 .896 .106 -.025
Confident about your knowledge about how vaccines work -.017 .872 -.009 .002
Confident you have the necessary information to make decisions about vaccines for my child -.011 .756 -.105 .028
Confident you are able to express your vaccine views to your childs pediatrician .028 .563 -.101 .017
The ingredients in vaccines are unsafe .112 -.059 .875 -.064
There are serious side effects of vaccination .058 -.022 .849 .006
There has not been enough research on the safety of vaccines .028 -.054 .743 -.011
My childs immune system could be weakened by too many vaccines -.157 -.039 .701 .026
Some vaccines cause autism in healthy children -.031 -.155 .635 -.076
Children get more vaccines than are good for them -.147 .007 .622 -.111
It is better for my child to get the natural disease than to get a vaccine -.380 .096 .481 -.037
Vaccines are safe .279 -.067 -.447 .254
In general, my family (e.g. sisters, brothers and cousins) have similar beliefs about vaccines as me -.082 -.023 .022 .912
In general, my parents have similar beliefs about vaccines as me .039 -.044 .012 .796
In general, most of my close friends have similar beliefs about vaccines as me -.019 .026 -.019 .511
In general, my obstetrician/childs pediatrician has similar beliefs about vaccines as me .087 .138 -.235 .380
In general, my spouse or partner has similar beliefs about vaccines as me .200 .109 .008 .296
60


Note: Bold text denotes highest factor value across all four factors. Those bolded in a column are in
that columns factor grouping.
Results from Step Three. Internal Consistency of Survey Items of the Primary
Cohort
The items under each factor produced by EFA performed very well in
internal consistency using Cronbachs Alpha and exceeded the threshold value of
>.70.
Table 3.9. Internal Consistency of Survey Items within Constructs
Scale Alpha # questions
Beliefs about vaccinating .832 6
Evaluation of VPD and VAE .921 8
Subjective norms about vaccinating .783 5
Perceived control of vaccinating decisions .864 4
Results from Step Four. Confirmatory Factor Analysis
Findings from the confirmatory factor analysis (CFA) demonstrated model
goodness of fit for the factor structure results of the EFA. The fit indices met or
exceeded the suggested cutoff values frequently cited in the statistical literature
(Bentler, 1990). The index for the Root Mean Square of Error of Approximation
(RMSEA) was 0.07 and Bentler Model Fit was 0.90.
Results from Step Five. Test Re-test Reliability
From 520 parents sent an invitation to participate in the survey study, 320
returned a baseline survey. A second survey was sent to the 320. Of these 222
returned a second completed survey within the 14-day window. This was a 69%
response rate for return of the second survey. The overall temporal stability of the
survey was very high, r = .930. Each of the four constructs had an acceptable
61


correlation coefficient of well above .70. This indicates that there is excellent test-
retest reliability for these data (Table 3.10).
Table 3.10. Results from Baseline and Second Survey Administered Two Weeks
Later.
Scale Number of items in Scale Test-Retest Reliability
BELIEFS ABOUT VACCINATING 6 .849
Many of the illnesses vaccines prevent are serious .589
My child could get a serious disease if not vaccinated .612
My child does not need vaccines for diseases that are not common .684
Getting vaccines is a good way to protect my child from infectious diseases .456
Vaccines strengthen the immune system .684
Confident I can protect my child from disease by vaccinating .689
EVALUATION OF VPD/VAE 8 .925
The ingredients in vaccines are unsafe .814
There are serious side effects of vaccination .728
There has not been enough research on the safety of vaccines .696
My child's immune system could be weakened by too many vaccines .787
Some vaccines cause autism in healthy children .807
Children get more vaccines than are good for them .767
It is better for my child to get the natural disease than to get a vaccine .773
Vaccines are safe .684
SUBJECTIVE NORMS ABOUT VACCINATING 5 .760
Friends have similar vaccine beliefs as me .617
Family have similar vaccine beliefs as me .679
Parents have similar vaccine beliefs as me .661
Spouse has similar vaccine beliefs as me .648
Pediatrician has similar vaccine beliefs as me .695
PERCEIVED CONTROL OF VACCINATING DECISIONS 4 .778
Confident about your knowledge about infectious diseases .716
Confident about your knowledge about how vaccines work .686
Confident you have the necessary information to make decisions about vaccines for my child .728
Confident you are able to express your vaccine views to your childs pediatrician .634
Overall 23 .930
62


Note: All significant at <.001 level
Phase two presented methods and results for construct validity and
reliability. After administration and analyses of the Pilot cohort of 120 pregnant
women, there was not a solution for factors using EFA. Internal consistency
demonstrated good correlation for two of the four theoretical constructs. Further
refinement of the survey instrument was necessary. The constructs of Subjective
Norms about Vaccinating and Perceived Control about Vaccination Decision were
revised by the addition of new survey items. Survey items that were similar in
wording and meaning were removed. A revised survey instrument was ready for
testing. Therefore, a new cohort of 320 pregnant women and parents of children
under twelve months of age were recruited. The baseline data were used to assess
factor analyses, internal consistency analyses, and reliability. With a larger cohort,
the revised survey instrument found a factor solution, performed very strongly in
internal consistency, had good model fit, and high test re-test reliability. Fleisch-
Kincaide readability statistics were performed on the survey, indicating a 5.2 grade
level. Next, phase three, Predictive Validity, is presented. The Primary cohort survey
constructs were assessed for association with the objective measure outcome of
undervaccinated.
Methods for Phase Three. Predictive Validity
In this third phase, the objective was to determine if the survey constructs
were associated with undervaccination. The Primary cohort was used for phase
three. There were four steps conducted in phase three. First, the survey data from
the parents survey responses was linked to their childs vaccination data. Second,
63


from the childs vaccine data, an average days undervaccinated (ADU) metric was
calculated. Third, the data was screened using descriptive statistics. Finally, in the
fourth step, model building occurred. These four steps are described in the methods
and results are then presented.
The following four steps were taken to assess predictive validity of the
survey instrument.
Step One. Linking Parent Data to Child Data
Using KPCOs electronic administrative data, the mother was linked to their
childs vaccine data. This was accomplished by associating the primary subscriber
(parent) unique family identification numbers and the childs date of birth, using the
healthplan data. Continuous enrollment was defined as no more than a thirty-one
day gap in insurance coverage. Children were also required to have been enrolled
into the health plan by three months of age and still be enrolled at six months of age
with no more than a thirty-one day gap in order to capture early childhood vaccine
doses administered during this time. In addition, children were required to be
actively using the KPCO health system as defined by having at least one well child
visit.
Step Two. Calculating Average Days Undervaccinated
Of those children who met enrollment criteria, average days under-
vaccinated (ADU) was calculated. ADU is a metric developed by Glanz, et al. (2013a),
which calculates the difference between when a vaccine dose was truly
administered and when a vaccine dose should have been administered according to
the recommended vaccine schedule. These differences were combined across all
64


vaccine doses that were due from birth to 200 days. This length of time was selected
to provide a fourteen-day window from the age of six months of the child, allowing a
two-week window for parents to bring their child to the clinic for the recommended
six-month vaccines.
Step Three. Screening the Data
Before the model building process, descriptive statistics and univariate
statistical methods (i.e. means, histograms, t-tests, and chi-square tests) were used
to screen the data. For univariate analyses, continuous variables were collapsed into
meaningful categories (See Appendix B for details). Survey construct scale variables
were transformed into dichotomous variables where values of 1 through 3.49 were
"1 (hesitant about vaccines) and 3.5 through 5 were transformed into "0 (not
hesitant about vaccines). The cut point of 3.5 was selected after careful
consideration of various response scenarios. This cut point allowed for leniency for
parents to respond to some survey items with hesitancy as the literature reports
parents who fully vaccinate have concerns about vaccines.
Step Four. Model Building
The models were constructed by the purposeful selection methods described
by Hosmer and Lemeshow (2013). First, all possible univariate logistic regression
models with each independent variable were fit. Variables that were significant at a
= .25 level were included into the multivariate model. Variables that were
determined to be meaningful to the model were also included and not removed
regardless of p value (variable constructs, age, education and race). The
independent variables used in analyses were Beliefs about Vaccinating, Evaluation
65


of Vaccine Preventable Disease/Vaccine Adverse Events, Subjective Norms about
Vaccinating, Perceived Control about Vaccination Decision, health literacy, number
of children, influences on vaccine decision, use of social media for health
information, age, education and race.
The independent variables were then removed sequentially based on their
statistical significance at a = .05 level using the log-likelihood ratio test. As they
were removed, their potential as a confounder was quantified by calculating a
change in the coefficients of the model with and without the variable. Covariates
that produced changes in the coefficient greater than 15% were considered
confounders and were left in the final model. Once the final model of main effects
was established, meaningful interaction terms were considered and not pursued
due to theoretical unlikeliness. The model building process concluded with
goodness of fit testing and diagnostics (Hosmer, Lemeshow & Sturdivant, 2013) to
assess the models fit and performance.
Results from Phase Three. Predictive Validity
Using the baseline survey data from the Primary cohort of 320 pregnant
women and parents of children under the age of twelve months and the vaccination
status of their child, predictive validity was assessed. The results from linking
parent data to the childs vaccine data, calculating undervaccinated status of the
child, data screening, and model building are presented below.
66


Results from Linking Parent Data with Child Data and Calculating
Undervaccinated Status
First, missingness of data from linking the mother to the child was evaluated.
Of the 320 children of parents who completed baseline surveys, 301 were linked.
The nineteen that did not link were due to: one fetal death, one parent lost health
insurance, six left KPCO health insurance plan prior to delivery, and eleven had
missing unique identifiers in the EHR. Of the 301, forty-seven children were
excluded for not being enrolled within three months of the birth and one was not
actively using the system. Using enrollment criteria, 253 were actively using the
KPCO health system. Of the 253, twelve did not have vaccine information available
at age 200 days. Therefore, 241 children were eligible for analyses (see Figure 3.3).
Of the 241, twenty-nine were undervaccinated before age 200 days.
Undervaccinated was a dichotomous variable where the metric average days
undervaccinated was either "0 days undervaccinated (fully vaccinated on time) or
">0 days undervaccinated (undervaccinated). After manual medical record review
of the twenty-nine undervacinated cases, two were excluded. For these two, the
reason for undervaccination was related to logistic barriers and not parental choice.
Two hundred thirty nine cases were included in the analyses; twenty-seven were
undervaccinated before age 200 days.
67


Figure 3.3. Study Flow Diagram of Primary Cohort Linking Mother to Child
Results from Data Screening
Overall, the cohort had an average age of thirty-two years. The majority were
white (88.7%), married (86.2%) and had a post-high school education (94%).
Additional descriptive information about the cohort is available in Table 3.11.
Table 3.11. Characteristics of Pregnant Women and Parent Cohort by Vaccine
Behavior, N=239
Characteristic3
Gender n (%)
Female
Age in years, m(SD)
Age in years, range
Race, n(%)
White
Non white
ADU>0, n=27 ADU = 0, n=212 p value
26 (96.3] 205 [96.7) .913
32.30 (4.65] 24.0-41.0 31.8 [4.79) 19.0-46.0 .593
26 (96.3] 1 [2.9) 186 [88.2) 26 [11.8) .999
3 [11.1) 24 [88.9) 31 [14.6) 181 [85.4) .823
Hispanic ethnicity, n(%)
Yes
No
68


Characteristic3 ADU>0, n=27 ADU = 0, n=212 p value
Education n (%) High school degree or less Some college or more 1 [3.7) 25 (96.3) 12 (5.7) 199 (94.3) .999
Employment n (%) Employed full time All other 12 (44.4) 15 (55.6) 150 (70.8) 62 (29.2) .426
Marital status n(%) Married All other 24 (88.9) 3 (9.01) 182 (85.8) 30 (14.2) .999
Income n (%) 80,000Kor less More than 80,000K 23 (57.5) 17 (42.5) 161 (48.2) 173 (51.8) .999
aFor age, independent t-test was used for this continuous variable. For all other variables, chi-square
was used for categorical variables and Fischers exact test was used when cell size was less than five.
Using the Pearson Chi-squared Test for Independence statistic or Fisher
Exact Test when cell size was less than five, parents of undervaccinated children
were less likely to be employed full-time than parents whose children were fully
vaccinated. Additionally, results from the univariate analysis (Table 3.12) found that
parents of undervaccinated children were more likely to have more children, have
adequate health literacy, use social media to search for health information and make
their own decisions without influence of their childs doctor than parents of children
who were fully vaccinated. In the univariate analysis, Beliefs about Vaccinating,
Evaluation ofVPD/VAE, Subjective Norms about Vaccinating, employment, number
of children, health literacy, physician influence, use of social media, parent control of
decision, were all significant in the model while Perceived Control about Vaccination
Decision, race, age, education and income were not statistically significant at p=<.05.
These characteristics are shown in the univariate analysis (Table 3.12). For
example, parents whose child was undervaccinated were more than seven times
69


more likely to have hesitant responses on the survey construct Beliefs about
Vaccinating than parents whose child was fully vaccinated on time.
Table 3.12. Univariate Analyses of Categorical Covariates from Primary Pregnant
Mother and Parent Cohort, N=239
Covariate OR 95% Cl p value
Beliefs about Vaccinating3 7.27 3.06,17.30 <.001
Evaluation of VPD/VAE3 17.96 4.15, 77.80 <.001
Perceived Control about Vaccination Decision3 1.90 0.83, 4.35 .127
Subjective Norms about Vaccinating3 6.68 2.85,15.63 <.001
Full time employment vs. all others (part -time, retired, student, unemployed, stay-at-home-) 0.33 0.14, 0.76 .009
2 or more children vs. less than 2 children 3.61 1.52,8.57 .004
Adequate health literacy vs. inadequate health literacy 4.65 1.06,20.31 .041
Influence of doctor in vaccine decision vs. other influences in vaccine decision (spouse, friends, family, no one, other-) 0.25 0.07, 0.88 .030
Use social media for health information 1/week or more vs. use of social media for health information once a month or less 3.54 1.15,10.84 .027
Parent makes vaccine decision vs. shared decision or doctor makes decision 5.09 1.16,22.21 .031
Race as white versus all other races 3.50 0.45,26.89 .229
Education as college versus high school or less 1.57 0.20,12.56 .672
Age as 30 years or older versus under 30 0.90 0.39,2.12 .816
Income 80K and above versus under 80K 1.02 0.46,2.28 .960
OR = odds ratio; Cl = confidence interval
Note: p-value significance = .05
aReferent is nonhesitancy
Results from Model Building
A logistic regression model was constructed where being undervaccinated
before age 200 days (yes/no) was the binary outcome. This model determined the
factors in the cohort data that contributed significantly to children being classified
as undervaccinated. Based on a p-value of .25 in the univariate analyses, the
independent variables considered in the full model were employment, race, number
of children, health literacy, physician influence, social media use for health
70


information, and preference for vaccine decision making. The four survey
constructs, Beliefs about Vaccinating, Evaluation ofVPD/VAE, Subjective Norms
about Vaccinating, Perceived Control about Vaccination Decision, and covariates
age, race, and education remained in the model regardless of significance.
The full model was built using the four survey constructs in the same model.
However, when the full model was run, three of the constructs and many covariates
that were significant in the univariate analyses were not significant. When a reduced
model using only survey constructs was built, two of the four constructs were
significant with an additional construct trended close to significance: Beliefs about
Vaccinating (adjusted odds ratio (AOR), 2.66; 95% Confidence Interval (Cl) 0.98,
7.24), Evaluation ofVPD/VAE (AOR, 11.00, Cl 2.40, 50.50), Subjective Norms about
Vaccinating (AOR, 3.05, Cl 1.15, 8.13), and Perceived Control about Vaccination
Decision (AOR, 0.44, Cl 0.25,1.81).
When each of the four constructs was modeled without any other constructs
in the model, each construct was significant by association with the outcome of
undervaccination. Thus, the results below (Table 3.13) use one construct per model,
adjusting the model by using the same covariates in each model.
71


Table 3.13. Logistic Regression Analyses of Independent Modeling of Survey
Constructs and Factors Associated with Undervaccination with Adjusted Odds
Ratios
Covariate AOR 95% Cl p value15
aBeliefs about Vaccinating
Hesitant 10.01 2.11,18.00 <.001
Nonhesitant Referent -
Evaluation ofVPD/VAE
Hesitant 24.92 5.22,119.01 <.001
Nonhesitant Referent
aPerceived Control about Vaccination Decision
Hesitant 0.39 0.16, 0.96 .040
Nonhesitant Referent -
Subjective Norms about Vaccinating
Hesitant 13.65 4.71,39.53 <.001
Nonhesitant Referent - -
aSurvey responses were binary where l-3.49=hesitant and 3.5-5=nonhesitant.
bSignificance level is <.05
Adjusted ORs are adjusted for employment, education, race, age and number of children
AOR = adjusted odds ratio; Cl = confidence interval
Hosmer and Lemeshow: Beliefs about vaccinating= .719; Evaluation of VPD/VAE = .708; Perceived
Control = .541; Subjective Norms = .294
In separate models, all four survey constructs were significantly associated
with undervaccination after controlling for employment status, education, race, age,
and number of children. Being classified as hesitant on Beliefs about Vaccinating
(adjusted odds ratio (AOR), 10.01; 95% Confidence Interval (Cl), 2.11,18.00),
Evaluation ofVPD/VAE (AOR, 24.92, Cl, 5.22,119.01) and Subjective Norms (AOR,
13.65, Cl, 4.71, 39.53) were significantly associated with undervaccinated status.
Being classified as nonhesitant on Perceived Control about Vaccination Decision was
negatively associated with undervaccinated status (AOR, .0.39, Cl, 0.16, 0.96).
Evaluation ofVPD/VAE had the highest odds ratio. Parents who were classified as
hesitant on the Evaluation ofVPD/VAE construct were almost twenty-five times as
likely to have an undervaccinated child as compared to those classified as
72


nonhesitant on the construct. Diagnostic testing using Cooks distance were all less
than one, indicating overall influence of a case on the model was not a factor (Cook
& Weisberg, 1982). Finally, goodness of fit was checked using the Hosmer and
Lemeshow metric. All models were nonsignificant, indicating good fit.
Next, a Validation cohort of two similar study groups was constructed. The
first group was the original cohort used (the Primary cohort) in the logistic
regression analyses. The second group consisted of the baseline usual care
participants of an online intervention to reduce vaccine hesitancy and increase
vaccine behaviors for pregnant mothers and parents of children less than twelve
months of age. Those children who were undervaccinated underwent medical
record review. Two cases were removed after medical record review as the cause of
undervaccination were logistical barriers rather than parental choice to decline or
delay vaccines. Table 3.14 describes the Validation cohort. Descriptive statistics (t-
tests and chi-square) were conducted to test for differences between the
undervaccinated and fully vaccinated groups.
Table 3.14. Characteristics of the Validation Cohort by Vaccine Behavior, N=374
ADU>0, ADU = 0,
n=40 n=334
Characteristic3
Gender n (%)
Female 39 (97.5) 323 (96.7)
Age in years, m (SD) 31.8 (4.23) 31.9 (4.70)
Age in years, range 24.0-41.0 19.0-46.0
Race, n(%)
White 39 (97.5) 292 (87.7)
Non white 1 (2.5) 41 (12.3)
Hispanic ethnicity, n(%)
Yes 5 (12.5) 41 (12.3)
No 35 (87.5) 293 (87.7)
Education n (%)
73


Characteristic3 ADU>0, n=40 ADU = 0, n=334
High school degree or less 1 (2.5) 14 (4.2)
Some college/college degree 39 (97.5) 319 (95.8)
Employment n (%)
Employed full time 18 (45.0) 232 (69.5)
All other 22 (55.0) 102 (30.5)
Marital status n(%)
Married 36 (90.0) 297 (88.9)
All other 4 (10.0) 37 (11.1)
Income n(%)
80,000Kor less 23 (57.5) 161 (48.2)
More than 80,000K 17 (42.5) 173 (51.8)
aFor age, independent t-test was used for this continuous variable. For all other variables, chi-square
was used for categorical variables and Fischers exact test was used when cell size was less than five.
In this Validation cohort, parents had similar characteristics as to the
previous Primary cohort. Parents were, on average, almost thirty-two years of age. A
majority of the parents were white (898%), highly educated (96%), and married
(89%).
There were statistical differences between those undervaccinated and those
who were fully vaccinated that were very similar to the smaller Primary cohort.
These characteristics are shown in the univariate analysis (Table 3.15). Parents of
undervaccinated children were less likely to be employed full-time than parents
whose children were fully vaccinated. Additionally, parents of undervaccinated
children were more likely to have more children, and make their own decisions
without influence of their childs doctor than parents of children who were fully
vaccinated. Parents with two or more children were 2.3 times as likely to have an
undervaccinated child as parents with fewer than two children. However, three
variables were not significant: health literacy, use social media to search for health
information, and influence of doctor on their vaccine decision.
74


Table 3.15. Univariate Analyses of Categorical Covariates from Validation Study
Cohort, N=374
Covariate OR 95% Cl P
aBeliefs about vaccinating 8.61 4.15,17.86 <.001
Evaluation of VPD/VAE 15.24 5.30, 43.84 <.001
aPerceived control about vaccination decision 2.29 1.14, 4.58 .020
Subjective norms about vaccinating 8.43 4.13,17.23 <.001
2 or more children versus less than 2 children 2.30 1.18, 4.46 .014
Full time employment vs. all others (part -time, retired, student, unemployed, stay-at-home] 0.36 0.19, 0.70 .003
Influence of doctor in vx decision vs. other influences in vaccine decision (spouse, friends, family, no one, other] 0.39 0.16, 0.95 .037
Parent makes vaccine decision vs. shared decision or doctor makes decision 5.42 1.63,17.99 .006
Race as white versus all other races 5.48 0.73, 40.94 .098
Education as college versus high school or less 1.71 0.22,13.37 .608
Age as 30 years or older versus under 30 0.70 0.35,1.37 .290
Income 80K and above versus under 80K 0.70 0.35,1.37 .268
Source of dataset: Primary versus Usual Care 1.20 0.60,2.40 .617
OR = odds ratio; Cl = confidence interval
Note: p-value significance = .05
aReferent is nonhesitancy
Next, a logistic model was constructed by again using undervaccinated at 200
days status (ADU) as the outcome. Independent variables were obtained from the
data screening processes. The methods used in the previous logistic regression were
applied. Covariates were assessed for confounding and education remained in the
model due to confounding. With the larger strata of forty undervaccinated data
points, the model results using all constructs and significant or clinically important
covariates are presented in Table 3.16.
75


Table 3.16. Logistic Regression Analysis of Survey Constructs and Factors
Associated with Undervaccination with Adjusted Odds Ratios, N = 374
Covariate AOR 95% Cl flower) p value
Beliefs about vaccinating
Hesitant 3.27 1.27, 8.42 .014
Nonhesitant Referent - -
Evaluation of VPD/VAE
Hesitant 9.34 2.79,31.29 <.001
Nonhesitant Referent - -
Perceived Control about Vaccination Decision
Hesitant 0.74 0.29,1.90 .533
Nonhesitant Referent - -
Subjective Norms about Vaccinating
Hesitant 4.90 1.97,12.14 .001
Nonhesitant Referent - -
Number of children
2 or more children 4.13 1.70,10.05 .002
< 2 children Referent - -
Employment
Full time employment 0.35 0.15, 0.82 .015
All other categories Referent - -
Education
College education 1.79 0.17,19.14 .630
High school or less Referent - -
Race
white 10.15 1.03,99.91 .047
All other races Referent - -
Age
Age 30 years or older 0.38 0.15,1.52 .041
Age <30 years Referent - -
Source of dataset
Primary cohort 1.24 0.50,1.97 .644
Usual care Referent - -
Survey responses were binary where l-3.49=hesitant and 3.5-5=nonhesitant
Hosmer and Lemeshow = .481
AOR = adjusted odds ratio; Cl = confidence interval
Three of the four survey constructs were significantly associated with
undervaccination after controlling for the other survey constructs, employment,
education, race, age, number of children and source of the dataset. Being classified
as hesitant on Beliefs about Vaccinating (AOR, 3.27, Cl, 1.27, 8.42), Evaluation of
VPD/VAE (AOR, 9.34, Cl, 2.79, 31.29) and Subjective Norms about Vaccinating (AOR,
76


4.89, Cl, 1.97,12.14) all had an increased likelihood of being associated with
undervaccinated status. Perceived Control about Vaccination Decision was not
significantly associated the likelihood of having undervaccinated status (AOR, .74,
0.29,1.90). Similar to the previous logistic regression using independent models on
the Primary cohort, in this Validation cohort, parents who were classified as
hesitant on Evaluation of VPD/VAE were 9.3 times more likely to have an
undervaccinated child than parents who were classified as nonhesitant. Cooks
distance were all less than one so no data points were outliers. Finally, goodness of
fit was assessed using the Hosmer and Lemeshow test of significance. The fit is good
as the test is insignificant at 0.481.
Discussion
The objective of this research was to develop and evaluate a survey
instrument designed to measure parents concerns, attitudes, beliefs and intentions
about vaccines for their child. To accomplish this, qualitative and quantitative
methods were used. First, a comprehensive literature review identified potential
survey items. Next, a panel of vaccine experts and parents from the target
population provided feedback on the survey. This step provided face and content
validity to the survey instrument. The pilot survey was administered to 120
pregnant women. Evaluation of internal consistency of the baseline data resulted in
two of the four constructs with acceptable alpha values.
The survey instrument was revised, and new items were added. Then, 320
pregnant women and parents of children under the age of twelve months were
administered the revised survey instrument. Data from the revised survey were
77


analyzed for factor structure, internal consistency, repeatability and model fit. These
results provided construct validity of the survey instrument. Finally, logistic
regression was performed to determine if there was an association between the
survey constructs developed through the factor analysis and undervaccination at
age 200 days. By modeling one survey construct per model, there were strong
associations with being undervaccinated at 200 days for three of the constructs:
Beliefs about Vaccinating, Evaluation ofVPD/VAE, and Subjective Norms about
Vaccinating. Using the approach of one construct per model, predictive validity was
established, yet it was not the traditional model approach. For this reason, a
validation was performed on a larger cohort and although the odds ratios were
lower when all survey constructs were in one model, it performed similar to the
smaller cohort. This provided validation that with larger sample, three of the four
constructs are associated with undervaccination.
The investigation had relatively high participation rates. For the Pilot cohort
of 120 pregnant women, the response rate was 62%. This was similar for the
Primary cohort of 320 (62%). This higher response rate helped shield against
missing data not at random (Evans, 1991).
Engaging subject matter experts to review the instrument and then
conducting cognitive interviews with parents to pilot the survey contributed to
further revision of the survey instrument. Analyzing the quantitative and qualitative
data obtained from these engagement processes produced an improved survey.
When designing and evaluating surveys, there are several important points
which should be considered. First, it is important to have a theoretical concept to
78


guide the decision making process in designing and evaluating a survey instrument.
Although statistical results provide some guidance in deciding what items remain in
the construct and which are removed, much of the decision-making and
interpretation of meaning in the data is made by the investigator (Blair, Czaja, &
Blair, 2013; Fabrigar, Wegener, McCullum, & Strahan, 1999).
Second, the time, cost and labor involved in primary data collection and data
analysis is intensive. Most surveys require technical expertise to either design a
paper-based copy or proficiency in survey software and code such as Structured
Query Language (SQL) Hyper Text Markup Language (html) to develop web-based
survey instruments. There are also costs of incentives to compensate participants
time for completion of surveys. Much labor is expended on mailing surveys,
monitoring web-based survey platforms, and follow-up with participants, and data
entry and data cleaning procedures. These processes have protracted timelines that
extend into years of research.
Third, what is a relevant vaccine concern to a parent today can evolve and
change over time. Survey instruments need to be designed so these new concerns
can quickly be assimilated into the instrument without sacrificing reliability and
validity of the instrument. This can be accomplished by adding the items to the next
administration and re-analyzing the baseline data using the steps outlined in this
chapter.
Survey development is a collaborative and iterative process. Development
that is followed by administration of the instrument and analyses, and then followed
by additional development provided applicable results. This iterative process was
79


used for this survey by piloting it on a cohort of pregnant women, analyzing the
data, revising the survey instrument, then administering the Concerns, Attitudes,
Beliefs, and Intentions about Vaccines (CABI-V) survey to a new cohort of parents.
This permitted careful development and understanding of the constructs and item
measurements.
This study adds to the knowledge base for future survey-based studies. One
of the most important operational findings in this study was that exploratory factor
analysis power estimates published in the literature for surveys do not always
converge towards a solution. In the literature, sample size estimates average from
100 as the low estimate to over 300 as the optimal estimate to produce stable
statistical results (Nunnally, Bernstein, & Berge, 1967; Nunnally, 1978; Comrey,
1988). Tinsley and Tinsely (1987) recommend five to ten subjects per survey item.
In this investigation, the 120 baseline surveys obtained in the Pilot cohort was not
an adequate sample did not result in a solution due to no convergence.
Additionally, when using undervaccinated as an outcome, it is imperative to
have adequate sample size. Undervaccinated is rare in the population and
calculating it precisely and accurately is complex. Because it is important to capture
only truly undervaccination by parental choice, careful medical record review is
required. In this study, medical record review validated the outcome of
undervaccination. Exclusion of reasons other than choice is necessary. This research
found that children on public or high deductible insurance do not have continuous
enrollment. Thus, they are removed from the final analyses for the outcome of
undervaccination. For this study, 320 pregnant women and parents were recruited.
80


Of these, 241 of their children had continuous enrollment (75%) from which to
calculate undervaccination status at 200 days of age. Opel et al. (2013b) had similar
proportions, 310 of the 437 children had continuous enrollment at nineteen months
of age (71%). However, Opel allowed a more liberal definition for breaks in
insurance coverage but had an extended observation period to nineteen months of
age. This highlights the importance of adequate sample size and provides valuable
information on expected data loss when using restrictions of enrollment for future
studies.
It is also important to consider the outcome measure of undervaccinated
produced by the ADU algorithm was associated with being hesitant on the survey
constructs. Although there have been numerous single study surveys designed to
measure parents hesitancy about vaccines or knowledge, attitudes and beliefs
about vaccines, there has not been extensive rigor in developing and evaluating an
instrument that can be used in multiple settings and research designs, such as
interventions.
There are several significant limitations to this investigation. The outcome
variable used in logistic regression was rare. Therefore, continuous enrollment of
239 children for 200 days after birth was available and the logistic regression had
only twenty-seven (11%) in the undervaccinated group. In addition, health behavior
survey data is highly correlated. It was hypothesized a priori that the survey
constructs would be multicollinear. Therefore, by using each survey construct
independent of the other constructs through logistic regression, the covariates were
added and statistical significance was observed. When a larger sample was
81


assembled by adding a cohort of similar parents from an intervention study, there
was a significant association between undervaccination and three survey constructs.
It is possible that the survey instrument is not generalizable to other
geographical regions and research. It is important to test the instrument in other
geographic regions and with other types of intervention studies to evaluate its
strength of association with undervaccination. The construct of Perceived Control of
Vaccination Decision was not associated with undervaccination in univariate and
multivariate analyses. There is a need to test additional survey items and assess
which items perform well in differentiating vaccine status.
There is an urgent need to develop and implement effective interventions to
improve vaccination and reduce parental concerns. It is also imperative to have a
survey instrument that accurately and reliably measures parental concerns about
vaccines over time. Through this development and evaluation process, an
instrument that does have reliability and validity was produced. The next steps are
to evaluate its use in interventions, across multiple settings and with diverse
populations to demonstrate its reliability across populations. Notably, these settings
and populations may drive the need to change the instrument. If effective across
geography, setting, and populations, the survey instrument would represent a useful
and broadly applicable resource to researchers, public policy and health care
practice. Using the same questions over ranges of populations by multiple studies
could potentially provide a repository of data for change in attitude and beliefs
about vaccines.
82


This chapter described the development and evaluation of a survey
instrument to measure parents concerns, attitudes, beliefs, and intentions about
childhood vaccines. The instrument performed well in the areas of face, content and
construct validity, internal consistency, and test re-test reliability. Analyses
demonstrated strong association between survey constructs and undervaccination.
In individual models, three constructs demonstrated significant association with
undervaccination. However, when the constructs were entered into the model
together, the results were different and two constructs were not significant.
In the next chapter, the Pilot cohort data was analyzed over four time points.
Repeated measures methods were used to assess changes of the survey items.
83


CHAPTER IV
DESCRIPTION OF PREGNANT MOTHERS VACCINE DECISION MAKING PROCESS
OVER TIME
In this chapter, the survey data from the Pilot cohort administered to 120
pregnant women at four time points is presented. The survey items were analyzed
over time using inferential statistics to assess significant changes in measures. First,
there is a brief introduction to decision making of parents, the sources of
information they rely on, and literature on previous investigations on parents
decision making over time. This is followed by the study design, methods, results
and conclusions.
Introduction
Parents face decisions about vaccines for their child within the first hours
after birth. For parents, this decision making process can seem overwhelming. They
often turn to multiple sources of information about vaccines, which can be
inaccurate, depending on the source. Additionally, parents must accurately
comprehend scientific information about vaccines and the immune system that may
be complex, poorly presented or lack details that allow for full understanding
(Downs, 2008). Influential sources of information include significant others (i.e.
family, friends), the childs pediatrician, and media sources. These sources can also
influence parents concerns, attitudes, beliefs and intentions about vaccines for their
child (Kennedy, 2011).
Although the topic of decision making about vaccines has been well covered
in the literature (Benin, et al., 2006; Serpell & Green, 2006; Sturm, Mays, & Zimet,
84


2005), there are few studies that examine the decision making process
prospectively. Most survey studies use retrospective attitude assessment and self-
reported vaccine status, which limit the reliability of the data and results. Several
longitudinal survey studies have been conducted to measure parents intentions to
vaccinate prior to or shortly after the birth of their child as compared to the parents
actual vaccine behavior. These studies used self-reported status of the parent or
review of available medical records to capture this behavior.
Wroe, et al. (2005) recruited 195 women in their third trimester of
pregnancy and measured intentions to vaccinate their child, perceptions of risks and
benefits of vaccinating, and feelings of regret of vaccine decision. She found that
feelings of responsibility and regret were strong predictors of decisions about
vaccination. Opel and colleagues (2013b) administered a vaccine hesitancy screener
to 437 parents of infants who were two months of age and followed them forward in
time to collect their vaccine behavior using the EHR. Of the 310 children who had
continuous enrollment and complete EHR records at age nineteen months, those
with higher hesitancy scores at two months of age had more days undervaccinated.
These early investigations use prospective decision-making and demonstrate the
importance of well-designed data collection and access to objective vaccine
behaviors.
This study provides longitudinal descriptive information on a cohort of
pregnant mothers. The aim of this study was to prospectively capture pregnant
womens concerns, attitudes, beliefs, and intentions about vaccines for their child
for four time pointsprior to the birth through the sixth month of age of the child.
85


First, a cohort of pregnant mothers identified through the EHR was contacted. Next,
the survey developed in chapter three was administered at four time points to
mothers. Additional descriptive variables were added to the survey (i.e. health
literacy measures, demographics, decision-making preferences, hesitancy about
vaccines). Finally, analyses were performed to describe the data over time.
Methods for Longitudinal Data Collection
Design of the Study
To address the need for further descriptive information collected
prospectively on parents decision-making about vaccines for their child over time, a
prospective survey study was conducted with 120 pregnant women who were
followed forward in time from pregnancy through six months of age of their child.
The Concerns, Attitudes, Beliefs and Intentions (CABI) survey, which was a pilot
survey developed for this study, was administered a four time points: second to
third trimester of the mother and then at two, four and six months of age of the
child. These time points were strategically selected to align the survey collection at
points when parents make decisions about vaccines for their child as they typically
attend well-child checks when the infant is two, four, and six months of age.
Demographic information was collected at the baseline survey. This prospective
approach protects the predictor measure from becoming influenced by the
knowledge of the outcome (vaccine behavior) and reduces recall bias (Figure 4.1).
86


Time
Figure 4.1. Design of Prospective Survey Data Collection, N=120
Study Population for the Longitudinal Study
Within the KPCO health care system, approximately 5,000 live births occur
each year. Prior to the birth, pregnant women receive care within the health system
at KPCO clinics. There were three steps to creating the longitudinal cohort. First,
using the EHR, women between twenty to thirty-two weeks gestation were
identified. In this step, women with unknown pregnancy status or serious medical
conditions of fetus or mother were excluded. Next, a random sample was drawn
from the eligible cohort. Finally, manual medical record review was performed to
confirm eligibility. The details of these three steps are provided below.
Step One. Identification of Study Cohort
87


Pregnant women who were in their late second to first half of their third
trimester (between twenty and thirty-two weeks of gestation) of pregnancy and
members of the integrated health delivery system, Kaiser Permanente Colorado,
were identified for this study. The estimated delivery date (EDD) is available in the
EHR to estimate pregnancy parameters. Exclusions were made through the initial
electronic data pull (using International Classification of Diseases, 9th Revision,
Clinical Modification (ICD-9-CM) diagnosis codes for inclusion and exclusion
criteria). The decision was made to begin recruitment after twenty weeks gestation
as most women receive an ultrasound prior to twenty weeks in which fetal genetic
abnormality screening is performed. Thus, those women with serious fetal
compromise were excluded from the cohort (i.e. fatal heart condition, anencephaly,
spontaneous abortion). Additionally, women with less than 60 days enrollment in
KPCO or less than one obstetric visit were excluded. Pregnant women age eighteen
or over were eligible for the study.
Step Two. Random Sample
Using the EHR, estimated delivery dates (EDD) were extracted from the data
files. Pregnant women falling within the parameters of twenty to thirty-two weeks
gestation were then compiled into a cohort. From this eligible cohort, a random
sample was drawn.
Step Three. Chart Review Verification
Manual chart review to confirm pregnancy and verify eligibility was
performed on the entire random sample and those ineligible were removed from
the cohort. Pregnant women with a serious health condition related to their
88


pregnancy (i.e. very high-risk pregnancy with long-term bed rest; high probability of
not carrying the child to full term) were excluded from the cohort through manual
medical record review. Pregnant mothers whose first language was other than
English were also excluded from the study sample as all materials were in English
only.
Enrollment
The research was conducted between July 2012 and December 2013 at
Kaiser Permanente Colorado (KPCO). Eligible pregnant women were initially
contacted to participate in the study by a mailed letter and were provided an opt-
out postcard to mail back to the investigator to indicate they were not interested in
participating. Once verbal consent was obtained, a validated vaccine hesitancy
screener, Parent Attitudes about Childhood Vaccines (PAC-V) was administered
over the telephone and the investigator transcribed answers. The PAC-V was
administered as part of the feasibility study and is described in detail elsewhere in
this chapter.
Consent forms and PAC-V responses were stored within locked storage at
KPCO, using unique study IDs as the only method of identification. PAC-V scores
were also entered into a password protected Access Database. Participants were
requested to return mailed survey within one weeks time from receipt of the survey
materials.
Participants received a gift card for ten dollars to a retail store for each
survey they completed, for a total of forty dollars in incentive across four surveys.
Gift cards were sent to participants via postal mail with a personalized thank you
89


letter. Subsequent surveys were mailed to parents when their children were two
months of age, four months of age, and six months of age. If a survey was not
received by postal mail within seven to ten days, parents were contacted by phone
or email and were then sent by postal mail a blank follow up survey to complete in
the event they misplaced the initial mailed survey. All surveys were tracked as to
when they were sent out in postal mail and when surveys were returned, using
unique study identifiers, an excel spreadsheet, and Access database.
Analytic methods
The analytic plan for survey results had two main steps. In the first step, the
survey measures were defined. The specific response category for each survey
question is available in Appendix C. In the second step, the statistical approaches to
analyze the survey results are described.
Step One. Defining Survey Measures
For the baseline survey, twenty-three survey items were created to measure
the concerns, attitudes, beliefs and intentions of parents about vaccinating their
child. Survey items consisted of continuous, categorical and dichotomous variables.
These survey items are described in detail in chapter three. Additional survey items
included measures of health literacy and numeracy, demographics (age, race,
ethnicity, income, number of children, education, marital status, and employment),
vaccine intention, decision-making preferences about vaccines for children, and
personal reference to someone who experienced a vaccine adverse event.
Health literacy and numeracy. Health literacy was measured using a
validated, self-reported one-question measure developed by Chew, etal. (2008). The
90


Full Text

PAGE 1

CONCERNS, ATTITUDES BELIEFS AND INTENTIONS OF PARENTS ABOUT VACCINES FOR THEIR CHILD: DEVELOPMENT AND EVALUATION OF A SURVEY INSTRUMENT IN AN INTEGRATED HEALTH CARE SYSTEM IN COLORADO B y JO ANN SHOUP B.A., Edinboro University of Pennsylvania M.A., Edinb oro University of Pennsylvania M.S.W., University of Pittsburgh M.S., Carnegie Mellon University A thesis submitted to the Faculty of the Graduate School of the University of Colorado Denver in partial fulfillment of the requirements for the degree of D octor of Philosophy Public Affairs 2015

PAGE 2

ii This thesis for the Doctor of Philosophy degree by Jo Ann Shoup has been approved for the Public Affairs Program by Mary Guy, Advisor Jessica Sowa, Chair Danielle Varda Jason M. Glanz November 18, 2015

PAGE 3

iii Shoup, Jo Ann ( Ph.D., Public Affairs ) Concerns, Attitudes, Beliefs and Intentions of Parents about Vaccines for their Child: Development and Evaluation of a Survey Instrument in an Integrated Health Care System in Colorado Thesis directed by Professor M ary Guy ABSTRACT R outine chil dhood vaccination has led to the eradication, elimination or control of previously common infectious diseases Through mandates vaccination rates have remained hi gh However, parents have concerns about the safety of vaccines. The purpose of this dissertation was to develop, implement and evaluate a theory driven survey instrument to measure parents concerns, attitudes, beliefs, and intentions about vaccines for their child. There were three phases to develop and evaluate the survey instrument. First, the instrument was developed through extensive literature review, use of special matter experts and cognitive interviews with the target population to establish face and content validity. Second, t he pilot survey instrument was administered to 120 pregnant mothers. After analyses, further revisions were made. The revised survey was then administered to 320 pregnant mothers and parents of children under twelve months of age The baseline responses were assessed using factor analys is (FA) and internal consistency (IC) methods t o establish construct validity and reliability F A yielded four factors : Beliefs about Vaccinating ; Evaluation of Vaccine Preventable Disease (VPD) and Vaccine Adverse Events (VAE) ; Subjective Norms about

PAGE 4

iv Vacc inating; Perceived Control over Vaccinating Decision. Using repeatability, temporal stability of the instrument was very high, r = .930. Finally, association of four survey constructs with undervaccinated status was examined using logistic regression (LR). U nivariate analyses demonstrated strong relationships between survey constructs and undervaccination status In separate model s controlling for covariates, significant associations between three survey constructs and undervacc ination were found : Beliefs about Vaccinating, (adjusted odds ratio (AOR), 10.01; 95% Confidence Interval (CI) 2.11, 18.00); Evaluation of VPD/VAE (AOR, 24.92, CI 5.22, 119.01); and Subjective Norms about Vaccinating (AOR, 13.65, CI 4.71, 39.53). Additional analyses assessed the surv ey items for trends over time, and different measures of hesitancy about vaccines were compared across survey items Overall, parents hesitancy about vaccines decreased after the birth of their child. Concerns that continued from pregnancy through six mon ths of age of the child were: concerns about vaccine side effects, ingredients in vaccines, and concerns that vaccines cause autism. Different h esitancy measures were consistent in differentiating survey item responses of those who were hesitant versus non hesitant. The form and content of this abstract are approved. I recommend its publication. Approved: Mary Guy

PAGE 5

v ACKNOWLEDGMENTS I would like to acknowledge the financial support of the George Bennett Dissertation G rant thro ugh the Informed Medical Decisio ns Foundation, and the scientific support of the Institute for Health Research, Kaiser Permanente Colorado. Without these support s I would not have been able to pursue the cohort recruitment and additional methods necessary for rel iability and validity. Additionally, I gratefully thank the University Scholars C ommittee at the University of Colorado for their ongoing financial commitment to my academic pursuits. I would like to thank a number of people who have been central to the c ompletion of this dissertation. First, to my dissertation committee of Dr. Mary Guy (chair), Dr. Danielle Varda, Dr. Jessica Sowa and Dr. Jason Glanz (mentor), thank you for your sage wisdom and advice through this process A multidisciplinary dissert ation committee has proven a very valuable asset in my development. Second, I would especially like to thank those who su pport my endeavors at work and school Komal Narwaney, Sophia Newcomer, and Nikki Wagner.

PAGE 6

vi TABLE OF CONTENTS CHAPTER I. REVIEW OF TH E LITEATURE ......................................................................................................... 1 Introduction to the Problem ............................................................................................................... 1 Purpose of the Dissertation ................................................................................................................ 3 Background on Vaccine Decision Making ..................................................................................... 4 Role of Public Policy in Vaccination ................................................................................................. 5 Vaccine Hesitancy ................................................................................................................................ 13 Decision Making about Vaccines .................................................................................................... 16 Interventions for Concerns and Hesitancy about V accines ................................................. 20 Research Questions ............................................................................................................................. 23 II. THEORIES AND FRAMEWORKS FOR HEALTH DECISION MAKING ........................... 24 Individual and Interpersonal Health Behavior Models ......................................................... 24 Concept Model for Vaccine Decision Making ............................................................................ 3 0 Hypotheses and Aims ......................................................................................................................... 32 III. DEVELOPMENT AND EVALUATION OF A MEASU REMENT INSTRUMENT: THE CONCERNS, ATTITUDES, BELIEFS AND INTENTIONS OF PARENTS ABOUT VACCINES FOR THEIR CHILD SURVEY ....................................................................................... 35 Reliability and Validity of Survey Instruments ........................................................................ 35 Methods for Phase One. Survey Development ......................................................................... 38 Results from Phase One. Survey Development ........................................................................ 43 Methods for Phase Two. Survey Evaluation .............................................................................. 47 Results from Phase Two. Survey Evaluation ............................................................................ 52 Methods for Phase Three. Predictive Validity .......................................................................... 63 Results from Phase Three. Predictive Validity ......................................................................... 66 Discussion ............................................................................................................................................... 77

PAGE 7

vii IV. DESCRIPTION OF PREGNANT MOTHERS VACCINE DECISI ON MAKING PROCESS OVER TIME ............................................................................................................................................. 84 Introduction ........................................................................................................................................... 84 Methods for Longitudinal Data Collection ................................................................................. 86 Results of the Longituinal Cohort .................................................................................................. 95 Summary of Longitudinal Findings ............................................................................................ 109 V COMPARING THREE MEASURES OF VACCINE HESITANCY ........................................ 114 Vaccine Hesitancy .............................................................................................................................. 114 Methods ................................................................................................................................................. 116 Resu lts .................................................................................................................................................... 121 Discussion ............................................................................................................................................. 128 VI CONCLUSIONS .............................................................................................................................. 131 REFERENCES ....................................................................................................................................... 135 APPENDIX A ................................ ................................ .......................................................................................................... 151 B ................................ ................................ .......................................................................................................... 153 C ................................ ................................ .......................................................................................................... 155 D ........................................................................................................................................................ 161

PAGE 8

viii LIST OF TABLES TABLE 2.1. Constructs, Definitions, and Application Examples of the Health Belief Model. 27 2.2. Constructs and Definitions of the Theory of Planned Behavior, 2008. ................. 28 2.3. Concepts and Defin itions of Social Network Theory .................................................... 29 2.4. Theoretical Constructs, Indirect Measures, Sub constructs, Definitions and Examples ........................................................................................................................................ 32 3.1. Terms and Definitions Related to Reliability and Validity of Sur vey Instruments .......................................................................................................................................................... 35 3.2. Methods Used to Determine Reliability and Validity ................................................... 39 3.3. Expert Panelist Expertis e and Institutional Affiliation ................................................ 41 3.4. Survey items (n=23), Response Scale, Theor etical Construct, and Source of Survey Item. .................................................................................................................................. 45 3.5. Interna l Consistency of Survey Items within Constructs ............................................ 55 3.6. Survey Items Removed after Ass essment of Internal Consistency ......................... 55 3.7. Survey Items Removed after Evaluation of Factor Structure.................................... 59 3.8. Factor Loadings From Principal Axis Factor Analsis With Direc t Oblim Orthogonal Rotation For A Four Factor Solution Fo r Baseline Survey Items (N=320) .......................................................................................................................................... 60 3.9. Internal Consistency of Survey Items within Constructs ............................................ 61 3.10. Results from Baseline and Second Survey Administered Two Weeks Later. .... 62 3.11. Characteristics of Pregnant Women and Parent Co hort by Vaccine Behavior, N=239 ............................................................................................................................................ 68 3.12. Univariate Analysis of Categorical Covariates from Pregnant Mother and Parent Cohort, N=239 .............................................................................................................. 70 3.13. Logistic Regression Analysis of Independent Modeling of Survey Constructs and Fac tors Associated with Undervaccination with Adjusted Odds Ratios ..... 72 3.14. Characteristics of Validation Study Co hort by Vaccine Behavior, N=374 ........... 73

PAGE 9

ix 3.15. Univariate Analysis of Categorical Covariates from Validation Study Cohort, N=374 ............................................................................................................................................ 75 3.16. Logistic Regression Analysis Survey Constructs and Factors Associated with Undervaccination with Adjusted Odds Ratios, N= 374 ............................................... 76 4.1. Chara cteristics of the Cohort, N=120 .................................................................................. 98 4.2. Means and Standard Deviations of Pilot Cohort, N=120. .......................................... 102 4.3. Trends over time of Pilot Cohor t Using CochranArmitage, N=120 ..................... 108 4.4. Trends over time of Pilot Cohor t Using CochranArmitage, N=120 ..................... 109 5.1. Characteristics of the Pilot Cohort Using PAC V Sc reener Grouping Variable, N=120 ............................................................................................................................................ 122 5.2. Characteristics of the Pilot Cohort Using S elf Reported Vaccine Intention Grouping Variable, N=120 ..................................................................................................... 122 5.3. Characteristics of the Cohort Using Average Days Undervaccinated Grouping Variable, N=80 ........................................................................................................................... 123 5.4. Measures of Vaccine H esitancy by Survey Items, N=120 ......................................... 124 5.5. Measures of Vaccine Hesitancy by Survey Items, N=239. ........................................ 127

PAGE 10

x LIST OF FIGURES FIGURE 2.1. Conceptual Model of Factors Contribu ting to Vaccine Decision Making .............. 31 2.2. Phases and Steps to Evaluating Validity and Reliability of Survey Instrument 34 3.1. Study Flow Diagram for Pilot Cohort .................................................................................. 53 3.2. Study Flow Diagram for Primary Cohort .......................................................................... 58 3.3. Study Flow Diagram of Primary Cohort Linking Mother to Child ........................... 53 4.1. Design of Prospective Survey Data Collection, N=120 ................................................. 87 4.2. Pregnancy Cohort Flow Diagram ......................................................................................... 97

PAGE 11

1 CHAPTER I REVIEW OF TH E LITERATURE Introduction to the Problem Vaccines are one of the most significant public health achievements in the past century (Centers for Disease Control and Prevention (CDC), 1999; 2011) As a result, routine childhood vaccination has led to the eradication, elimination or control of common infectious diseases such as small pox and polio (Plotkin & Plotkin, 2004) In the United States, mandates requiring childhood vaccines prior to entering public school has contributed to the control of infectious dis ease, and therefore decreased morbidity and mortality (Roush & Murphy, 2007) Despite the success of vaccin es and extensive research demonstrating the safety of vaccines (Baggs, et al., 2011; McNeil, et al., 2014) parents have numerous concerns. These con cerns include the ingredients in vaccines, the number and timing of vaccines, and short and long term health complications r esulting from vaccination (Freed, Clark, Butchart, Singer & Davis 2010) Although most parents follow the recommended childhood v accine schedule when vaccinating their children, c oncerns about vaccines have generated caution and hesitancy in parents. As a result, this hesitanc y has caused some parents to delay or decline certain or all vac cines for their child (Freed, et al 2010; Dempsey, Schaffer, Singer, Butchart, Davis & Freed, 2011) T here has been a recent surge of interest and activity in the research literature and mainstream media surrounding parents decision making about childhood vaccines These decision making factors

PAGE 12

2 include interpretation of scientific evidence about the safety and effectiveness of vaccines, the role of trust, influence of vaccine policies and mandates influence of media, social groups, cultural norms and belief systems ( Brown, et al 2012; Brunson, 2013; Glanz, et al 2013b; MacDonald, Smith, & Appleton, 2012) The public s trust in vaccine governance is weakening (Feudtner & Marcus, 2000; Larson, Cooper, Eskola, Katz, & Ratzan, 2011 ). In part, this is attributable to the very success of vaccine policies and program s designed to protect the publics safety from infectious disease. Due to the success of vaccines p arents are unlikely to experience vaccinepreventable diseases in their lifetime, and the perceived risk and seriousness of these diseas es diminish (Chen & Hibbs, 1998; Kennedy, Brown, & Gust, 2005). Subsequently p u blic concerns have changed from spread of disease to the safety of vaccines. As vaccine coverage declines in communities the risk for infectious disease outbreaks increases (F ine, 1993). There have been recent outbreaks in the United States that have contributed to the rapid spread of disease (Halsey & Salmon, 2015). C onsequently, infants who cannot vaccinate due to age, children who cannot receive vaccination due to medical co ntraindications and older adults whose immunity to disease has waned are atrisk for contracting a vaccinepreventable disease. Interventions are needed to facilitate change in health beliefs, attitudes, and concerns, and ultimately increase vaccination r ates. These interventions can be delivered through web based resources and decision aids, parentphysician communication approaches, inperson educational outreach, or other traditional behavioral intervention techniques. However, such interventions need to be

PAGE 13

3 rigorously evaluated for cost effectiveness and the ability to induce behavior change. In order to measure change in attitudes and beliefs towards vaccines, well developed, theory driven survey measures that can be implemented longitudinally are needed ( Kaufman, et al., 2013; Sadaf, Richards, Glanz, Salmon, & Omer, 2013). Purpose of the Dissertation The purpose of this dissertation was to develop, implement and evaluate a survey instrument to measure parents concerns, attitudes, beliefs, and intentions about vaccines for their child. Using Kaiser Permanente Colorados (KPCO) robust electronic health record (EHR ) and sel f reported information from KPCO members, an investigation was conducted that expanded on survey measurements in several ways. First, t he instrument was carefully constructed using theory, assessment from experts, and input from the target population to guide its development. Second, the survey was designed to measure change in attitudes and beliefs about vaccines over time. Third it was designed to be used in mul tiple interventions rather than for use in a particular study. Finally the survey underwent systematic reliability and validity evaluation. This chapter provides background information on vaccine policy, vaccine hesitancy, deci sion making risk perception and sources of influence and informationin the context of vaccines. Next, interventions for vaccine concerns and hesitancy are described. Finally, there is a brief discussion about why prospective measures of concerns, attitud es, beliefs and intentions about childhood vaccines are needed. The research question is discussed at the end of the chapter.

PAGE 14

4 Background on Vaccine Decision Making T here have been increasing vaccination rates overall in the United States C onversely there has been increasing exemption rates from vaccination (Omer et al 2006; Thompson, et al. 2007). An emerging public health concern is the deviation away from the recommended childhood vaccine schedule. Parents are increasing their use of delayed or alte rnative vaccination schedules, where parents selectively decide when and which vaccines will be administered to their child It is estimated that more than one in ten parents of young children use delayed vaccine schedules (Dempsey, et al 2011). These sc hedules are varied and do not follow consistent patterns (Glanz et al 2013a), which poses challenges for pediatricians as they communicate the safety of the childhood vaccine sc hedule to parents (Maglione, et al 2014). These challenges include lack of scientific research on the safety of alternative schedules, concerns about susceptibility to disease due to delaying or omitting vaccines, and unknown outcomes when deviations from the recommended schedule occur. For parents, the process of d ecision mak i ng about vaccination for their child has several important factors. These factors incl ude perceptions of risk of vac cination and risk of infectious disease, available vaccine information, and infl uence of sources of vaccine information such as family, internet, and media that ac t as personal networks. While it is well known that parents have concerns ab out the risks of vaccination, little is known of the prospective decision making process of pregnant women about vaccines for their child This prospective decision making process is important to the timing of when to implement practicebased

PAGE 15

5 interventions that strive to improve vaccination rates By describing the decis ion making process before and after the birth of the child information can be obtained to assist in optimal planning, timing, and information content in future research and practice. Role of Public Health Policy in Vaccination The health of the nation is essential to the social, economic and organizational growth of its people ( Shi & Singh, 2 011) Social determinants of health such as socioeconomic status, access to preventive health care (i.e. vaccines ) and availability to nutritional foods shape the health outcomes of individuals and communities (Marmot & Wilkinson, 2005) Public health po licy strives to achieve improvement of individual, group, and population health status. Examples of p ublic health p olicies include tobacco control policies, school nutrition policies, drug and alcohol laws, and vaccine mandates. Implementation of health po licies can be complex, time consuming, and have variable outcomes (Longest, 2002). Stakeholder buy in and ongoing support are necessary to the success of policy processes. Additionally, health policies that are flexible enough to accommodate shifts in societal perceptions about health issues are more successful in achieving long term sustainability (Pluye & Denis, 2004) Vaccine Policy and Protection of Community Health Va ccine policy is designed to mitigate disparities in health (Shonkoff, Boyce, & McEwen 2009) This policy design strives to assure population based protection against infectious disease in individuals and communities regardless of social

PAGE 16

6 factors that may impede individual health outcomes. However, vaccine policy does not guarantee coverag e against disease for all individuals and communities. Defining the problem within the context of the policy agenda is essential to understanding the causes of vaccine delay, how severe the problem is, consequences associated with delaying vaccines, and the incidence or scope of the problem (Rochefort & Cobb, 1994). Decision making about vaccination happens at the individual level Individuals can choose to comply with vaccine mandates that provide protection to the individual and the larger community or o pt out from vaccines, thus depending on others to protect them. Vaccine mandates were designed to ensure high coverage against vaccinepreventable diseases in the population (Cooper, Larson & Katz, 2008). History of Vaccine M andates The origins of the curr ent vaccine mandates in the United States are patterned from the laws related to smallpox disease control. Smallpox is an infectious disease that results in fluid filled pustules on the skin. The disease can have s evere complications such as scarring, blindness and encephalitis The disease also had a 35% mortality rate (Fenner, Henderson, Arita, Jezek, & Ladnyi, 1988) Initially, vaccination against smallpox was required when the disease was widespread. In th e United Kingdom in the mid nineteenth centur y, small pox vaccination of infants was publically mandated with a penalty of imprisonment of the parent if this mandate was not followed. This resulted in a change in the relationship between the state and its citizens Public mandates for smallpox vaccination caused

PAGE 17

7 protests against the states au thority over individual citizens decisions about vaccination. Eventually, law was enacted that allowed for exemption from vaccination through a conscientious objection statement (Durbach, 2000; Wolfe & Sharp, 2002). Shift from F ede ral to State A uthority in the United States In the nineteenth century compulsory vaccination was generally accepted. H owever, some believed the requirement interfer e d with human autonomy and liberty, invited unwarranted governmental interference, infringed on personal and religious b eliefs and induced medical safety concerns (Hodge & Gostin, 2002). Congress passed the Vaccine Act of 1813 to encourage vaccination against smallpox. It was the first federal program designed to improve the h ealth of the public. The goals of the program were to ensure a reliable source of smallpox vaccination, authori ze distribution to citizens and distribute the vaccine. With this ambitious public policy in place, the struggle to implement the program proved to be daunting After an incident of contaminated vaccine supply, the a ct was repealed and vaccine policy shifted from federal to states authority (Griffin, 2009). Compulsory School Vaccine Mandates Beginning in 1818, compulsory child hood vacci nation was introduced in England. N oncompliance with vaccination included annual financial penalties and not permitting an unvaccinated child entry into public school (Hodge & Gostin, 2002). In the United States, the first school entry vaccine mandate was enacted i n 1827. H owever, these mandates were not upheld until the smallpox epidemics that occurred in the mid 1890 s (Colgrove, 2006; Hodge & Gostin, 2002).

PAGE 18

8 Individual Autonomy versus Societal Obligations Vaccine policy represents a study in challenging tensions between societal interest and individual interest, much like Hardins The Tragedy of the Comm ons (Hardin, 1968; 1998). The common s in this policy example is a community with high vaccination rates, resulting in reduction in risk from infectious disease and a community relatively free from vaccine preventable disease. H erd immunity (when enough people are vaccinated in the community to stop transmission of the disease) of the community also transfer s to those who do not vaccinate due to medical, re ligious or other reasons. O ther reasons for not va ccinating include avoidance of the risk of adverse events from vaccines or the assertion of individual choice. In effect, the individual who does not vaccinate free rides on the risk of vaccinating taken by other s in the community Those who do not vaccinate are afforded the same herd protection from disease as those who vaccinate. However, there are community level immunity threshold levels to maintain in order to prevent significant outbreaks of infectious disease (Offit, 2011). At the individual level, the decision not to vaccinate increases the risk of infectious disease minimally due to community immunity prote ction provided to the individual. H owever, this individual decision not to vaccinate weakens the ove rall herd protection for the entire community. If too many individuals choose to do what is in their best individual interest, the common is at risk of depletion and will f ail to protect the community. Thus, infectious disease outbreaks will emerge. A de crease in community protection also puts vulnerable individuals who cannot receive vaccines and were previously protected by the community protectionat

PAGE 19

9 greater risk for infectious diseases (Malone & Hinman, 2003). This is a tragedy of the commons where herd immunity is depleted when individual interests prevail in order to avoid risk of adverse events following vaccination or to assert individ ual choice To avoid a tragedy of the commons, legal mandates have been imposed at the state level for certain v accines (Orenstein & Hinman, 1999). There is a risk of an unmanaged commons and potentially social disaster (Hardin, 1998) as mor e states allow for exemptions from va ccine requirement mandates. If this scenario occurs, the common is used up by self inter ested parents, or more appropriately described as concerned parents. As Hardin pointed out in his Tragedy of the Commons revision manuscript in 1998: Individualism is cherished because it produces freedom, but the gift is conditional: The more the population exceeds the carrying capacity of the environment, the more freedoms must be given up (p. 683) Dietz and colleagues (2003) propose a governance framework for managing the commons in modern society This framework stresses the importance of critical and ongoing informed dialogue among interested parties Additionally, the framework encourages engagement as a way to deal with conflicts, discussion of social norms and compliance, and adaptation and change to maximize the commons. Although this gover n ance framework has valuable approaches to dealing with the

PAGE 20

10 tragedy of the commons it is much more difficult to put into practice with a highly charged topic such as vaccines Vaccination of children by their parents has been likened to a social contract (Hodg e & Gostin, 2002; Rousseau 1920) where individual decision making about vaccines benefits the social order of the community. The act of vaccination has duality of benefits as it protects the individual child against infectious disease and contributes to the protection of the publics health from infectiou s disease outbreaks Vaccination by most individuals in the community protect the medically vulnerable who cannot vaccinate, those who do not vaccinate by choice and those who are un aware they are unprotected when the vaccine is not effective on the individual level. However, at the individual level, parents are concerned about the safety of vaccines and perceived risks asso ciated with vaccination (Freed, et al., 2010; Kennedy, LaVail, Nowak, Basket, & La ndry, 2011). These tensions frame the challenges of implementation and sustainability of vaccine policy today. Current Vaccine Mandates In the United States, death from common childhood infectious disease has been essentially eliminated including polio and rubella ( Roush & Murphy, 2007). For example, there has been greater than 98% decline in the incidence of vaccine preventable diseases such as polio, mumps, Haemophilus influenzae type b, and meas le s since vaccines for these diseases became available (O renstein, Douglas, Rodewald, & Hinman, 2005; Roush & Murphy, 2007). This has been successful, in part, due to the implementation of vaccine policy in the United States (Orenstein & Hinman, 1999; Orenstein, et al 2005; Pickering &

PAGE 21

11 Orenstein, 2002). In 19 77, the American Academy of Pediatrics (AAP) made vaccine policy a top priority, calling for universal vaccination (American Academy of Pediatrics (AAP) 1977; Wood, 2003). AAP strongly supported enactment of public policy mandates to require up to date vaccinatio ns of all children entering public school (National Vaccine Advisory Committee (NVAC) 1999; NVAC, 2013). Exemption from Vaccines States determine the requirements for school mandated vaccination. These requirements include the documentation necess ary to enter public school, the documentation required to show evidence of vaccination, and conditions under which parents can exempt their children from vaccination (Diekema, 2014). There are three main categories of exemption from vaccination: medical ex emption, religious exemption and personal or philosophical exemption. Not all states allow rel igious or personal exemptions. Currently, three states (California, Mississippi, and West Virginia) do not recognize religious or personal exemptions. Nineteen st ates permit personal and religious exemptions A ll states permit medical exemption from vaccination (Sandstrom, accessed 9 22 2015) States rely on the recommended immunization schedule determined by the Advisory Committee on Immunization Practices (ACIP) to determine individual s tate vaccination requirements. ACIP is a group of fifteen voting medical and p ublic health experts who determine the recommended doses and use of vaccines to control infectious disease in the United States (Smith, 2010). Mandates that permit personal exemption (also known as ph ilosophical exemption) from school required vaccination vary in complexity from state to state.

PAGE 22

12 However, the general format includes parents acknowledgement of a personal objection to vaccination (Rota, Salm on, Rodewald, Chen, Hibbs, & Gangarosa, 2001; Blank, Caplan, & Constable, 2013). There has been an overall increase in statelevel rates of personal exemptions. E ase of exemption from schoolrequired vaccination has been shown to influence rates of underv accination. Personal exemptions rose from an average of 0. 99 percent in 1999 to 2.54 percent in 2004 in states that have this policy option (Omer, et al 2006; Thompson, et al 2007). For example, in the state of Arkansas, the number of personal exemptio ns increased from 529 in 2001 to 1,145 in 2004, following a vaccine policy change that permitted personal exemption from required school entry vaccination (Salmon, et al ., 2006; Thompson, et al 2007). Omer et al (2012) recently updated their landmark r esearch in vaccine policies and rates of exemption from vaccines. They found that in 2011, nonmedical exemption rates in states that had relatively easy exemption policies were higher than in states with difficult exemption policies. Geographic Location a nd Vaccine Exemptions R ates Geographic clustering of exemptions from vaccines occurs within states, which may reach upwards of one quarter to onethird of the community. In the state of Washington in 2007, county level vaccine exemption rates ranged from 1 .2 to 26.9% (Omer, et al. 2008). There are wide variations in exemption from vaccination rates across states. In the 20142015 school year, Louisiana had less than 0.6% personal exemption rate from vaccination among kindergartners while Vermont had an ex emption rate

PAGE 23

13 of 5.9% ( Seither, et al., 2015 ). Furthermore, these rates are based on ages of children at least five years old, and does not account for rates of undervaccination of children younger than five years of age These children may have been on alt ernative vaccination schedules and have subsequently caught up with the age appropriate recommended childhood vaccine schedule. There is growing concern among public health officials, policymakers, school officials, parent groups, and medical providers that the upward trend in personal exemptions from vaccination may contribute to the reemergence and rise of otherwise populationcontrolled infectious diseases (Chen & Hibbs, 1998; Lantos, et al. 2010). In fact, there have been recent measles and pertuss is outbreaks linked to children of parents who did not vaccinate their child (Atwell, et al., 2013; Halsey & Salmon, 2015) Vaccine Hesitancy The International Vaccine Hesitancy Working Group defines vaccine hesitancy as delay in acceptance or refusal of vaccines despite availability of vaccine services. Vaccine hesitancy is complex and context specific, varying across time, place and vaccines. It is influenced by factors such as complacency, convenience and confidence (WHO SAGE Working Group, 2014, p. 7) Vaccine Refusal and D elay Some studies have found that parents who choose not to vaccinate their children differ demographically from parents who vaccinate their children. They tend to be older and have higher levels of education (Gust, et al., 2008). C onversely, the national rates of vaccination for children under the age of three are lower for

PAGE 24

14 those living in poverty. V accination rates for most series are lower for children who are black (Elam Evans, et al., 2014). It is important to clarify that there are distinct differences between parents who chose not to vaccinate their child and those who experience logistic barriers such as transportation. This current investigation focuses on undervaccination by parental choice. Studies using individual level d ata have shown that children of parents who declined vaccination were approximately twenty three times more likely to acquire pertussis (Glanz, et al., 2009) and nine times more likely to contract varicella than vaccinated children (Glanz, et al., 2010). W hen compared to parents who accept vaccines, parents who decline vaccines are more likely to believe their children are not at risk for vaccinepreventable diseases and that vaccinepreventable diseases are not dangerous (Salmon, et al., 2005; Smith, et al ., 2011). One in ten parents delay one or more vaccines for their child (Dempsey, et al 2011; Offit & Moser, 2009). There is sparse scientific information available on the safety of delaying vaccines. T here is some scientific indication that delaying one childhood vaccine poses greater risks. Hambidge and colleagues (2014) found that delaying the first dose of measles, mumps and rubella ( MMR ) vaccine past fifteen months of age res ulted in higher risk of seizures. While this provides some evidence that dev iation from the recommended childhood vaccine s chedule poses increased ad verse events to the individual at least for one vaccine, it also poses an interesting dilemma. If a parent refuses any MMR vaccine, their child has a decrea sed risk of seizure and an increased risk of the disease of measles.

PAGE 25

15 Vaccine C oncerns and Trust Although scientific evidence has refuted many of the misconceptions regarding vaccine safety (see DeStefano, 2007; Thompson, et al 2007; Price, Thompson, Goodson, & Weintraub, et al 2010), there is continuing mistrust in mainstream media sources as well as misunderstanding of scientific information. General concerns of parents regarding vaccination include a perceived causal linkage between receipt of the measles, mumps, and rubella vaccine and neurological disorder (Smith, et al 2008); autism (Thompson, et al 2010; Freed, 2010); parents worries that the human papillomavirus vaccine may lead to sexual promiscuity (Dempsey, Zimet, Davis, & Koutsky, 2006); fears regarding the manu facturing processes and ingredients in vaccines (Benin, Wisler Scher, Colson, Shapiro, & Holmboe, 2006); and concerns that too many vaccines overload or weaken young childrens immune systems (Benin, et al 2006). Va ccine safety concerns may reduce parents willingness to vaccinate their children. This is particularly true for new vaccines that are added to the childhood vaccine schedule (Freed, et al 2010). Several studies have examined the role of trust in vaccine decision making. Benin and colleagues (2006) conducted qualitative, open ended interviews with new mothers (one to three days postpartum) and then again in three to six months time. The theme of trust was an important factor in parents decisionmaking about vaccination of their infants Th ose who vaccinated their children had more t rust in their physician, wanted to follow social norms to vaccinate, a nd wanted to protect their child and others from infectious disease through vaccination. Mothers who did

PAGE 26

16 not vaccinate their child verbalized dis trust in their physician, had a trusting relationship with a complementary medicine provider or others who endorsed their beliefs about not vaccinating their child were worried about vaccine adverse events, and believed that vaccine preventable diseases are not serious. The parents also considered that others vaccinate and therefore their child is not at risk for a vaccine preventable disease (Benin, et al 2006). Other studies have found similar findings in relation to trust and their childs provide r. Parents who refused and delay vaccines are more likely to have low levels of trust in their childs pediatrician than parents who accepted vaccines (Glanz, et al., 2013b). Information from health care professions, health departments and governmental agencies are viewed as less trustworthy by parents who do not vaccinate their child (Kennedy, et al., 2005; Salmon, et al., 2005). Studies have shown that trust in their child s pediatrician influence their decision to vaccinate their child in those who previ ously deviated or considered deviating from the recommended c hildhood vaccine schedule (McCauley, K ennedy, Basket, & Sheedy, 2012) From these published studies, it appears that their childs pediatrician can mitigate parents vaccine concerns, to some ext ent. Decision Making about Vaccines With the increasing trend of parents who delay some or all vaccines for their children, research has focused on how parents approach this decision and what factors are associated with delay (Omer, et al., 2006; Robison, Groom, & Young, 2012; Omer, et al., 2012). These parents are often concerned that childr en receive too many vaccines over a short period of time (Salmon, et al., 2005). Consequently,

PAGE 27

17 parents request vaccination schedules that both increase the time between vaccinations and reduce the number of vaccine do ses in a single medical visit. S tudies suggest that children with delayed vaccines may have different healthcare utilization patterns than vaccinated children. Glanz and colleagues (2013a) found that unvaccinated children of parents who refused vaccines were 50% less likely to have a clinic visit for an upper respiratory infection compared to fully vaccinated children (Glanz, et al 2009). Furthermore, in a matched cohort analysis, it was found that underva ccinated children had lower outpatient visit rates and higher inpatient admission rates compared to ageappropriately vaccinated children (Glanz, et al 2013a). Other research has shown that children who do not receive some or all vaccines for personal or medical reasons had 12% fewer clinic visits than children who were up to date (Wei, Mullooly, Goodman et al 2009). Risk Perceptions A shift occurred from concerns of infectious disease to concerns about adverse events associated with receipt of vaccines. This was fueled by the proliferation of media outlets recounting personal, emotional, and affective stories of parents whose children were reported as harmed by vaccines (See Gonzalez, 1982; Offit, 2011). Kennedy and colleagues (2011) used the 2010 He althStyles data which surveyed parents of children six years of age or younger. Parents endorsed conc erns about the pain of vaccines during a do ctors visit most frequently ( 38% ) followed by concerns that their child received too many vaccines in one doctors visit (36%).

PAGE 28

18 Additional concerns endorsed by at least 25% of the parents included th e number of vaccines for children, symptoms of fever after vaccination, concerns that vaccines may caus e autism, and concerns that vaccine i ngredients are unsafe. In a study of 1500 parents of young infants, parents knowledge, attitudes, and beliefs were stratified by vaccine decision. Parents stated that vaccine side effects were of highest concern (McCauley, et al 2012). These findings confirm previous studies (Fr eed, et al 2010). Weinstein, et al (2007) prospectively measured beliefs about risk as a predictor to influenza vaccination in college students. Anticipated regret of getting influenza was the strongest predictor of getting the influenza vaccine. Additionally, it was found that risk perceptions framed as feelings predicted the relationship between risk and vaccine better than rational statements (Weinstein, et al 2007). Finally, Brewer, et al (2007) conducted a metaanalysis o f vaccine risk perceptions in thirty four studies with 15,988 data points finding consistent relationships between risk perceptions and behavior. They found linear relationships in risk likelihood (r=.26), susceptibility (r=.24), and severity (r=.16) as significantly predicting v accination behavior (Brewer, et al. 2007). Research in the area of risk perception and health behaviors is r elatively undecided. A majority of empirical research does find positive associations between perception of risk and behaviors (See Brewer, Weinstein, Cuite, & Herrington, 2004; Weinstein, et al., 2007; Wroe, et al 2005; Wroe, et al., 2004 as examples). Yet, individual studies report a range of outcomes, relationships and effect sizes.

PAGE 29

19 Sources of Influence Parents access information about vaccines t hrough a wide variety of ways including internet, books, social media, personal contacts (neighbors, family, and groups ), physician or complimentary health provider and television as some common examples. If t he attitudes towards vaccination are similar between pediatrician and parent the pediatrician can potentially contribute to parents decisions about vaccination (Mergler, et al 2013). Furthermore, Opel and colleagues (2013a) found that when physicians use a presumptive narrative that assumes pare nts will vaccinate, the compliance with vaccination increased compared to a participatory narrative that encouraged discussio n about the decision. This new area of research investigates provider influences on parent decision making. Often, media sources and personal contacts contradict one another, which can lead to conflicting and erroneous information. Media can have both positive and negative effects on parents decisionmaking about vaccines for their children (Smith, et al 2008; Nyhan, Reifler, Rich ey & Freed, 2014). An increasing number of adults use the internet to obtain health care information. Recent Pew survey data show that nearly 81% of adults in the United States are regular internet users, 72% of those regular internet users looked online for health information in the past year (Fox & Duggan, 2013). Half of online health and medical information searches are related to someone elses health (Fox & Duggan, 2013). Although most parents identify physicians as their most trusted source for hea lth information regarding their child (Freed, et al., 2010), parents use

PAGE 30

20 the internet as a source of information about vaccine over their childs physician to guide their decision ( Benin, 2006; Downs, Bruine de Bruin, & Fischoff, 2008). Additionally, Downs (2008) and others have found that parents lack basic knowledge about how vaccines work They do not find the vacc ine information provided by their childs pediatrician very helpful or to be enough (see also Gellin, Mailbach, & Marcuse, 2000; Gust, et al. 2005; Benin, et al 2006). These parents were less confident in the safety of vaccines and endorsed communication barriers with their childs pediatrician. These sources of influence and deficits in knowledge add to the complex ity of decisionmaking In terventions for Concerns and Hesitancy about Vaccines Given the sweeping changes in the ways in which people receive and retrieve health information (e.g. internet, social media), it is imperative to design interventions that meet the needs of individuals and provide flexible frameworks from which evidencebased health strategies can be implemented. A promising approach to disseminating health information to individuals is the use of interactive health information technologies. Health technology offers a sc alable platform that can help improve the quality, cost effectiveness, capacity and efficiency in the health system (Bennett & Glasgow, 2009), and specifically vaccine decision making. However, interventions using technology need to be rigorously evaluated for effectiveness (Shoup, et al. 2015). Several interventions designed to improve vaccination rates have been conducted. In a study by Wroe, parents who received a decision aid for vaccination had significant increases in ontime vaccination of their ch ild, decreased perceptions

PAGE 31

21 of risk of vaccines, increased perception of risk of vaccine preventable diseases, reduced anxiety and increased satisfaction ten weeks after the birth of their child as compared to those receiving usual care (Wroe, et al 2005). In another study, researchers implemented a web bas ed MMR decision aid that significantly increased parents knowledge over time and decreased conflict over the MMR vaccine decision. Most parents decided to vaccinate their child for MMR however the sample size was small (Jackson, Cheater, Peacock, Leask, & Trevena, 2010). From the same study team, a randomized control trial of 142 study participants where the intervention arm was an informational pamphlet on MMR and a parentled group meeting, and the control group was pamphlet only Those in the intervention arm had higher vaccination rates for MMR ; however, decisional conflict remained the same (Jackson, et al. 2011). There is a need to know when the optimal timing is to implement interventions aimed to increase vaccines and reduce vaccinerelated concerns. There is sparse evidence as to how and when to intervene (Sadaf, et al., 2013). Further knowledge in the area of decision making and risk perception regarding vaccination has strong implications f or public policy, as vast amounts of time and resources are invested in our public infrastructure in planning for disease pandemics (Miller, Viboud, Blalinska, & Simonsen, 2009). Clearly, there is a need for interventions designed for parents to increase v accination rates and decrease vaccine concerns. However, there is a parallel need for well designed theory driven and reliable measurement instruments to assess the subjective outcomes of interventions.

PAGE 32

22 Need for Prospective Measures Although there are many survey instruments in the literature that measure vaccine concerns, most have not been developed for longitudinal use or evaluated for association with vaccine behavior. Opel (2011a b; 2013b) has designed a va lidated survey that is a vaccine hesitancy s creener administered to parents in the clinical setting (Opel, et al 2013b). It measures parents intentions to vaccinate and quantifies a score of hesitancy about vaccinating. This research holds promise in assisting medical personnel to tailor information and messages in near real time to parents who are concerned about vaccines or are hesitant about vaccinating their child It also provides a measure to predict future vaccine decisions and potentially mediate the potential outcome of vaccine hesitancy through info rmation, education, and emotional element of vaccine information and decision making that parents experience. In summary, there are three major gaps in the literature when considering the measurement of parentassociated vaccine concerns and behaviors: validated measur es specific to vaccine concerns attitudes, behaviors and intentions; measurement of these domains prospectively and measure change over time in these domains ; and use of objective measures to quantify the association with the su rvey measures The literature reviewed in this chapter included vaccine mandates, rates of exemption from mandates, and parents hesitancy about vaccines for their child. Specifically, the chapter ide ntified factors that influence parents decision making about vaccines. First, parents have numerous concerns about the safety of vaccines.

PAGE 33

23 These include concerns that vaccines cause negative shortand long term health conditions such as autism. Second, sources of information that parents access about vaccines are influential. However, these sources differ for parents who have concerns about vaccines. Physicians are the most influential sources for parents who decide to vaccinate their child, but to a lesser extent for parents who decide to delay or not vaccinate their child. Finally, perceptions of risk of disease and risk of vaccination were reviewed. Vaccine behavior is associated with perceived risk of getting a disease or experiencing an adverse event from vaccination. Vaccine behavior is also associated with how severe the disease or adverse event might be. Next, the investigation research question is presented. Research Question s The research question focuses the dissertation on the development and evaluation of a survey instrument to measure parents cha nges in concerns, attitudes, beliefs and intentions about vaccines for their child. Research Question 1: How can changes in attitudes and beliefs about childhood vaccines be measured? Research Question 2 : What vaccine d ecisionmaking factors, as measured subjectively (by survey items) in a cohort of parents, are associated w ith undervaccination behavior? Next, in chapter two, the theories that guided the design and the developme nt of a survey instrument that measures the factors identified in the literature review are presented For this investigation, the three major theories

PAGE 34

24 applied were Health Belief Model (HBM), Theory of Planned Behavior (TPB) and Social Network Theory (SNT). HBM and TPB guided the development of the survey constructs, while SNT guid ed the addition of descriptive questions.

PAGE 35

25 CHAPTER II THEORIES AND FRAMEWORKS FOR HEALTH DECISION MAKING Traditional theories of decisionmaking have focused on the rational actor, bounded rationality, public choice anal ysis, deliberative democracy and discourse theory, organizational influence, and policy networks, to name a few (Simon 1959; Kingdon & Thurber, 1984; Koppenjan & Kliji, 2004; March, 1994; Miller, 1982 ) These theories range in unit of analysis from the individual actor to broad social groups. More recently, theories and models of health behavior have emerged, drawing upon several disciplines such as sociology, psychology, epidemiology and biological sciences. These theories help researchers guide the examination of reasons why people engage in particular health behaviors or not, assist in planning and development of public health programs, and contribute strategies to health communication planning. Most significantly, health behavior theories have focused o n effective health education and health communication strategies that lead to recommended behavior change (Glanz, Rimer, & Viswanath, 2008). Policy makers, health care institutions, parents, and schools h ave identified vaccine decisionmaking as an important public health issue over the last decade. In order for vaccine behavior change to occur in an effective manner, methods should be designed with emphasis on the individual and their social characteristics, beliefs, norms and their environment. Individua l and Interpersonal Health Behavior Models The current study uses several theoretical perspectives to create a conceptual model. The model integrates individual and interpersonal factors that

PAGE 36

26 are theorized to contribute to parents decision making regarding vaccines for their child. Below, the three health behavior models used for this investigation are described. Health Belief Model (HBM) The Health Belief Model (HBM) is one of the most widely applied health behavior conceptual framework used to explain b oth behavior change and maintenance of health related behaviors, and to guide health behavior interventions (Champion & Skinner, 2008). It was developed in the 1950s by social psychologists in the United States Public Health Service to explain poor partic ipation by the public in health prevention programs. Rosenstock and colleagues (1959) refined the factors by conducting a systematic review of factors on why people did or did not accept polio vaccination. The main factors of the HBM emerged through this research (Rosenstock, Derryberry, & Car riger, 1959). The model was expanded upon to include the study of symptoms and behaviors related to illness (Conner & Norman, 2005). HBMs central focus is health motivation, which makes it amenable to addressing behaviors that induce health c oncerns. HBM assumes that the individual can make their own decisions and to act upon those decisions. The key aspect of the model is that an individual will take a health related action (such as vaccination) if there is the belief and expectation that a negative health co ndition can be avoided and the recommended health action can be acted upon (Champion & Skinner, 2008). Six constructs, as described in Table 2.1 contribute to HBM as key elements in predicting why people act to p revent, screen for, or control illness. These include

PAGE 37

27 susceptibility, seriousness, benefits and barriers to a behavior, cues to action, and self efficacy. Table 2. 1 Constructs, Definitions, and Application Examples of the Health Be lief Modela. Construct Definition Application Perceived susceptibility Belief about the chances of experiencing a risk or getting a condition or disease Define populations at risk Individual risk based on behaviors Perceived severity Belief about how s erious a condition and its sequelae are Specify consequences of risks and conditions Perceived benefits Belief in efficacy of the advised action to reduce risk or seriousness of impact Define the action to take and positive effects that are expected Per ceived barriers Belief about the tangible and psychologi cal costs of the advised action Identify and reduce perceived barriers through correction of misinformation or incentives Cues to action Strategies to activate readiness Provide how to informatio n, promote awareness ; reminders to health behaviors Self efficacy Confidence in ones ability to take action Provide training in recommended action, reinforcement, reduce anxiety aGlanz, et al., 2008 A significant limitation of HBM is its exclusive foc us on the individual without taking into consideration the environmental context in which the individual resides. Without a broader context of the individual, there is failure to consider other influences such as social netw orks. In addition, HBM is a cogn itivebased model and does not consider the affective or emotional components to behavior. Given the recent advances in understanding the importance of emotion and risk perception (Slovic, 2010), this represents a gap in the model that should be added.

PAGE 38

28 Th eory of Planned Behavior (TPB) The Theory of Planned Behavior (TPB) is another individual level health model, which focuses on individual motivating factors that act as determinants to performing an intended behavior. TPB is an extension of the Theory of R easoned Action (TRA), adding perceived control to the models constructs. The underlying tenet of the theory is its assumption that behavioral intention (motivation) determines behavior. Therefore, it is also important to determine the extent of control th e person has over the behavior (Montano & Kasprzyk, 2008). Three main constructs contribute to the TPB as key elements in explaining how behavioral intention determines behavior. These constructs are attitude towards behavior, subjective norm, and perceiv ed behavioral control as influencing behavioral intention. Table 2.2 provides the d efinitions of the constructs. Table 2 .2 Constructs and Definitions of the Theory of Planned Behavior, 2008a. Construct Definition Behavioral inten tion Perceived likelihood of performing the behavior Attitude Overall affective evaluation of the behavior Subjective norm Belief about whether most people approve or disapprove of the behavior -and an individuals motivation to comply with expectation s of others Perceived control Overall measure of perceived control over behavior likelihood and effect of a facilitating or constraining condition aGlanz, et al., 2008 However, there is conflicting views in the literature regarding TPBs intentions as a predictor of behavior (see Sheeran, Trafimow, Finaly, & Norman, 2002; Webb & Sherran, 2004). Some studies demonstrate a relationship while others fail to find a relationship.

PAGE 39

29 Social Network Theory (SNT) The Social Network Theory (SN T ), an interpersonal c onceptual framework, has been described as a powerful influence on health. The concept of SN T emerged in the 1950s, through the work of Barnes (1954; 1969). From early work in the field, it was found that closeknit or homogenous networks exchange more af fective, emotional support and provide more social influence on members to conform to social norms (Heaney & Israel, 2008). SN T as a conceptual framework illustrates the influence of social relationships on health status, health behaviors, and health decis ion making. There are three concepts of SN T; each concept has sub concepts and definitions (Heaney & Israel, 2008). Table 2.3 provides the c oncepts and definitions of SN. Table 2. 3 Concepts and Definitions of Social Network Theor ya. Concepts Definitions Structural characteristics of social networks: Reciprocity Extent to which resources and support are both given and received in a relationship Intensity or strength Extent to which social relationships offer emotional closeness Complexity Extent to which social relationships serve many functions Formality Extent to which social relationships exist in the context of organizational or institutional roles Density Extent to which network members know and interact with each oth er Homogeneity Extent to which network members are demographically similar Geographic dispersion Extent to which network members live in close proximity to focal person Directionality Extent to which members of the dyad share equal power and influenc e Functions of social networks: Social capital Resources characterized by norms of reciprocity and social trust

PAGE 40

30 Concepts Definitions Social influence Process by which thoughts and actions are changed by actions of others Social undermining Process by which others expres s negative affect or criticism or hinder ones attainment of goals Companionship Sharing leisure of other activities with network members Social support Aid and assistance exchanged through social relationships and interpersonal transactions Types of social support: Emotional support Expressions of empathy, love, trust and caring Instrumental support Tangible aid and services Informational support Advice, suggestions and information Appraisal support Information that is useful for self evaluat ion aGlanz, et al., 2008 SN T lacks empirical evidence on how to enhance social networks and communication exchange within a network. This culminates in the question of who should provide what to whom and when (Heaney & Israel, p. 207, 2008). However, S N T offers both the broader environment and affective components in relation to behavior changes. Concept Model for Vaccine Decision Making The cu rrent investigation uses components of three theories (Health Belief Model, Theory of Planned Behavior, and S ocial Network Theory) t o create a conceptual model about vaccine decisionmaking that integrates individual and social factors Figure 2. 1 depicts the conceptual model designed for this study based on the literature review. For each model construct, Table 2.4 provides a definition and example. In this model, t he linear pathway of the concept model implies that

PAGE 41

31 attitudes and beliefs about vaccines lead to formation of intentions about vaccinating. These intentions influence vacci ne behavior (vaccinating or not vaccinating on time). In this study the constructs from HBM and TPB are used to develop and evaluate a survey instrument. In the concept model, the following conceptual definitions are used : beliefs about perceived susceptibility and severity contr ibutes to Beliefs about Vaccinating and beliefs about effectiveness and costs of taking action contributes to Evaluation of VPD/VAE. For TPB, the following constructs are used: Subjective Norms about Vaccinating, Perceived Control of Vaccination Decision, and Intentions about Vaccinating. Social Network Theory concepts a re used for descriptive purposes These concepts are Social Influence and Informational Support. The SNT constructs are used for descriptive questions. These are: Functions of Social Network s and Types of Social Networks. Figure 2. 1. Conceptual Model of Factors Contributing to Vaccine Decision Making

PAGE 42

32 Table 2. 4 Theoretical Constructs, Indirect Measures, Sub cons tru cts, Definitions aHBM = Health Belief Model bTPB = Theor y of Planned Behavior cVPD = V accine preventable disease dVAE = V accine adverse event Hypotheses and Aims To answer the questions, h ow can changes in attitudes and beliefs about childhood vaccines be measured and what vaccine decisionmaking factors, as m easured subjectively (by survey items) in a cohort of parents, are associated with undervaccination behavior, a study was conducted with pregnant women and parents of children under the age of twelve months The hypothese s and aims are presented below. Hy potheses 1) Survey constructs developed through this investigation (Beliefs about Vaccinating, Evaluation of VPD/VAE, Perceived Control of the Vaccine Decision, and Subjective Norms about Vaccination) are associated with objective vaccine behavior (vaccinated or undervaccinated) Model Construct/ Predictors of Intention Theoretical Model from sub-construct (HBM a /TPB b) Definition Beliefs about Vaccinating HBM /TPB This measures the parents' perception of effectiveness of vaccines or perception of physical and psych ological costs of vaccination. Evaluation of VPD c /VAE d HBM /TPB This measures parents' beliefs about whether or not their children are at risk for or seriousness of VPD/VAE Perceived Control of Vaccinating Decision TPB This measures the parents' belief in their ability to overcome barriers to vaccination and perceptions of difficulty or ease in vaccination. Subjective Norms about Vaccinating TPB Beliefs or motivation of parent about complying with social norms

PAGE 43

33 2) Concerns, attitudes, beliefs and intentions of parents about vaccines for their child change over time from pregnancy to six months after birth of the child. Aims 1) Develop a longitudinal survey instrument that measures parents concer ns, attitudes, beliefs and intentions about vaccines for their child. 2) Administer the survey instrument on two cohorts of parents. a. Cohort of pregnant women over time b. Cohort of pregnant mothers and parents of children under twelve months of age from an integ rated health system KPCO 3) Evaluate the survey instrument for reliability and validity. a. Conduct exploratory factor analysis (EFA) to determine the survey items that measure latent variables and develop factors b. Calculate Cronbachs alpha on survey items identified in EFA to evaluate internal consistency of the related group of items c. Conduct testretest reliability to measure stability of the instrument over time 4) Estimate the association between subjective survey constructs and objective vaccine behavior (vac cinated or undervaccinated) 5) Compare survey variables at four different time points to evaluate changes over time. 6) Compare different measures of vaccine hesitancy using cohort data.

PAGE 44

34 T here are three phases to the investigation. First, the methods and results of the development of the survey instrument are presented in phase one. This is followed by phase two, the methods and results of the evaluati on of the survey instrument. Finally, in phase three, the methods and results of testing the association between the survey constructs with undervaccination are presented. These phases are discussed in Chapter III. Figure 2.2 shows the layout of each of the subsequent chapters. Chapter III presents phases one, two and three. Chapter IV presents the longitudinal anal yses of each survey item from the pilot survey. Chapter V presents comparison of different measures of hesitancy. Figure 2.2. Phases and Steps to Evaluating Validity and Reliability of Survey Instrument

PAGE 45

35 CHAPTER III DEVELOPMEN T AND EVALUATION OF A MEASUREMENT INSTRUMENT: THE CONCERNS, ATTITUDES, BELIEFS AND INTENTIONS OF PARENTS ABOUT VACCINES FOR THEIR CHILD SURVEY In this chapter, the process of developing, evaluating, and validating a theory driven survey instrument that c an be used to measure the effectiveness of interventions targeting vaccine concerns of parents is described First, the process of developing and piloting the survey instrument is described Next, the results of the process are presented Then, the methods and results from evaluating the validity and reliability of the instrument on a cohort of parents are described. Finally, methods and analytic results of how the survey constructs ar e associated with vaccine behavior are shown Reliability and Validity o f Survey Instrument Comprehensive development of a survey instrument requires assessment of r eliability and validity from pilot ing through finalization of the instrument. In order to establish common definitions and provide a clear understanding of the methods and results, a table of basic definitions was created (Table 3.1) Table 3.1. Terms and Definitions Related to Reliability and Validity of Survey I nstrumentsa Term Definition Reliability Survey instrument performs in consistent, predictable ways; sc ores represent a true meaning of the survey item. Internal consistency reliability Measures the strength of relationship betwe en survey items in a scale. I tems should be highly correlated. Cronbachs Alpha is a measure of reliability.

PAGE 46

36 Term Definition Test re test reli ability Test retest reliability is a measure of consistency and stability of the survey instrument over time. It is assessed by administer ing the same test to the same sample on two different occasions. There should be no substantial difference in response s between the two time points. Validity Assesses whether the survey items predict specific events or its relationship to measures of other constructs. Face validity Assesses whether, on surface, the survey items appear to represent the survey construct. Content validity Assessed by recognized subject matter experts to evaluate whether survey items define content. Criterion validity Assesses how well one instrument stacks up against another instrument or predictor. Predictive validity Assesses the su r vey instruments ability to predict an outcome based on survey scores Concurrent validity A ssesses the new instrument against a gold standard Construct validity Assesses whether the survey items truly reflects and measures the construct b when it is put into real world practical use. Convergent validity Assesses different methods for obtaining the same information about a trait or concept and the results are similar aLitwin, 1995; Fowler, 2013 bA construct is an abstract idea, theme or subject matter measured by survey items. It is also called a latent variable in some literature (DeVellis, 2012). Objectives of the Study The primary objective of this stud y was to design a theory driven survey instrument to measure parents concerns, attitudes, beliefs and intentions about childhood va ccines. Additional aims include administration of the survey to two cohort s. The P ilot cohort was pregnant mothers exclusively and the P rimary cohort was pregnant mothers and parents of children under twelve months of age F actor analyses and reliability testing were used to evaluate the survey. Theory Used to Develop the Survey The survey instrument was developed by primarily using the theoretical constructs and definitions from Theory of Planned Behavior (TPB) and Health Belief Model (HBM), two health behavior theories that have been previously used in health services research to explain vaccine decision making and behavior. Social Network

PAGE 47

37 Theory (SNT) was used to guide the development of descriptive survey items that were not used in determining the sur v ey constructs. Two constructs used conceptual definitions from the HBM (Champion & Skinner, 2008) for development of the survey instrument. These c onstructs were Beliefs about Vaccinating, and Evaluation of Vaccine Preventable D iseases (VPD) and V ac cine Adverse Events (VAE) Beliefs about V accinating explain the benef its and barriers of vaccinating It measures the parents perceived psychological and physical costs associated with vaccinating their child and parents perc eptions of the effectiveness of vaccination. Ev aluation of VPD and VAE explain a parents belief about how serious a parent considers the disease or potential adverse events after vaccinating. It measures parents belief s about ho w likely it is to contract a disease or experience a vaccine adverse ev ent (Montano & Kasprzyk, 2008) An additional two constructs from TPB (Glanz, 2008) were included. These constructs were Perceived Control of Vaccinating Decision and Subjective Norms about Vaccinating. Perceived Control of Vaccinating Decision explains the parents belief in their ability to overcome barriers to vaccination. It measures parents perceptions of difficulty o r ease in vaccination. Subjective Norms about Vaccinating explains parents beliefs or motivation about vaccines. It measures parents compliance with social norms about vaccinations. Together, these four constructs provided the theoretical infrastructure to develop a measurement instrument to pilot on pregnant mothers about their concerns, attitudes, beliefs and intentions about vaccines for their child.

PAGE 48

38 Methods for Phase One. Survey Development To address the need for a survey instrument that can be used to measure changes over time in parents vaccine concer ns, an evaluation of reliability and validity of a novel survey instrument was conducted There were two cohorts used in this phase of the investigation. The P ilot cohort consisted of 120 pregnant women. Aft er the instrument was revised it was administered to a different cohort the Primary cohort of 320 pregnant women and parents of children under the age of twelve months A third cohort, the Validation cohort, was added as a validation check. It consisted of 374 pregnant mothers and parents of children under the age of twelve months. Below is a description of each cohort that will guide the reader throughout the dissertation. Pilot C ohort This was the first cohort assembled to test the pilot survey. The Pilot cohort consi sted of 120 pregnant women at the time they were recruited Participants were administered a baseline survey when pregnant; Primary C ohort This wa s a cohort assembled after the Pilot c ohort baseline survey was analyzed and revisions were made to the survey instrument. It consisted of 320 pregnant women and parents of children less than twelve months of age. The Primary cohort was used to finalize the survey instrument. Validation C ohort This cohort of 374 pregnant mothers and parents of children under the age of twelve months included the Primary cohort described above and the no

PAGE 49

39 treatment arm of a randomized control trial for behavior change about childhood vaccines The purpose of this Validation cohort was to analyze a larger sample of children in the undervaccinated strata and test the association between the outcome of undervaccination and the survey constructs. Table 3.2 shows the types reliability and validity and the methods used to assess reliability and validity. Table 3.2. Methods Used to Determine Reliability and Validity of the CABI V Survey Type of Reliabilit y or Validity Method Used Reliability Internal consistency reliability Cronbachs Alpha Test re test reliability Pearson r Validity Face validity Literature review; review by investigator and study team ; cognitive interviews with target population Content validity Expert panel Construct validity Exploratory and Confirmatory Factor Analysis Predictive validity Logistic Regression Criterion validity Multiple measures of hesitancy Procedures All regulatory compliance (Research Review Committee app roval, Institutional Review Boar d (IRB) approval) were obtained for the study conducted at Kaiser Permanente Colorado, Institute for Health Research. The University of Colorado Denver, Colorado Multiple Institutional Review Board approval (COMIRB) ceded re gulatory oversight to Kaiser Permanente Colorado IRB. All data was stored at Kaiser Permanente Colorado Institute for Health Research, within a secured, locked facility. Data was stored in a password protected Access database. The limited dataset contained a unique study identification number (ID) which was used to link to EHR data.

PAGE 50

40 Using a theory driven approach, a twenty three item survey instrument was developed to measure participants beliefs, concerns, intentions, control, and subjective norms about childhood vaccines The first phase established face and content validity. This involved the following steps: conduct a literature review assemble an expert panel to review the survey items, and conduct cognitive interviews with the target population. The survey instrument comprised previously established items that were modified as well as novel items. The steps are described below. The following three steps were conducted to establish face and content validity when developing the measurement instrument. Step O ne : L iteratu re R eview First, a broad based literature review was conducted. Survey items that have been previously published were categorized by year, author, theoretical model and behavioral constructs and entered into a database. This item invento ry provided easy organization of the existing survey items, linked the theoretical constructs to survey items and revealed gaps in measurement of theoretical constructs that required creation of novel survey items. Step T wo : Expert P anel Next, a panel of six national subject matter experts (SME) was convened. The SMEs had expertise in the areas of vaccine policy, research, epidemiology, and clinical practice. Their areas of profession included pediatricians, behavioral scientists, public health epidemiolog ists, and pol icy academicians (Table 3.3 ). The SME s received a link to an online rating system and received no incentive for their

PAGE 51

41 expert participation. The SME s independently rated each survey item on its strength of contribution to the identified theoretical construct using a five point scale (strongly unfavorable to the concept to strongly favorable to the concept). SMEs also had the opportunity to add qualitative comments about the wording, structure and measure of the survey item. Mean scores from the six SMEs for each survey item were calculated. Those items scoring less than three out of a five point scale were removed from the eligible pool of survey items. At the conclusion of these steps, a pilot survey instrument was developed and formatted. Table 3.3. Expert Panelist Expertise and Institutional A ffiliation Expert Panelist Area of Expertise Institutional Affiliation Douglas Opel, MD Vaccine safety, provider communication strategies, organizational issues underlying ethical conflict Scholar Treum an Katz Center for Pediatric Bioethics and Research Institute at Seattle Childrens Hospital; Professor in the Division of Bioethics, Department of Pediatrics at the University of Washington School of Medicine Saad Omer, PhD Vaccine safety, infectious dis ease epidemiology, policy analysis and pandemic disease Investigator Atlanta Clinical and Translational Science Institute; Emory Center for AIDS Research; Associate Professor Emory Vaccine Center; Global Health and Epidemiology, Rollins School of Public He alth, Emory University Hank Jenkins Smith, PhD Political science, social research, risk perception and vaccines Professor Center for Applied Social Research and Department of Political Science, University of Oklahoma Julie Downs, PhD Social influences and decision making, vaccine decision making Director Center for Risk Perception and Communication and Center for Behavioral Decision Research in the Department of Social and Decision Sciences, Carnegie Mellon University Simon Hambidge, MD, PhD Vaccine s afety and delivery, immunology, surveillance, provider communication Chief Ambulatory Office r Denver Health; Professor Colorado School of Public Health, University of Colorado Denver Health systems, management and policy and Department of Pediatrics, School of Medicine, University of Colorado Denver; Researcher Institute for Health Research, Kaiser Permanente Colorado

PAGE 52

42 Expert Panelist Area of Expertise Institutional Affiliation Matthew Daley, MD Vaccine safety and delivery, survey design, provider communication Senior Researcher Institute for Health Research, Kaise r Permanente Colorado; Professor Department of Pediatrics, School of Medicine, University of Colorado Denver St ep T hree: Cognitive I nterviews Then, the survey instrument was piloted on a sample of ten pregn ant women and parents of children under the age of twelve months who were identified using the EHR. Parents participated inperson for a oneto one thirty minute interview. Parents provided written consent to participate in the study and were given a tendollar gift card to a retail store for their par ticipation. Cognitive interviews were conducted to evaluate potential sources of response error in survey instruments. A protocol for administration of the interview was developed so that each session was conducted in a similar way. Techniques were used t o assess participants understanding of the intent, comprehension, and usability of the survey instrument (Willis, 2004) These techniques ask ed the participant to think out loud when answering the survey questions. The interviewer took careful notes as to what processes were taken to arrive at the answer. This was followed by verbal probing, where participants were asked probes such as What does this mean to you? and Tell me what this question is asking using your own words. Based on parents feedbac k, survey items were subsequently modified to improve clarity and readability. The qualitative data was organized by survey question and analyzed for themes (Patton, 2005) Modifications to the survey were made after the qualitative and quantitative expert panel and

PAGE 53

43 parent cognitive interview feedback was analyzed. These steps resulted in a finalized pilot survey. Results from Phase One Survey Development Results from Step One Literature Review, Expert Panel, and Cognitive Interviews The results from the development of the survey instrument are described below. These results from the literature review, expert panel and cognitive interviews establish face and content validity of the survey instrument. The initial survey item pool generated from the literature review was quite large. See Appendices A C and D for detailed survey development steps and survey items used at each study phase. There were over 150 potential survey items contributing to the constructs, not including the descriptive and demographic questions The investigator eliminated one hundred and nine survey items because they were duplicate items or irrelevant to the theoretical constructs. There were now forty three survey items. Through extensive literature review followed by assessment by the investigator and collaborative study team, the item pool was reduced. Using this method, the su rvey item pool was reduced to thirty items associated with the constructs in addition to other descriptive and demographic questions. The item pool wa s furt her reduced to twenty three items after the panel of six subject matter experts rated the questions and provided qualitative feedback. Fo r survey items with a mean rating of three or under, the questions were disqualified from the item pool. The qualitativ e feedback from the SMEs had two

PAGE 54

44 main themes: reorganization of survey items under construct headings and provide additional description and clarification by using qualifiers within the survey items to improve comprehension of the question. Qualitative feedback from the SMEs included the following quotes: Some questions are perfectly good questions, but don't fit perfectly well into any exact construct. That is OK As you can tell by my comments, sometimes it may make more sense to rethink what the construct is, and then go back to the questions (to see if they capture the construct). Great work; this is a really excellent integration of a number of important sources and concepts. Additional examples of qualitative feedback included suggestions to impro ve the content, context and understandability of the question. Several SMEs suggested a preference of one question over another For example, one SME commented, I wonder if the use of the term "disease" is too general and too vague here. It may be possibl e to replace with "several infectious diseases" or "infectious diseases such as diphtheria, polio, and measles." Finally, the results from the cognitive interviews were incorporated into the survey instrument. Overall, there were no significant comprehension difficulties

PAGE 55

45 across the sample. Parents provided specific word changes to the survey questions such as I think asking if the illnesses are serious may yield a better idea of whether or not someone agrees that all/none of the illnesses are serious many gets into the grey area to begin with and The question didnt sound right. It should say serious illnesses I intend to protect my child from. These and other suggestions were deliberated in a collaborative team discussion and some were incorporated into the f inal version of the survey instrument. Table 3.4 shows each survey item, survey response scale, theoretical construct thought to be associated with the survey item, and source of the survey item. See Appendix A for the survey items included prior to expert panel and cognitive interviews. Table 3.4. Survey Items Response Scale, T h eoretical Construct, and Source of Survey I tem (n=23) Survey item Response scale Theoretical construct Source of item Generally I intend to do what my childs doctor reco mmends about vaccines for my child Strongly disagree to strongly agree Subjective Norm s Freed, et al 2010 Most of the parents I know vaccinate their children Strongly disagree to strongly agree Subjective Norm s New item developed As a parent, I have given a lot of thought about vaccinations for my child a Strongly disagree to strongly agree Subjective Norm New item developed How confident are you that you have the necessary information to make decisions about vaccination for your child Very confide nt to not confident at all Perceived Control New item developed How confident are you that you will be able to protect your child from some types of infectious disease by vaccinating him or her Very confident to not confident at all Perceived Control New item developed Parents should have the right to refuse vaccines that are required for school for any reason Strongly disagree to strongly agree Perceived Control Freed, et al 2010

PAGE 56

46 Survey item Response scale Theoretical construct Source of item I believe many of the illnesses vaccines prevent are serious a Stron gly disagree to strongly agree Evaluation of VPD/VAE Opel, et al 2011 a,b Getting vaccinated will greatly reduce the chance of getting infectious diseases like polio or measles, but will not eliminate the chance of getting these diseases completely Stro ngly disagree to strongly agree Evaluation of VPD/VAE Song, et al 2014 I believe vaccines are generally safe Strongly disagree to strongly agree Evaluation of VPD/VAE Modified from Gellin, et al 2000 I believe my child could get a serious disease i f he or she were not vaccinated Strongly disagree to strongly agree Evaluation of VPD/VAE Kennedy, et al 2011 Children get more vaccines than are good for them a Strongly disagree to strongly agree Evaluation of VPD/VAE Gellin, et al 2000 My child will not need vaccines for diseases that are not common anymore, like polio a Strongly disagree to strongly agree Evaluation of VPD/VAE Adapted from Freed, et al 2010 I am concerned about serious infectious diseases like whooping cough or measles Stro ngly disagree to strongly agree Evaluation of VPD/VAE Kennedy, et al 2011 I am concerned that the ingredients in vaccines are unsafe a Strongly disagree to strongly agree Evaluation of VPD/VAE Gust, et al 2005 I am concerned that some vaccines cause autism in healthy children a Strongly disagree to strongly agree Evaluation of VPD/VAE Freed, et al 2010 I am concerned that there are serious side effects of vaccines a Strongly disagree to strongly agree Evaluation of VPD/VAE Freed, et al 2010 I believe it is better for my child to develop immunity by getting sick than to get a shot a Strongly disagree to strongly agree Beliefs about vaccines Salmon, et al 2005 I believe there has not been enough research on the safety of vaccines a Strongly disagree to strongly agree Beliefs about vaccines Freed, et al 2010 I believe it is better for my child to get the natural disease than to get a vaccine a Strongly disagree to strongly agree Beliefs about vaccines Modified from Salmon, et al 2005

PAGE 57

47 Survey item Response scale Theoretical construct Source of item Va ccines strengthen the immune system Strongly disagree to strongly agree Beliefs about vaccines Salmon, et al 2005 Getting vaccines is a good way to protect my child from infectious diseases Strongly disagree to strongly agree Beliefs about vaccines Freed, et al 2010 I am concerned that my childs immune system could be weakened by too many vaccines a Strongly disagree to strongly agree Beliefs about vaccines Gellin, et al 2000 I am concerned that it would be painful for my child to receive so m any shots during one doctors visit a Strongly disagree to strongly agree Beliefs about vaccines Gust, et al 2005 aReverse scored for EFA and Cronbachs Alpha After the survey was developed through a theory driven systematic process, face and content val idity was es tablished through assessment by cognitive interviews with parents and SMEs respectively (Litwin, 1995) Fleisch Kincaide r eadability statistics were performed on the pilot survey, indicating a 5.1 grade level. Altho ugh caution in interpreting the grade level statistic is suggested by Streiner and Norman ( 2014), this metric indicated an instrument that could be used in populations with lower literacy. T he pilot survey was now ready for administration on a cohort of pregnant women. The piloting o f the survey was essential to knowing what to refine within the survey Methods for Phase Two. Survey Evaluation In this second phase, the survey instrument was administered to a sample of the target population and then evaluated for construct validity and reliability (Thompson & Daniel, 1996) The following five steps were taken to evaluate the survey instrument. Step one identified the cohort and administered the survey instrument. Step two explored the facto rs underlying the survey items. Step three assessed internal

PAGE 58

48 consistency of the theoretical factors. Step four assesses the relationship between the observed survey items and the underlying constructs, as a model fit. Finally, in step five, two identical survey instruments were administered within fourteen days of one another to the same participant to establish test retest reliability. S tep O ne : Identify the C ohort and Administer the Survey I nstrument By using KPCO mem bership data, the P ilot cohort of pregnant women was identified. From these wom en, a random sample was drawn. They wer e then contacted to participate in the survey st udy. The identification of the Pilot cohort and the process of administering the survey instrument are described in detail below. Study S etting The study was conducted between April 2012 and December 2014 at Kaiser Permanente Colorado (KPCO). KPCO is a nonprofit integrated health care delivery system serving the healthcare needs of app roximately 600,000 members in twenty primary care clinics across the metropolitan Denv er area. KPCO uses an electronic health record (EHR), which captures demographic data, health plan enrollment information, encounter data including diagnosis codes, and vaccination administration information. It is usual care for parents of young infants to receive information about childhood development that includes standard handouts about vaccines, the diseases they prevent, and common adverse events of vaccines, such as swelling at the site of injection. Pregnant women receive verbal education and infor mational handouts about vaccines recommended during pregnancy and are offered these vaccines at their regularly scheduled obstetric appointments.

PAGE 59

49 At KPCO, most vaccines are delivered at routinely scheduled well child visits and recorded in real time into the EHR. The EHR also allows providers to track which vaccinations a child has received and which ones are due at the well child visit. Finally, the EHR allows pediatricians to code parents decisions about vaccinations directly into the m edical record, i.e. selecting some vaccin es and not others, electing for a different vaccine schedule from the recommended schedule, or declining vaccination for their child not due to a medical contraindication. The EHR also a llows researchers to obtain accurate informatio n regarding vaccination status such as the number of vaccines received compared to the recommended schedule and any reasons for delay of vaccination such as an illness, logistical barrier, or parental concerns. Study P opulation Using the EHR, an eligible c ohort of pregnant women between twenty and thirty two weeks gestation for the pilot survey and an eligible coh ort of pregnant women between twenty and thirty two weeks gestation and parents of children under the age of twelve months for the administration of the revised survey were identified for the study. From each of these cohort s, a random sample was drawn. An electronic data pull using International Classification of Diseases, 9th Revision, Clinical Modification (ICD 9 CM) codes for medical exclusions and manual medical record review was undertaken to confirm pregnancy and ascertain any medical conditions of the fetus (i.e. medical or elective abortion, fetal deaths, fetal abnormalities with high probability of not carrying child to full term). Exclusio ns includ ed the parent being under age eighteen, disenrollment in the health plan prior

PAGE 60

50 to planned contact about the study, or first language of the parent was other than English. Statistical comparisons to see if there were differences between those who a greed to participate and those who did not partic ipate (declined participation ineligible or did not respond) were conducted using chi square test of significance and ttest test of significance. R ace, ethnicity, age, insurance type, and number of days un dervaccinated were compared. For the Pilot cohort, participants were contacted by telephone and after consenting, and then sent a survey through postal mail. Parents were provided a tendollar gift card to a national retail store for completion of the surv ey. For the Primary cohort, participants were contacted up to two times by postal mail and asked to complete two surveys: a baseline survey followed by an identical survey fourteen days later. At each contact, a cover letter explaining the purpose of the study, a paper survey, and a postagepaid return envelope were included in the postal mailing. For completion of both surveys, participants received a fifteendollar gift card to a national retail store. The local human subjects review board approved the s tudy. At this point, the survey instrument was developed and tested on the study population. Using the baseline survey data from the study population who participated in the P ilot survey and the Primary survey, responses were entered into an Access databas e. Separate analyses were performed on these baseline data to assess factor structure, internal consistency, and goodness of fit using IBM SPSS Statis tics for Windows, version 22.0 (Armonk, NY: IBM Corp.) and SAS 9.2 (Cary, NC: SAS Institute Inc.). The inv estigator hypothesized that four factors would em erge

PAGE 61

51 from the data, based on the conceptual model developed for this investigation. The concept model incorporates the Theory of Planned Behavior and the Health Belief Model. I t was anticipated that the fact ors would be highly correlated as the survey items were measuring attitudes that are consistent across theoretical domains Step T wo Exploratory Factor A nalysis First, exploratory factor analyses were performed to determine the structure of the data. Bec ause health behavior data is correl ated, principal axis factoring (PAF) with direct oblim rotation were used Correlated survey data was expected, as most responses about behaviors and attitudes are similar, regardless of the construct. Normality of the data, independence of the observations, and linearity were tested. The Kaiser Meyer Olkin (KMO) measure of sampling adequacy was used to determine if there were sufficient items for each factor (>.70) and Bartletts test of sphericity was used to determine if correlations in the data exist to allow factor analysis (<.05) (Leech, Barrett, & Morgan, 2008) Step Three. Internal Consistency of Survey I tems Aft er factor analysis, Cronbachs A lpha was performed on the constructs. This method measured how well the items measure the latent construct. High correlations with other items in the same construct indicate the items should provide similar scores. A reliability value of .70 and above is considered acceptable (Cronbach, 1951). Step Four. Confirmatory Factor A nalysis Then, to determine if the Primary cohort data fit the hypothesized measurement model determined by EFA and Cronbachs Alpha, confirmatory factor

PAGE 62

52 analysis (CFA) was used (Thompson, 2004). To evaluate the model fit, root mean square error of approxim ation (RMSEA) and Bentler comparative fit index (CFI) were assessed RMS EA ranges from 0 to 1, with a smaller value indicating a better (Thompson, 2004) Step Five. Test Retest Reliability Finally, in the Primary cohort, repeatability was measured, us ing the method of test retest. For those participants who completed the baseline survey, the same instrument was mailed seven days later. This allowed for a fourteenday period of time from the baseline survey to the receipt and return of the second survey. Pearsons reliability statistic was used to determine intraindividual variability. A coefficient of .70 or higher is considered acceptable reliability. Results from Phase Two. Survey Evaluation The results from the evaluation of the measurement instrum ent are described below. These results included i dentification of the Pilot cohort and Primary cohort, factor analyses, assessment of internal consistency, and test retest reliability to establish construct validity and reliability of the survey instrument. Results from Step O ne: Identify the P ilot C ohort and Administer the Survey I nstrument For the Pilot cohort, the study population consisted of 120 pregnant women. Of the 194 eligible, 120 participated. Participants and nonparticipants were significantl y different with regard to race (X2=10.38, df=5, N=229, p<.05). White mothers were more likely than expected to participate than other race categories.

PAGE 63

53 Also, participants were significantly different from nonparticipants regarding age, t(209.06)=2.94, p<.01. The average age for participants (28.90) was significantly higher than the age (27.22) of nonparticipants. There were no statistically significant differences between those who participated and those who did not participant in ethnicity, days enrolled in Kaiser Permanente health insurance, and childs average days undervaccinated. Figure 3.1 shows the specific exclusions. Figure 3.1. Study Flow Diagram for Pilot Cohort Results f rom Step Two. Exploratory Fact or A nalysis of Pilot C ohort Exploratory Factor Analysis (EFA) is robust against skewed data. One variable was slightly skewed (many of the illnesses vaccines prevent are serious at 1.39). This variable was not transformed, as interpretation of transformed variables Study Pop ulation 555 Pregnant women in second half of second trimester or first half of third trimester within KPCO managed care organization from August 15, 2012 to December 15, 2012 Random sample 229 Pregnant women in second half of second trimester or first hal f of third trimester Ineligible after medical record review 25 Pregnant women ineligible 11 discontinued KP insurance 6 did not speak English as first language 4 serious health conditions 2 resided outside of Colorado 1 fetus had life threatening condition 1 elective abortion Invited for participation 204 Pregnant women Declined participation 22 Pregnant women declined after letter contact 40 Pregnant women unable to contact after 3 attempts Ineligible 10 infants will not have KP insurance Agreed to participation 132 Pregnant women consented and enrolled Baseline (T0) participation 120 Pregnant women Dropped out of participation 12 had child prior to completing baseline survey

PAGE 64

54 in facto rs is not possible. The initial EFA used Principal Axis Factor (PAF) analysis with varimax rotation to explore the dataset. KMO was .888, indicating the sample could yield factors from the data. Bartletts test of sphericity was significant, indicating there was adequate correlation between variables. Two extracted communalities were under .30. Using the criteria of eigenvalues at one or greater, there were six factors extracted. However, the first factor had twelve items, and the last two factor s had only two items per factor. As this exploratory approach produced too many survey items in one factor and too few survey items in other factors, the next step was to restrict the number of factors to fit the theoretical model. Therefore, the next iteration of f act or analyses was set to extract four variables, leaving all other criteria the same as previously described. Since the theoretical model defined four constructs, this provided a justification for setting extraction at four This rotation produced fo ur fa ctors, but with unfavorable results Again, the first factor had thirteen survey items and the last two fact ors had two survey items within each factor. The next iteration used PAF with direct oblim. Similar results were obtained. When extracted communal ities were removed, further solution was not possible within the twenty five rotation cycle Therefore, Cronbachs Alpha was used to explore the theoretical constructs.

PAGE 65

55 Results from Step Three. Internal Consistency of Survey I tems of the Pilot C ohort Cro nbachs A lpha was used to see how well an item in a construct correlated with the other items in the construct. High inter item correlations imply robust associations between items and the latent variable (Litwin, 1996) Using Cronbachs Alpha as a m easure of reliability, Table 3.5 below indicates reliable constructs in two of the four constructs (Beliefs about V accinating and Evaluation of VPD and VAE). Two constructs performed poorly (Sub jective Norms about V accinating and Perceived Control of Vaccinating D ecisions). These two constructs had only two questions in each co nstruct. Table 3.5 Internal Consistency of Survey Items within C onstructs Scale Alpha # questions Beliefs about vaccinating .769 5 Evaluation of VPD and VAE .839 8 Subjective norms about vaccinating .352 2 Perceived control of vaccinating decisions .671 2 Items were removed if the total score correlation improved without the item. The se items are listed in Table 3.6 below.

PAGE 66

56 Table 3.6 Survey Items Removed After Assessment of Internal C onsistency, n = 8 These exploratory analyses provided insights as to additional direction the survey instrument needed in order to assess reliability. Additional survey items were needed for th e two constructs that performed poorly (Perceived Control of Vaccinating Decision and Subjective Norms about Vaccinating) The survey instrument was revised by reassessing the literature on subjective norms and perceived control. Additional questions were added to the survey instrument in Survey Item Latent Construct Notes Parents should have the right to refuse vaccines that are required for school for any reason Perceived Control Deleted after pregnant cohort analysis after previous item deletions ) Most of the parents I know vaccinate their child Subjective Norm Deleted after pregnant cohort analysis af ter previous item deletions) Getting vaccinated will greatly reduce the chance of getting infectious diseases like polio or measles, but will not eliminate the chance of getting these diseases completely Evaluation of VPD/VAE Deleted after pregnant cohort analysis \205\212\203\220\211\207\003\213\220\003\003\210\224\221\217\003yzv\003 to .832 after deletion) I am concerned about serious infectious diseases like whooping cough or measles Evaluation of VPD/VAE Deleted after pregnant cohort analysis \205\212\203\220\211\207\003\213\220\003\003\210 rom .832 to .839 after deletion) That it would be painful for my child to receive so many shots during one doctors visit Beliefs about vaccines Deleted after pregnant cohort analysis \205\212\203\220\211\207\003\213\220\003\003\210\224\221\217\003yrr\003 to .769 after deletion) It is better for my child to develop immunity by getting sick than to ge t a shot Beliefs about vaccines Deleted due to duplicity with other survey item

PAGE 67

57 these constructs. Revisions to the instrument also included rewordin g of the middle response from N either agree nor disagree to Not sure. Twenty nine items were on the second iteration of the survey. Results from St ep One: Identify the Primary Cohort and Administer the Survey Instrument For the Primary cohort, the study population consisted of 320 pregnant women and parents of children under the age of twelve months. Participants and nonparticipants were significantl y different with regard to race (X2=13.92, df=5,N=520, p<.05). White mothers were more l ikely to participate than other race categories. Also, participants were significantly different from nonparticipants regarding age, t(341.01)=2.93, p<.01. The average age for participants (30.39) was significantly higher than the age (29.00) of nonparticipants. There were no significant differences between the two groups in ethnicity, days enrolled in Kaiser Permanente health insurance, and childs average days undervac cinated. In Figure 3.2, of the Pri mary cohort o f 520, there were 320 participants recruited and who completed a baseline survey for a response rate of 62% Two hundred and twenty two participants returned a second survey within the fourteenday period (69% response rate)

PAGE 68

58 Figure 3.2. Study Flow Diagram for Primary Cohort Results from Step Two. Exploratory Factor A nalysis of Primary C ohort Exploratory Factor Analysis (EFA) was conducted. One variable was moderately skewed (vaccines protect my child at 1.65). This variable was not transformed because interpretation of transformed variables in factors is not possible and EFA is robust against skewed data. The EFA used Principal Axis Factor (PAF) analysis with direc t oblim to explore the dataset. KMO was .933, indicating Study Population 834 Pregnant women in second or third trimes ter and 4,718 Parents of children less than 12 months of age within KPCO managed care organization in March, 2013 N=5,552 Random sample 534 Pregnant women and Parents of children less than 12 months of age Ineligible after medical record review 14 Pregnant women ineligible 5 did not speak English as first language 4 significant genetic c onditions of fetus 3 fetal demise outcomes 2 elective abortions Invited for participation N = 520 298 Pregnant women 222 Parents of children less than 12 months of age Declined participation (n = 184) Ineligible (n = 16) 14 no forwarding address 1 did not have KP insurance 1 in another study Returned baseline survey 130 Parents of children less than 12 months Returned second survey within 2 week period 87 Parents N = 222 Surveys e ligible for test re test reliability analysis Returned baseline survey 190 Pregnant women Returned second survey within 2 week period 135 Pregnant women

PAGE 69

59 the sample could yield factors from the data. Bartletts test of sphericity was significant, so there was enough correlation between variables to yield results. Eight extracted communalities were und er .30 and one question was duplicative o f another. The nine survey items in Table 3.7 were deleted after EFA. Using the criteria of eigenvalues at one or greater, there were four factors extracted. Table 3.7 Survey Items Removed after Evaluation of Facto r S tructure, n = 6 Table 3.8 shows the factor structure using EFA. For all of the items that load onto a particular construct, the other loading values are low. This factor solution r esulted in six items in the Belief s about Vaccinating construct, four in th e Perceived Control about Vaccination Decision construct, eight in the Evaluation of VPD/VAE construct, and five in the Subjective N orms about Vaccinating cons truct for twenty three items. The EFA explained 56.63% of the variance, which is acceptabl e in health behavior research (Hair, Black, Babin, Anderson, & Tatham, 2005). Survey Item Latent Construct Notes Information on the internet about vaccines helps me to make decisions about vaccinating my child Subjective Norm Deleted after secondary cohort analysis I think that there are other parents, like me, struggling with the decision about vaccines for their child Subjective Norm Deleted after secondary cohort analysis School laws requiring that children have up to date vaccines to enter daycare or public school influence my decisions about vaccinating my child Subjective Norm Deleted after secondary cohort analysis Allowing parents to delay vaccine doses or skip some vaccines lets parents be more in charge of their childrens health care Subjective Norm Deleted after secondary cohort analysis Parents who skip or delay certain vaccines are relying on other people in the community being vaccinated to protect their unvaccinated children from getting sick Subjective Norm Deleted after secondary cohort analysis The risks from getting a vaccine outweigh the ri sks from getting a disease Evaluation of VPD/VAE Deleted after secondary cohort analysis

PAGE 70

60 Table 3.8 Factor Loadings From Principal Axis Factor Analysis With Direct Oblim Orthogonal Rotation For A F our Factor Solution For Baseline Survey Items (N=320). Item Factor Beliefs Perceived Con trol Eval uation Subjective Norm Many of the illnesses vaccines prevent are serious .785 .006 .059 .058 My child could get a serious disease if not vaccinated .732 .088 .095 .015 My child does not need vaccines for diseases that are not common .521 .005 .244 .010 Getting vaccines is a good way to protect my child from infectious diseases .511 .115 .173 .149 Vaccines strengthen the immune system .416 .070 .23 2 .002 Confident I can protect my child from disease by vaccinating .406 .260 .187 .180 Confident about your knowledge about infectious diseases .009 .896 .106 .025 Confident about your knowledge about how vaccines work .017 .872 .009 .002 Confiden t you have the necessary information to make decisions about vaccines for my child .011 .756 .105 .028 Confident you are able to express your vaccine views to your childs pediatrician .028 .563 .101 .017 The ingredients in vaccines are unsafe .112 059 .875 .064 There are serious side effects of vaccination .058 .022 .849 .006 There has not been enough research on the safety of vaccines .028 .054 .743 .011 My childs immune system could be weakened by too many vaccines .157 .039 .701 .026 S ome vaccines cause autism in healthy children .031 .155 .635 .076 Children get more vaccines than are good for them .147 .007 .622 .111 It is better for my child to get the natural disease than to get a vaccine .380 .096 .481 .037 Vaccines are sa fe .279 .067 .447 .254 In general, my family (e.g. sisters, brothers and cousins) have similar beliefs about vaccines as me .082 .023 .022 .912 In general, my parents have similar beliefs about vaccines as me .039 .044 .012 .796 In general, most of my close friends have similar beliefs about vaccines as me .019 .026 .019 .511 In general, my obstetrician/childs pediatrician has similar beliefs about vaccines as me .087 .138 .235 .380 In general, my spouse or partner has similar beliefs about va ccines as me .200 .109 .008 .296

PAGE 71

61 Note: Bold text denotes highest factor value across all four factors. Those bolded in a column are in that columns factor grouping. Results from Step Three. Internal Consistency of Survey I tems of the Primary C ohort The items under each factor produced by EFA performed very well in internal consistency using Cronbachs Alpha and exceeded the threshold value of >.70. Table 3.9 Internal Consistency of Survey Items within Constructs Scale Alpha # questions Beliefs abou t vaccinating .832 6 Evaluation of VPD and VAE .921 8 Sub jective norms about vaccinating .783 5 Perceived c ontrol of vaccinating decisions .864 4 Results from Step Four. Confirmatory Factor A nalysis Findings from the confirmatory factor analysis (CFA) demonstrated model goodness of fit for the factor structure results of the EFA. The fit indices met or exceeded the suggested cutoff values frequently cited in the statistical literature (Bentler, 1990). The index for the Root Mean Square of Error of Appro ximation (RMSEA) was 0.07 and Bentler Model Fit was 0.90 Results from S tep Five. Test Retest R eliability From 520 parents sent an invitation to participate in the survey study, 320 returned a baseline survey. A second survey was sent to the 320. Of these 222 returned a second completed survey within the 14day window This was a 69% response rate for return of the second survey The overall temporal stability of the survey was very high, r = .930. Each of the four constructs had an acceptable

PAGE 72

62 correlation coefficient of well above .70. T his indicates that there is excellent testretest reliability for these data (T able 3.10). Table 3.10. Results from Baseline and Second Survey Administered Two Weeks Later. Scale Number of items in Scale Test Retest Reliability BELIEFS ABOUT VACCINATING 6 .849 Many of the illnesses vaccines prevent are serious .589 My child could get a serious disease if not vaccinated .612 My child does not need vaccines for diseases that are not common .684 Getting vaccines is a go od way to protect my child from infectious diseases .456 Vaccines strengthen the immune system .684 Confident I can protect my child from disease by vaccinating .689 EVALUATION OF VPD/VAE 8 .925 The ingredients in vaccines are unsafe .814 There ar e serious side effects of vaccination .728 There has not been enough research on the safety of vaccines .696 My childs immune system could be weakened by too many vaccines .787 Some vaccines cause autism in healthy children .807 Children get more vaccines than are good for them .767 It is better for my child to get the natural disease than to get a vaccine .773 Vaccines are safe .684 SUBJECTIVE NORMS ABOUT VACCINATING 5 .760 Friends have similar vaccine beliefs as me .617 Family have similar vaccine beliefs as me .679 Parents have similar vaccine beliefs as me .661 Spouse has similar vaccine beliefs as me .648 Pediatrician has similar vaccine beliefs as me .695 PERCEIVED CONTROL OF VACCINATING DECISIONS 4 .778 C onfident about your knowledge about infectious diseases .716 Confident about your knowledge about how vaccines work .686 Confident you have the necessary information to make decisions about vaccines for my child .728 Confident you are able to express your vaccine views to your childs pediatrician .634 Overall 23 .930

PAGE 73

63 Note: All significant at <.001 level Phase two presented methods and results for construct validity and reliability. After administration and analyses of the Pilot cohort of 120 pregnant women there was not a solution for factors using EFA. Internal consistency demonstrated good correlation for two of the four theoretical constructs. Further refinement of the survey instrument was necessary. The constructs of Subjective Norms about V accinating and Perceived Control about Vaccination Decision were revised by the addition of new survey items. Survey items that were similar in wording and meaning were removed. A revised survey instrument was ready for testing. Therefo re, a new cohort of 320 pregnant women and parents of children under twelve months of age were recruited. The baseline data were used to assess factor analyses, internal consistency analyses, and reliability. With a larger cohort, the revised survey instrument found a factor solution, performed very strongly in internal consistency, had good model fit, and high test retest reliability. Fleisch Kincaide readability statistics were performed on th e survey, indicating a 5.2 grade level. Next, phase three Predictive Validity, is presented. The Primary cohort survey constructs were assessed for association with the objective measure outcome of undervaccinated. Methods for Phase Three. Predictive Validity In this third phase, the objective was to determine if the survey constructs were associated with undervaccination. The Primary cohort was used for phase three. There were four steps conducted in phase three. First, the survey data from the parents survey responses was linked to their childs vaccination data. Second,

PAGE 74

64 from the c hild s vaccine data, an average days undervaccinated (ADU) metric was calculated. Third, the data was screened us ing descriptive statistics. Finally, in the fourth step, model building occurred. These four steps are described in the methods and results are then presented. The following four steps were taken to assess predictive validity of the survey instrument. Step One. Linking Parent Data to Child D ata Using KPCO s electronic administrative data, the mother was linked to their childs vaccine data. Th is was accomplished by associating the primary subscriber (parent) unique family identification numbers and the childs date of birth, using the healthplan data. Continuous enrollment was defined as no more than a thirty one day gap in insurance coverage. Children were also required to have been enr olled into the health plan by three months of age and still be enrolled at six months of age with no more than a thirty one day gap in order to capture early childhood vaccine doses administered during this time. In addition, children were required to be actively using the KPCO health system as defined by having at least one well child visit. Step Two. Calculating Average Days U ndervaccinated Of those children who met enrollment criteria, a verage days under vacci nated (ADU) was calculated ADU is a metric developed by Glanz, et al (2013a), which calculates the difference between when a vaccine d ose was truly administered and when a vaccine dose should have been administered according to the recommended vaccine sc hedule. These differences were combined across all

PAGE 75

65 vaccine doses that were due from birth to 200 days This length of time was selected to provide a fourteenday window from the age of six months of the child allowing a two week window for parents to bring their child to the clinic for the recommended six month vaccines Step Three. Screening the Data Before the model building process, descriptive statistics and univariate statistical methods (i.e. means, histograms, ttests, and chisquare tests) were use d to screen the data. For univariate analyses, continuous variables were collapsed into meaningful categories (See Appendix B for details) Survey construct scale variables were transformed into dichotomous variables where values of 1 through 3.49 were ( hesitant about vaccines) and 3.5 th rough 5 were transformed into 0 (not hesitant about vaccines). The cut point of 3.5 was selected after careful consideration of various response scenarios. This cut point allowed for leniency for parents to respond to some survey items with hesitancy as the literature reports parents who fully vaccinate have concerns about vaccines. Step Four. Model Building The models were constructed by the purposeful selection methods described by Hosmer and Lemeshow (2013). First all possible univariate log istic regression models with each independent variable = .25 leve l were included into the multivariate model. Variables that were determined to be meaningful to the model wer e also included and not removed regardless of p value (variable constructs, age, educatio n and race). The independent variables used in analyses were Beliefs about V accinating Evaluation

PAGE 76

66 of Vaccine Preventable Disease/Vaccine A dverse E vents Subjective Norms about V accinating Perceived Control about Vaccination Decision, h ealth literacy n umber of children i nfluences on vaccine decision, u se of social media for health information, a ge e ducation and r ace The independent variables were then removed sequentially based on their likelihood ratio test. As they were removed, their potential as a confounder was quantified by calculating a change in the coefficients of the model with and witho ut the variable. Covariates that produced changes in the coefficient greater than 15% were considered confounders and were left in the final model. Once the final model of main effects was established, meaningful interaction terms were considered and not p ursued due to theoretical unlikeliness. The model building pr ocess concluded with g oodness of fit testing and diagnostics (Hosmer, Lemesho w & Sturdivant, 2013) to assess the models fit and performance. Results from Phase Three. Predictive Validity Using the baseline survey data from the Primary cohort o f 320 pregnant women and parents of children under the age of twelve months and the vaccination status of their child, predictive validity was assessed. The results from linking parent data to the childs v accine data, calculating undervaccinated status of the child, data screening, and model building are presented below.

PAGE 77

67 Results from Linking Parent Data with Child D ata and Calculating Undervaccinated S tatus First, missingness of data from linking the mot her to the child was evaluated. Of the 320 children of parents who completed baseline surveys, 301 were linked. The nineteen that did not link were due to : one fetal death, one parent lost health insurance, six left KPCO health insurance plan prior to delivery, and eleven had missing unique identifiers in the EHR. Of the 301, fortyseven children were excluded for not being enrolled within three months of the birth and one was not actively using the system. Using enrollment criteria, 253 were actively using the KPCO health system. Of the 253, twelve did not have vaccine information available at age 200 d ays. Therefore, 241 children were eligible for analyses (see Figure 3.3) Of the 241, twenty nine were undervaccinated before age 200 days Undervaccinated was a dichotomous variable where the metric average days undervaccinated was either 0 days undervaccinated (fully vaccinated on time) or >0 days undervaccinated (undervaccinated). After manual medical record review of the twenty nine undervacinated cas es two were excluded. For these two, the reason for undervaccination was related to logistic barriers and not parental choice. Two hundred thirty nine cases w ere included in the analyses; twenty seven were undervaccinated before age 200 days.

PAGE 78

68 Figure 3.3. Study Flow Diagram of Primary Cohort Linking Mother to Child Results from Data S creening Overall, the cohort had an average age of thirty two years. The majority were white (88.7%) married (86.2%) and had a post high school ed ucation (94%) Additional descriptive information about the cohort is available in Table 3.11. Table 3.11. Characteristics of Pregnant Women and Parent Cohort by Vaccine Behavior, N=239 Characteristica ADU>0, n=27 ADU = 0, n=212 p value Gender n (%) F emale 26 (96.3) 205 (96.7) .913 Age in years, m (SD) 32.30 (4.65) 31.8 (4.79) .593 Age in years, range 24.0 41.0 19.0 46.0 Race, n (%) White 26 (96.3) 186 (88.2) .999 Nonwhite 1 (2.9) 26 (11.8) Hispanic ethnicity, n (%) Yes 3 (11.1) 31 (14.6) .823 No 24 (88.9) 181 (85.4 )

PAGE 79

69 Characteristica ADU>0, n=27 ADU = 0, n=212 p value Education n (%) High school degree or less 1 (3.7) 12 (5.7) .999 Some college or more 25 (96.3) 199 (94.3) Employment n (%) Employed full time 12 (44.4) 150 (70.8) .426 All other 15 (5 5.6) 62 (29.2) Marital status n (%) Married 24 (88.9) 182 (85.8) .999 All other 3 (9.01 ) 30 (14.2 ) Income n (%) 80,000K or less 23 (57.5) 161 (48.2) .999 More than 80,000K 17 (42.5) 173 (51.8) aFor age, independent t test was use d for this continuous variable. For all other variables, chisquare was used for categorical variables and Fischers exact test was used when cell size was less than five. Using the Pearson C hi square d T est for I ndependence statistic or Fisher Exact Test when cell size was less than five, p arents of undervaccinated children were less likely to be employed full time than parents whose children were fully vaccinated Additionally, results from the univariate analysis (Table 3.12) found that parents of underv accinated children were more likely to have more children, have adequate health literacy, use social media to search for health information and make their own decisions without influence of their childs doctor than parents of children who were fully vacci nated. In the univariate analysis, Beliefs about Vaccinating, Evaluation of VPD/VAE, Subjective Norms about Vaccinating, employment, number of children, health literacy, physician influence, use of social media, parent control of decision, were all significant in the model while Perceived Control about Vaccination Decision, race, age, education and income were not statistically significant at p=<.05. These characteristics are shown in the univariate analysis (Table 3.12 ). For example, parents whose child was undervaccinated were more than seven times

PAGE 80

70 more likely to have hesitant responses on the survey construct Beliefs about Vaccinating than parents whose child was fully vaccinated on time. Table 3.12. Univariate Analyses of Categorical Covariates from Pri mary Pregnant Mother and Parent Cohort, N=239 OR = odds r atio; CI = confidence interval Note: p value sign ificance = .05 aReferent is nonhesitancy Results from Model B uilding A logistic regression model was constructed where being undervaccinated before age 200 days (yes/no) was the binary outcome. This model determined the factors in the cohort data that contributed significantly to children being classified as undervaccinated Based on a p value of .25 in the univariate analyses, the independent variables considered in the full model were employment, race, number of children, health literacy, physician influence, social media use for health Covariate OR 95% CI p value Beliefs about V accinating a 7.27 3.06, 17.30 <.001 Evaluation of VPD/VAE a 17.96 4.15, 77.80 <.001 Perceived Control about Vaccination D ecision a 1.90 0.83, 4.35 .127 Subjective Nor ms about V accinating a 6.68 2.85, 15.63 <.001 Full time employment vs. all others (part time retired, student, unemployed, stay at home) 0.33 0.14, 0.76 .009 2 or more children vs. less than 2 children 3.61 1.52, 8.57 .004 Adequate health literacy vs inadequate health literacy 4.65 1.06, 20.31 .041 Influence of doctor in vaccine dec ision vs. other influences in vaccine decision (spouse, friends, family, no one, other) 0.25 0.07, 0.88 .030 Use social media for health information 1/week or more vs. u se of social media for health information once a month or less 3.54 1.15, 10.84 .027 Parent makes vaccine decision vs. shared decision or doctor makes decision 5.09 1.16, 22.21 .031 Race as white versus all other races 3.50 0.45, 26.89 .229 Education as college versus high school or less 1.57 0.20, 12.56 .672 Age as 30 years or older versus under 30 0.90 0.39, 2.12 .816 Income 80K and above versus under 80K 1.02 0.46, 2.28 .960

PAGE 81

71 information, and prefere nce for vaccine decision making The four survey constructs, Beliefs about Vaccinating, Evaluation of VPD/VA E, Subjective Norms about Vaccinating, Perceived Control about Vaccination Decisio n, and covariates age, race, and education remained in the model regardless of significance. The full model was built using the four survey constructs in the same model However, when the full model was run, three of the constructs and many covariates that were significant in the univariate analyses were not significant. When a reduced model using only survey constructs was built, two of the four constructs were significant with an additional construct trended close to significance: Beliefs about Vaccinating ( adjusted odds ratio (AOR), 2.66 ; 95% Confi dence Interval (CI) 0.98, 7.24), E valuation of VPD/VAE (AOR, 11.00, CI 2.40, 50.50), Subjective Norms about Vaccinating (AOR, 3.05, CI 1.15, 8.13) and Perceived Control about Vaccination Decision (AOR, 0.44, C I 0.25, 1.81) When each of the four construct s was modeled without any other constructs in the model, each construct was significant by association with the outcome of undervaccination. Thu s, the results below (Table 3.13 ) use one construct per model, adj usting the model by using the same covariates in each model.

PAGE 82

72 Table 3.13. Log istic Regression Analyses of Independent Modeling of Survey Constructs and Factors Associated with Undervaccination with Adjusted Odds Ratios aSurvey responses were binary where 1 3.49=hesitant and 3.5 5=nonhesitant bSignificance level is <.05 Adjusted ORs are adjusted for employment, education, race, age and number of children A OR = adjusted odds r atio; CI = confidence interval Hosmer and Lemeshow: Beliefs about vaccinating= .719; Evaluation of VPD/VAE = .708 ; Perc eived Control = .541; Subjective Norms = .294 In separate models, all four survey constructs were significantly associated with undervaccination after cont r olling for employment status education, race, age, and number of children. Being classified as hesi tant on Beliefs about V accinating (adjusted odds ratio (AOR), 10.01; 95% Confidence Interval (CI), 2.11, 18.00) Evaluation of VPD/VAE (AOR, 24.92, CI, 5.22, 119.01) and Subjective Norms (AOR, 13.65, CI, 4.71, 39.53) were significantly associated with unde rvaccinated status Being classified as nonhesitant on Perceived Control about Vaccination D ecision was negatively associated with undervaccinated status (AOR, .0.39, CI, 0.16, 0.96). Evaluation of VPD/VAE had the highest o dds ratio. Parents who were class ified as hesitant on the Evaluation of VPD/VAE construct were almost twenty five times as likely to have an undervaccinated child as compared to those classified as Covariate A OR 95% CI p value b a B eliefs about V accinating Hesitant 10.01 2.11 18 .00 <.001 Nonhesitant Referent --a Evaluation of VPD/VAE Hesitant 24.92 5.22, 119.01 <.001 Nonhesitant Referent --a Perceived Control about Vaccination D ecision Hesita nt 0.39 0.16, 0.96 .040 Nonhesitant Referent --a Subjective Norms about V accinating Hesitant 13.65 4.71, 39.53 <.001 Nonhesitant Referent --

PAGE 83

73 nonhesitant on the construct. Diagnostic testing using Cooks distance were all less than o ne, indicating overall influence of a cas e on the model was not a factor (Cook & Weisberg, 1982). Finally, goodness of fit was checked using the Hosmer and Lemeshow metric. All models were nonsignificant, indicating good fit. Next a Validation cohort of two similar study groups was constructed The first group was the original cohort used (the Primary cohort) in the logistic regression analyses. The second group consisted of the baseline usual care participants of an online intervention to reduce vaccine hesitancy and increase vaccine behaviors for pregnant mothers and parents of children less than twelve months of age. Those children who were undervaccinated underwent medical record review. Two cases were removed after medical record review as the cause o f undervaccination were logistical barriers rather than parental choice to decline or delay vaccines. Table 3.14 describes the Validation cohort. Descriptive statistics (ttests and chisquare) were conducted to test for differences between the undervaccinated and fully vaccinated groups. Table 3.14. Characteristics of the Validation C ohort by Vaccine Behavior, N=374 Characteristica ADU>0, n=40 ADU = 0, n=334 Gender n (%) Female 39 (97.5) 323 (96.7) Age in years, m (SD) 31.8 (4.23) 31.9 (4.70) A ge in years, range 24.0 41.0 19.0 46.0 Race, n (%) White 39 (97.5) 292 (87.7) Nonwhite 1 (2.5) 41 (12.3) Hispanic ethnicity, n (%) Yes 5 (12.5) 41 (12.3) No 35 (87.5) 293 (87.7 ) Education n (%)

PAGE 84

74 Characteristica ADU>0, n=40 ADU = 0, n=334 High school degree or less 1 (2.5 ) 1 4 (4.2 ) Some college/college degree 39 (97.5 ) 319 (95.8 ) Employment n (%) Employed full time 18 (45.0) 232 (69.5) All other 22 (55.0) 102 (30.5) Marital status n (%) Married 36 (90.0 ) 297 (88.9) A ll other 4 (10.0 ) 37 (11.1 ) Income n (%) 80,000K or less 23 (57.5) 161 (48.2) More than 80,000K 17 (42.5) 173 (51.8) aFor age, independent t test was used for this continuous variable. For all other variables, chisquare was used for categorical variables and Fischers exact test was u sed when cell size was less than five. In this Validation cohort, parents had similar characteristics as to the previous Primary cohort. Par ents were, on average, almost thirty two years of age. A majority of the parents were white (898%) highly educated (96%) and married (89%) T here were statistical differences between those undervaccinated and those who were fully vaccinated that were very similar to the smaller Primary cohort. These characteristics are shown in the univariate analysis (Table 3.15 ). P arents of undervaccinated children were less likely to be employed full time than parents whose children were fully vaccinated. Additionally, parents of undervaccinated children were more likely to have more children, and make their own decisions without influence of their childs doctor than parents of child ren who were fully vaccinated. Parents with two or more children were 2.3 times as likely to have an undervaccinated child as parents with fewer than two children. However, three variables were not significant: health literacy, use social media to search for health information, and influence of doctor on their vaccine decision.

PAGE 85

75 Table 3.15. Univariate Analyses of Categorical Covariates from Validation Study Cohort, N=374 OR = odds r atio; CI = confidence interval Note: p val ue significance = .05 aReferent is nonhesitancy Next, a logistic model was constructed by again using undervaccinated at 200 days status (ADU) as the outcome. Independent variables were obtained from the data screening processes. The methods used in the p revious logistic regression were applied. Covariates were assessed for confounding and education remained in the model due to confounding With the larger strata of forty underva ccinated data points, the model results using all constructs and significant o r clinically important covariates are presented in Table 3.16 Covariate OR 95% CI P a Beliefs a bout vaccinating 8.61 4.15, 17.86 <.001 a Evaluation of VPD/VAE 15.24 5.30, 43.84 <.001 a Perceived control about vaccination decision 2.29 1.14, 4.58 .020 a Subjective norms about vaccinating 8.43 4.13, 17.23 <.001 2 or more children versus less than 2 c hildren 2.30 1.18, 4.46 .014 Full time employment vs. all others (part time retired, student, unemployed, stay at home) 0.36 0.19 0.70 .003 Influence of doctor in vx dec ision vs. other influences in vaccine decision (spouse, friends, family, no one, other) 0.39 0.16, 0.95 .037 Parent makes vaccine decision vs. shared decision or doctor makes decision 5.42 1.63, 17.99 .006 Race as white versus all other races 5.48 0.73, 40.94 .098 Education as college versus high school or less 1.71 0.22, 13.37 .608 Age as 30 years or older versus under 30 0.70 0.35, 1.37 .290 Income 80K and above versus under 80K 0.70 0.35, 1.37 .268 Source of dataset: Primary versus Usual Care 1.20 0.60, 2.40 .617

PAGE 86

76 Table 3.16. Logistic Regression Analysis of Survey Constructs and Factors Associated with Undervaccination with Adjusted Odds Ratios, N = 374 Survey response s were binary where 1 3.49=hesitant and 3.5 5=nonhesitant Hosmer and Lemeshow = .481 A OR = adjusted odds r atio; CI = confidence interval Three of the four sur vey constructs were significantly associated with undervaccination after controlling for the other survey constructs, employment, education, race, age, number of children and source of the dataset. Being classified as hesitant on Belie fs about V accinating (AOR, 3.27, CI, 1.27, 8.42) Evaluation of VPD/ VAE (AOR, 9.34, CI, 2.79, 31.29) and Subjective Nor ms about V accinating (AOR, Covariate A OR 95% CI (lower) p value Beliefs ab out v accinating Hesitant 3.27 1.27, 8.42 .014 Nonhesitant Referent --Evaluation of VPD/VAE Hesitant 9.34 2.79, 31.29 <.001 Nonhesitant Referent --Perceived Control about Vaccination D ecision Hesitant 0.74 0.29, 1. 90 .533 Nonhesitant Referent --Subjective Norms about V accinating Hesitant 4.90 1.97, 12.14 .001 Nonhesitant Referent --Number of children 2 or more children 4 .13 1.70, 10.05 .002 < 2 children Referent --Emplo yment Full time employment 0 35 0.15, 0.82 .015 All other categories Referent --Education Col lege education 1.79 0.17, 19.14 .630 High school or less Referent --Race white 1 0.15 1.03, 99.91 .047 All other races Referent --Age Age 30 years or older 0.38 0.15, 1.52 .041 Age < 30 years Referent --Source of dataset Primary cohort 1.24 0.50, 1.97 .644 Usual care Referent --

PAGE 87

77 4.89, CI, 1.97, 12.14) all had an increased likelihood of being associated with undervaccinated s tatus. Perceived Control about Vaccination D ecision was not significantly associated the likelihood of having undervaccinated status (AOR, .74, 0.29, 1.90) Similar to the previous logistic regression using independent models on the Primary cohort, in this Validation cohort, parents who were classified as hesitant on Evaluation of VPD/VAE were 9.3 times more likely to have an undervacc inated child than parents who were classified as nonhesitant Cooks distance were all less than one so no data points were outliers. Finally, goodness of fit was assessed using the Hosmer and Lemeshow test of significance. The fit is good as the test is i nsignificant at 0. 481. Discussion The objective of this research was to develop and evaluate a survey instrument designed to measure parents concerns, attitudes, beliefs and intentions about vaccines for their c hild. To accomplish this, qualitative and quantitative methods were used. First, a comprehensive literature review id entified potential survey items. Next, a panel of vaccine experts and parents from the target population provided feedback on the survey. This step provided face and content validity to the survey instrument. The pilot survey was administered to 120 pregnant women. Evaluation of internal consistency of the baseline data resulted in two of the four constructs with acceptable alpha values. T he survey instrument was revised and new items were added. Then, 320 pregnant women and parents of children under the age of twelve months were administered the revised survey instrument. Data from the revised survey were

PAGE 88

78 analyzed for factor structure, internal consistency, repeatability and model fit. These results provided construct validity of the survey instrument. Finally, logistic regression was performed to determine if there was an association between the survey constructs developed throu gh the factor analysis and undervaccination at age 200 days. By modeling one survey construct per model, there were strong associations with being undervaccinated at 200 days for three of the constructs: Beliefs about Vaccinating, Evaluation of VPD/VAE, and Subjective Norms about Vaccinating. Using the approach of one construct per model, predictive validity was established yet it was not the traditional model approach For this reason, a validation was performed on a larger cohort and although the odds ratios were lower when all survey constructs were in one model, it performed similar to the smaller cohort. This provided validation that with larger sample, three of the four constructs are associated with undervaccination. The investigation had relatively hig h participation rates. For the Pilot cohort of 120 pregnant women, the response rate was 62%. This was similar for the P rimary cohort of 320 (62%). This higher response rate helped shield against missing data not at random (Evans, 1991). Engaging subject matter experts to review the instrument and then conducting cognitive interviews with parents to pilot the survey contributed to further revision of the survey instrument. Analyzing the quantitative and qualitative data obtained from these engagement processes produced an improved survey When des igning and evaluating surveys, there are s everal important points which should be considered. First, it is important to have a theoretical concept to

PAGE 89

79 guide the decision making process in designing and evaluating a survey instrument. Although statistical results pro vide some guidance in deciding what items remain in the construct and which are removed, much of the decisionmaking and interpretation of meaning in the data is ma de by the investigator (Blair, Czaja, & Blair, 2013; Fabrigar, W e gener, McCullum, & Strahan, 1999) S econd, t he time cost and labor involved in primary data collection and data analysis is intensive. Most surveys require technical expertise to either design a paper based copy or proficiency in survey software and code such as Structured Query L anguage (SQL) Hyper Text Markup Language (html) to develop web based survey instruments. There are also costs of incentives to compensate participants time for completion of surveys. Much labor is expended on mailing surveys, monitoring web based survey p latforms and follow up with participants, and data entry and data cleaning procedures. These processes have protracted timelines that extend into years of research. Third what is a relevant vaccine concern to a parent today can evolve and change over tim e. S urvey instruments need to be designed so these new concerns can quickly be assimilated into the instrument without sacrificing reliability and validity of the instrument. This can be accomplished by adding the items to the next administration and reanalyzing the baseline data using the steps outlined in this chapter. Survey development is a collaborative and iterative process. Development that is followed by administration of the instrument and analyses, and then followed by additional development prov ided applicable results. This iterative process was

PAGE 90

80 used for this survey by piloting it on a cohort of pregnant women, analyzing the data, revising the survey instrument, then administering the Concerns, Attitudes, Beliefs, and Intentions about Vaccines ( C ABI V ) survey to a new cohort of parents. This permitted careful development and understanding of the constructs and item measurements. This study adds to the knowledge base for future survey based studies. One of the most important operational findings in this study was that exploratory factor analysis power estimates published in the literature for surveys do not always converge towards a solution. In the literature, sample size estimates average from 100 as the low estimate to over 300 as the optimal estimate to produce stable statistical results (Nunnally, Bernstein, & Berge, 1967; Nunnally, 1978 ; Comrey, 1988) Tinsley and Tinsely (1987) recommend five to ten subjects per survey item. In this investigation, the 120 baseline surveys obtained in the Pil ot coh ort was not an adequate sample did not result in a solution due to no convergence. Additionally, when using undervaccinated as an outcome, it is imperative to have adequate sample size Undervaccinated is rare in the population and calculating it precisely and accurately is complex Because it is important to capture only truly undervaccination by parental choice, careful medical record review is required. In this study, medical record review validated the outcome of undervaccination. Exclusion of reasons other than choice is necessary. This research found that children on public or high deductible insurance do not have continuous enrollment. Thus, they are removed from the final analyses for the outcome of undervaccination. For this study 320 pregnant women and parents were recruited

PAGE 91

81 Of these, 241 of their children had continuous enrollment (75%) from which to calculate undervac cination status at 200 days of age. Opel et al. (2013b) had similar proportions, 310 of the 437 children had continuous enrollment at nineteen months of age (71%). However, Opel allowed a more liberal definition for breaks in insurance coverage but had an extended observation period to nineteen months of age. This highlights the importance of adequate sample size and provides valuable information on expected data loss when using restrictions of enrollment for future studies. I t is also important to consider the outcome measure of undervaccinated produced by the A DU algorithm was associated with being hesitant on the survey c onstructs. Although there have been numerous single study surveys designed to measure parents hesitancy about vaccines or knowledge, attitudes and beliefs about vaccines, there has not been extensive rigor in developing and evaluating an instrument that c an be used in multiple settings and research designs, such as interventions. There are several significant limitations to this investigation. The outcome variable used in logistic regression was rare. Therefore, continuous enrollment of 239 children for 2 00 days after birth was available and the logistic regression had only twenty seven (11%) in the undervaccinated group. In addition, h ealth behavior survey data is highly correlated. It was hypothesized a priori that the survey constructs would be mul t i collinear. Therefore, by using each survey construct independent of the other constructs through logistic regression, the covariates were added and statistical significance was observed. When a larger sample was

PAGE 92

82 assembled by adding a cohort of similar parents from an intervention study, there was a significant association between undervaccination and three survey constructs. It is possible that the survey instrument is not generalizable to other geographical regions and research. It is important to test the instrument in other geographic regions and with other types of i ntervention studies to evaluate its strength of association with undervaccin a tio n. The c onstruct of Perceived Control of Vaccination Decision was not associated with undervaccination in univar iate and multivariate analyses. T here is a need to test additional survey items and assess which items perform well in differentiating vaccine status There is an urgent need to develop and implement effective interventions to improve vaccination and reduce parental concerns. It is also imperative to have a survey instrument that accurately and reliably measures parental concerns about vaccines over time. Through this development and evaluation process, an instrument that does have reliability and validity was produced. The next steps are to e valuate its use in interventions, across multiple settings and with diverse populations to demonstrate its reliability across populations Notably, these settings and populations may drive the need to change the instrument. If effective across geography, setting, and populations, the survey instrument would represent a useful and broadly applicable resource to researchers, public policy and health care practice. Using the same questions over ranges of populations by m ultiple studies could potentially provide a repository of data for change in attitude and beliefs about vaccines.

PAGE 93

83 This chapter described the development and evaluation of a survey instrument to measure parents concerns, attitudes, beliefs, and intentions about childhood vaccines. The instrument performed well in the areas of face, content and construct validity, internal consistency, and test retest reliability. Analyses demonstrated strong association between survey constructs and undervaccination. In i ndividual models, three constructs demonstrated significant association with undervaccination. However, when the constructs were entered into the model together, the results were different and two constructs were not significant. In the next chapter, the Pilot cohort data was analyzed over four time po ints. Repeated measures methods were used to assess changes of the survey items

PAGE 94

84 CHAPTER IV DESCRIPTION OF PREGNANT MOTHERS VACCINE DECISION MAKING PROCESS OVER TIME In this chapter, the survey data from the Pilot cohort administered to 120 pregnant women at four time points is presented. The survey items were analyzed over time using inferential statistics to assess significant changes in measures. First, there is a brief introduction to d ecis ion making of parents, the sources of information they rely on, and literature on previous investigations on parents decision making over time. This is followed by the study design, methods, results and conclusions. Introduction Parents face decisions ab out vaccines for their child within the first hours after birth. For parents, this decision making process can seem overwhelming. They often turn to multiple sources of information about vaccines which can be inaccurate, depending on the source. Additionally, parents must accurately comprehend scientific information about vaccines and the immune system that may be complex, poorly presented or lack details that allow for full understanding (Downs, 2008) I nfluential sources of information include significant others (i.e. family, friends), the childs pediatrician, and media sources. These sources can also influence parents concerns, attitudes, beliefs and intentions about vaccines for their child (Kennedy, 2011) Although the topic of decision making about vaccines has been well covered in the literature (Benin, et al. 2006; Serpell & Green, 2006; Sturm, Mays, & Zimet,

PAGE 95

85 2005) there are few studies that examine the decision making process prospectively. Most survey studies use retrospective attitude assessment and self reported vaccine status, which limit the reliability of the data and results. Several longitudinal survey studies have been conducted to measure parents intentions to vaccinate prior to or shortly after the birth of their child as compared to the parents actual vaccine behavior. These studies used self reported status of the parent or review of available medical records to capture this behavior. Wroe, et al (2005) recruited 195 women in their third trimester of pregnancy and measured intentions to vaccinate their child, perceptions of risks and benefits of vaccinating, and feelings of regret of vaccine decision. She found that feelings of responsibility and regret were strong predictors of decisions about vaccination. Opel and colleagues (2013b) administered a vaccine hesitancy screener to 437 parents of infants wh o were two months of age and followed them forward in time to collect their vaccine behavior using the EHR. Of the 310 children who had continuous enrollment and complete EHR records at age nineteen months, those w ith higher hesitancy scores at two months of age had more days undervaccinated. These early investigations use prospective decision making and demonstrate the importance of well designed data collection and access to objective vaccine behaviors. This study provides longitudinal descriptive information on a cohort of pregnant mothers. The aim of this study was to prospectively capture pregnant womens concerns, attitudes, beliefs, and intentions about vaccines for their child for four time points prior to the birth through the sixth month of age of the child.

PAGE 96

86 First, a cohort of pregnant mothers identified through the EHR was contacted. Next, t he survey developed in chapter three was administered at four time points to mothers. Additional descriptive variables were added to the survey (i.e. health literacy measures, demographics, decisionmaking preferences, hesitancy about vaccines). Finally, analyses were performed to describe the data over time. Methods for Longitudinal Data Collection Design of the Study To address the need for further descriptive information collected prospectively on parents decisionmaking about vaccines for their child over time, a prospective survey study was conducted with 120 pregnant women who were f ollowed forward in time from pregnancy through six months of age of their child. The Concerns, Attitudes, Beliefs and Intentions ( CABI ) survey, which was a pilot survey developed for this study was administered a four time points: second to third trimester of the mother and then at two, four and six months of age of the child. These time points were strategic ally selected to align the survey collection at points when parents make decisions about vaccines for their child as they typically attend well child checks when the infant is two, four, and six months of age. Demographic information was collected at the baseline survey. This prospective approach protects the predictor measure from becoming influenced by the knowledge of the outcome (vaccine behav ior) a nd reduces recall bias (Figure 4.1)

PAGE 97

87 Figure 4.1. Design of Prospective Survey Data Collection, N=120 Study Population for the Longitudinal Study Within the KPCO health care system, approximately 5,000 live births occur each year. Prior to the birth, pregnant women receive care within the health system at KPCO clinics. There were three steps to creating the longitudinal cohort. First, using the EHR, women between twenty to thir ty two weeks gestatio n were identified. In this step, women with unknown pregnancy status or serious medical conditions of fetus or mother were excluded. Next, a random sample was drawn from the eligible cohort. Finally, manual medical record review was performed to confirm el i gibility. The details of these three steps are provided below. Step O ne Identification of S tudy Cohort

PAGE 98

88 Pregnant women who were in their late second to first half of their third trimester (between twenty and thirty two weeks of gestation) of pregnancy and members of the integrated health delivery system, Kaiser Permanente Colorado, were identified for this study. The estimated delivery date (EDD) is available in the EHR to estimate pregnancy parameters. Exclusions were made through the initial electronic d ata pull (using International Classification of Diseases, 9th Revision, Clinical Modification (ICD 9 CM) diagnosis codes for incl usion and exclusion criteria). The decision was made to begin recruitment after twenty weeks gestation as most women r eceive an ultrasound prior to twenty weeks in which fetal genetic abnormality screening is performed. Thus, those women with serious fetal compromise were excluded from t he cohort (i.e. fatal heart condition, anencephaly, spontaneous abortion) Additionally, women with less than 60 days enrollment in KPCO or less than one obstetric visit were excluded. Pregnant women age eighteen or over were eligible for the study. Step T wo Random S ample Using the EHR, estimated delivery dates (EDD) were extracted from the data files. Pregnant women falling within the parameters of twenty to thir ty two weeks gestation were then compiled into a cohort. From this eligible cohort, a random sample was drawn. Step Three. Chart Review V erification Manual chart review to confirm pregnancy and verify eligibility was performed on the entire random sample and those ineligible were removed from the cohort. Pregnant women with a serious health condition related to their

PAGE 99

89 pregnancy (i.e. very high risk pregnancy with long term bed rest; high probability of not carrying the child to full term) were excluded from the cohort through manual medical record review. Pregnant mothers whose first language was other than English were also excluded from the study sample as all materials were in English o nly. Enrollment The research was conducted between July 2012 and December 2013 at Kaiser Permanente Colorado (KPCO). Eligible pregnant women were initially contacted to participate in the study by a mailed letter and were provided an optout postcard to m ail back to the investigator to indicate they were not interested in participating. Once verbal consent was obtained, a validated vaccine hesitancy screener, Parent Attitudes about Childhood Vaccines (PAC V) was administered over the telephone and the investigator transcribed answers The PACV was administered as part of the feasibility study and is described in detail elsewhere in this chapter. Consent forms and PAC V responses were stored within locked storage at KPCO, using unique study IDs as the only method of identification. PAC V scores were also entered into a pass word protected Access Database. Participants were requested to return mailed survey within one weeks time from receipt of the survey materials. Participants received a gift c ard for ten dollar s to a retail store for each survey th ey completed, for a total of forty dollars in incentive across four surveys. Gift cards were sent to participants via postal mail with a personalized thank you

PAGE 100

90 letter. Subsequent surveys were mailed to parents w hen their children were two months of age, four months of age, and six months of age. If a survey was not r ec eived by postal mail within seven to ten days, parents were contacted by phone or email and were then sent by postal mail a blank follow up survey to complete in the event they misplaced the initial mailed survey. All surveys were tracked as to when they were sent out in postal mail and when surveys were returned, using unique study identifier s, an excel spreadsheet, and Access database. Analytic met hods The analytic plan for survey results had two m ain steps. In the first step, the survey measures were defined The specific response category for each survey question is available in Appendix C. In the second step, the statistical approaches to analyze the survey results are described S tep O ne. Defining Survey M easures For th e baseline survey, twenty three survey items were created to measure the concerns, attitudes, beliefs and intentions of parents about vaccinating their child. Survey items consis ted of continuous, categorical and dichotomous variables. These survey items are described in detail in chapter three. Additional survey items included measures of health literacy and numeracy, demographics (age, race, ethnicity, income, number of children, education, marital status, and employment), vaccine intention, decisionmaking preferences about vaccines for children, and personal reference to someone who experienced a vaccine adverse event. Health literacy and numeracy Health literacy was measured using a validated, self reported onequestion measure developed by Chew, et al. (2008). The

PAGE 101

91 survey item How confident are you filling out medical forms by yourself? has responses on a fivepoint Likert scale from extremely to not at all. The survey item classifies responses into two categories of inadequate (somewhat, a little bit or not at all) or adequate (extremely, quite a bit) health literacy. This one question screener has been shown to be as sensitive and specific as other lengthier measur es (Chew, et al, 2008) Health numeracy was also measured u sing a validated self reported threeitem measure. This shortened version of the Subjective Numeracy Scale (SNS) corre lated highly with the original eightitem measure. The brief screen asks for the participant to provide subjective ratings on the following questions How good are you with fractions?; How good are you at figuring out how much a shirt will cost if it is 25% off? and How often do you find numerical information to be useful? The respon ses are on a six point Likert scale that ranges from not at all good (1) to extremely good (6) for two questions and never (1) to very often (6) for one question (Osborn, Wallston, Shpigel, Cavanaugh, Kripalani, Rothman, 2013 ; M cNaughton, Cavanaugh, Kripalani, Rothman & Wallston, 2015 ). The three questions are summed (and in some studies, have been averaged), so that higher scores rep resent higher health numeracy. The range of summed scores is 3 to 18. Vaccine hesitancy and intention Parents hesitancy about vaccines for their child was measured by a validated survey, Parent Attitude about Childhood Vaccines (PAC V hesitancy screener). PAC V i s a fifteenitem survey develop by Opel, et al (2011a b; 2013b) to screen for parental hesitancy about chil dhood vaccines. It has been validated in one regional clinical setting, demonstrated high

PAGE 102

92 reliability, and shown to predict childhood vaccine status. I n collaboration with the developer of the PAC V Douglas Opel, MD the investigator modified the screener to function as a prospective hesitancy measure. The survey was modified by re wording the questions for future decisions and intentions To illustrate this change from present or past tense to future tense, the following example is provided : How sure are you that following the recommended schedule will be a good idea for your child versus the original How sure are you that following the recommended schedule is a good idea for your child. The investigator administered each survey question and response s et of the PAC V over the phone to parents who consented to the study, and prior to parents taking the paper and pencil baseline CABI V survey. The investigator into an Access database transcribed answers A one question intention to vaccinate question was developed for the study and used at all four time points As an expecting parent, I intend to followed by seven responses Get all vaccines on time Get some vaccines on time to Not sure Responses were dichotomized into 0 vaccinate according to the recommended childhood schedule and 1, not vaccinate according to the recommended childhood schedule. For the latter, this includes responses of getting some, none, delaying vaccines as well as not sure responses for pregnant participants who had not yet made their vac cine decision for their child. Influences on vaccine decision An original description question was developed to assess who influences parents decisions about vaccines Which person had the greatest influence on your decision about vaccines for the child you are currently pregnant with? This qu estion was administered at all four survey

PAGE 103

93 time points. The question was originally designed to allow participants to select all influences and then, in the subsequent question, select the most important influence. Only the most influential person selected as the response was analyzed. Additionally, a question developed by Opel (2011a) for his initial work in the area of vaccine hesitancy was added to ask, Do you personally know anyone who has h ad a bad reaction to a vaccine? and answered a response of either yes or no. Sources of information In order to capture information about other influenc es and sources of information, six original questions about use of internet and social media were created. The response scale ranged from not at all to every day. A question that asked If I had access to vaccine information, blogs and discussion boards to chat with other parents on a Kaiser Permanente vaccine information website, I think my level of concern about vaccines would to measure hypothetical influence of a moderated social media information site. The response scale was increase, stay the same decrease. Decision making preference. An original question was developed to measure parents style and preference of dec ision making about vaccines. A five point scale ranged from I prefer to make the final decision about which vaccines my child receives or does not to I prefer to leave all decisions about which vaccines my child receives or does not receive to my doctor. It was hypothesized that parents who selected responses reflecting preference to control the decision making process would be more likely to be hesitant about vaccines.

PAGE 104

94 Step Two. Survey Response Analytic Plan Next, the analytic plan for each survey question is discussed. First, data entered into databases can introduce errors. To assess this, a second person entered 10% of the data and th e samples were compared at all four survey time points using Kappa percent agreement. The investigation used a longitudinal study design with repeated measures for the same individuals survey data, looking at variables over time to detect any change in variables. Comparisons between those who agreed to participate and those who did not p articipate (declined, ineligible or did not respond) were conducted. This was done to investigate if there were differences between the two groups not due to chance. The source of data for the comparisons was extracted from the EHR for both groups. This wa s done to compare data from the same source of information, reducing bias in the data. However, this approach also resulted in missing data at the individual level. In the EHR, data such as demographics are not consistently recorded and populated across th e insurance membership. These comparisons included age of mother, days of enrollment, mothers type of insurance plan, mothers race, mothers ethnicity (classified as Hispanic or non Hispanic), and child average days undervaccinated (ADU). Age of mother, insurance type, and days of enrollment were determined using the index date of September 1, 2012. The age of the mother was calculated at September 1, 2012, as was insurance type regardless of chang e in insurance later For enrollment, up to a thirty one d ay gap was allowed, so that someone who l ost his or her health coverage for thirty days was counted as continuously enrolled. In calculating average days

PAGE 105

95 undervaccinated, vaccine doses were counted up to 365 days from birth. For categorical variables, chisquare test of significance was used. This statistical test is used when categorical variables have two or more possible values. Chi square test shows whether the proportion for one variable differs among the values of the other variables. When expec ted ce ll counts were less than five, Fishers exact test of significance was used. For continuous variables, ttest of significance was used. The independent ttest for equal means determines if two sets of data are significantly different from one another. The Satterthwaite unequal variances test of significance was used for all ttest results, as the population variance was unknown and assumed unequal or different from one to the other (Moser & Stevens, 1992). The statistical significance of <.05 was the cutoff criteria. Using multiple measures on the same individual over time is known as repeated measures data. Using means, ttests, Cronbach Amitage test of trends, analysis of variance (ANOVA) and chisquare analyses were conducted. Repeated measures methods p rovided information on whether there was a change in mean scores for each survey question over time using two time points as comparison. Results of the Longitudinal Cohort The pregnancy cohort f low diagram is shown in figure 4.2 below. Using a least restrictive approach to enrollment rates, 62% participated. This was derived by taking the 204 invited for participation and subtracting ten who were determined ineligible. One hundred and twenty who took a baseline survey were divided by 1 94 who were eligible t o take the survey The twelve individuals who did

PAGE 106

96 not take a baseline survey were considered dropped from the study although they consented to the study and took the PAC V vaccine hesitancy screener. Retention rates were high in all three subsequent d ata collection points. For the two month surv ey (when the mothers infant was two months of age), 91.7% of the cohort retur ned a completed survey. At the four month survey, 95% of the mothers returned their completed survey, and at six months, 94.2% returned their completed survey. Surveys were considered eligible for the data collection points if they were returned within a window of more than twenty one days or less than twenty one days of the age of the child (two months of age, f our months of age, and six months of age). If a parent returned a survey late and it fell within the eligibility of the subsequent survey window, it was used for that later data collection point if another survey for that data collection point was not returned. The percent agreement between the 10% sample of double data entry by two separate individuals for the overall pregnant cohort was 98.7%. Specifically, for the baseline data, th e percent agreement was 98.4%; two month data, 98.8%; for the four month data, 98.0%; and for the six month data, 99.6%. These agreement percentages demonstrate low data entry errors and high accuracy of the datasets. Statistical comparisons to see if there were differences between those who agreed to participate and those who did not partic ipate were co nducted. R ace, ethnicity, age, insurance type, number of days undervaccinated were compared

PAGE 107

97 Figure 4.2. Pregnancy Cohort Flow Diagram There were two statistically different comparisons between those who participated and those who did not. Participants and nonparticipants were significantly different with regard to race (X2=10.38, df=5, N=229, p<.05). White Study Population 555 Pregnant women in second half of second trimester or first half of third trimester within KPCO managed care organization from August 15, 2012 to December 15, 2012 Random sample 229 Pregnant women in second half of second trimester or first half of third trimester Ineligible after medical record review 25 Pregnant women ineligible 11 discontinued KP insurance 6 did not speak Engli sh as first language 4 serious health conditions 2 resided outside of Colorado 1 fetus had life threatening condition 1 elective abortion Invited for participation 204 Pregnant women Declined participation 22 Pregnant women declined after letter contact 40 Pregnant women unable to contact after 3 a ttempts Ineligible 10 infants did not have KP insurance Agreed to participation 132 Pregnant women consented and enrolled Baseline (T0) participation 120 Pregnant women Dropped out of participation 12 had child prior to completing baseline survey Child aged 2 months (T1) participation 110 Pregnant women Dropped out of participation (T1) 4 unable to contact 6 did not respond to T1 survey request Dropped out of participation (T2) 1 unable to contact 5 did not respond to T2 survey request Child aged 4 months (T2) participation 114 Pregnant women Child aged 6 months (T3) participation 113 Pregnant women Dropped out of participation (T3) 1 unable to contact 6 did not respond to T3 survey request

PAGE 108

98 mothers were more likely than expected to participate than other race categories. Also, participants were significantly different from nonparticipants regarding age, t(209.06)=2.94, p<.01. The average age for participants (28.90) was significantly higher than the age (27.22) of nonparticipants. There were no differences in ethnicity, d ays enrolled in Kaiser health insurance, and childs average days undervaccinated. Table 4.1. Characteristics of the Cohort of Pregnant Mothers, N=120 Variable N (%) Age Mean (SD) 29.37 (3.9) Missing/No response 0 (0.0) Range 20.0 43.0 18 to 24 1 7 (14.2) 25 to 34 98 (81.7) 35 and older 5 (4.2) Hispanic Yes 23 (19.2) No 97 (80.8) Race American Indian/Alaska Native 0 (0.0) Asian 1 (0.8) Black/ African American 3 (2.5) Native Hawaiian or Other Pacific Islander 1 (0.8) White 11 0 (91.7) Multi racial 5 (4.2) Education Less than High School 1 (0.8) High School Graduate 12 (10.0) Some College 18 (15.0) Bachelors Degree 48 (40.0) Post baccalaureate Degree 40 (33.3) Employment Status Employed full time 84 (70.0) Employed part time 13 (10.8) Student 2 (1.7) Stay at home parent 18 (15.0) Unemployed 2 (1.7)

PAGE 109

99 Variable N (%) Missing/No response 1 (0.8) Income Less than 40K 22 (18.3) 40K to 80K 47 (39.2) 80K to 120K 30 (25.0) 120K to 150K 13 (10.8) Over 1 50K 4 (3.3) Prefer not to answer 3 (2.5) Missing/No response 1 (0.8) Marital Status Married 96 (80.0) Living with partner 15 (12.5) Not married 2 (1.7) Separated 2 (1.7) Single 4 (3.3) Divorced 1 (0.8) Number of children None 80 (66.7) 1 27 (22.5) 2 or more 13 (10.8) How often do you have someone like a family member, friend, hospital or clinic worker help you read hospital materials a Always 5 (4.2) Often 11 (9.2) Sometimes 28 (23.3) Occasionally 22 (18.3) Never 53 (44.2) Missi ng/ No response 1 (0.8) Mean (SD) 2.10 (1.20) How confident are you at filling out forms by yourself Extremely 62 (51.7) Quite a bit 47 (39.2) Somewhat 9 (7.5) A little bit 1 (0.8) Not at all 0 (0.0) Missing/ No response 1 (0.8) Mean (SD) 1.58 (0.67) Chew et al (2008) validated the one question screener where somewhat, a little bit, or not at all scored as inadequate health literacy a 10 (8.3) How often do you have problems learning about your medical condition because of difficulty understanding written information a Always 0 (0.0)

PAGE 110

100 Variable N (%) Often 2 (1.7) Sometimes 11 (9.2) Occasionally 42 (35.0) Never 64 (53.3 ) Missing/ No response 1 (0.8) Mean (SD) 1.58 (0.74) Summary score 8 and below 112 (93.3) Summary score 9 and above 7 (5.8) Missing 1 (0.8) How good are you with fractions b 1 Not at all good 4 (3.3) 2 5 (4.2) 3 15 (12.5) 4 32 (26 .7) 5 38 (31.7) 6 Extremely good 26 (21.7) Missing/ No response 0 (0.0) Mean (SD) 4.44 (1.26) How good are you at figuring out how much a shirt will cost if it is 25% off b 1 Not at all good 2 (1.7) 2 3 (2.5) 3 5 (4.2) 4 18 (15.0) 5 41 (34.2) 6 Extremely good 51 (42.5) Missing/ No response 0 (0.0) Mean (SD) 5.05 (1.11) How often do you find numerical information to be useful b 1 Never 1 (0.8) 2 1 (0.8) 3 8 (6.7) 4 28 (23.3) 5 37 (30.8) 6 Very often 45 (37.5) Missing/ No response 0 (0.0) Mean (SD) 4.95 (1.04) Subjective Numeracy Summary Score b Mean (SD) 14.44 (2.82) If I had access to vaccine information, blogs and discussion boards to chat with other paren ts on a Kaiser Permanente vaccine information website, I think my level of concern about vaccines would Increase 17 (14.2)

PAGE 111

101 aChew, et al. (2008) validated one question for health literacy screening. bThe Subjective Numeracy Scale (SNS) has been validated as a 3 question scale to screen for health numeracy. The response scale uses numbers anchored by text at extreme ends 1=not at all to 6=extremely. Higher s cores indicate higher numeracy skills with a composite scale score ranging from 1 to 16. Descriptive statistics were used to characterize the study population. As seen in Table 4.1, study participants were, on average, twenty nine years of age, highly ed ucated, white and more likely to be first time moms. Validated subjective literacy and num eracy scales were used. The cohort had high rates of literacy and numeracy Next, each survey item was described over time, using frequency counts, percentages, and m eans with standard deviations. Table 4.2 shows the descriptive data. In general, the survey item mean scores reflect the direction of each question. For example, I believe there has not been enough research on the safety of vaccines has a mean score at b aseline of 2.58; two month mean score of 2.34; four month mean score of 2.47; and six month mean score of 2.50. These scores are within the range of disagree with the survey question. Repeated measures analyses of variance (ANOVA) were conducted on each survey item When sphericity was violated, correction was made. In Table 4.2, when there was a significant effect of time on survey item means, a symbol for p value cutoff of either .05 using * or <.001 using ** was indicated. For the concerns about vaccines survey items, it is just as interesting that these did not change over time as the other survey items that did have significant change over time. Survey items with concerns Variable N (%) Stay the same 78 (65.0) Decrease 25 (20.8) Missing/ No response 0 (0.0)

PAGE 112

102 about ingredients, autism and side effects stayed stable, statistically insignificant and within the unsure response selection. P arents think less ab out vaccines after the child is two months of age. Parents confidence in vaccine information and ability to protect their child from disease by vaccinating increases after the birth of the child. Pregnancy seems to be a time of uncertainty about childhood vaccines. Table 4.2. Means and Standard Deviations of Pilot Cohort, N=120 Survey Question Baseline N (%) N = 120 Child 2 mos. N (%) N=110 Child 4 mos. N (%) N=113 Child 6 mos. N (%) N=114 Generally I intend to do what my childs doctor recommends about vaccines for my child Strongly disagree 0 (0.0) 0 (0.0) 0 (0.0) 2 (1.7) Disagree 2 (1.7) 2 (1.7) 2 (1.7) 1(0.8) Neither agree or disagree 6 (5.0) 4 (3.3) 2 (1. 7) 3(2.5) Agree 54 (45.0) 42 (35.0) 51 (42.5) 54(45.0) Strongly agree 58 (48.3) 62 (51.7) 59 (49.2) 52(43.3) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 8 (6.7) Mean (SD) 4.39 (0.68 ) 4.47 ( 0 .66) 4.26 (0.68 ) 4.40 (0.69 ) Most of the p arents I know vaccinate their children Strongly disagree 0 (0.0) 0 (0.0) 0 (0.0) 1(0.8) Disagree 4 (3.3) 1 (0.8) 4 (3.3) 2(1.7) Neither agree or disagree 12 (10.0) 5 (4.2) 4 (3.3) 6(5.0) Agree 63 (52.5) 57 (47.5) 65 (54.2) 68(56.7) Strongly agree 41 (34.2) 47 (51.8) 41 (34.2) 35(29.2) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 8 (6.7) Mean (SD) 4.18 (0.74) 4.36 (.62) 4.25 (0.69) 4.20 (0.70) I have given a lot of thought about vaccinations for my child * Strongly disagree 2 (1.7) 1 (0.8) 2 (1.7) 1 (0.8) Disagree 17 (14.2) 5 (4.2) 2 (1.7) 7 (5.8) Neither agree or disagree 27 (22.5) 21 (17.5) 31 (25.8) 14 (11.7) Agree 38 (31.7) 42 (35.0) 44 (36.7) 62 (51.7) Strongly agree 36 (30.0) 4 1 (34.2) 35.0 (29.2) 28 (23.3) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 8 (6.7) Mean (SD) 2.26 (1.09) 1.93 ( 0 .91) 2.05 (0.90) 2.03 (0.84) Parents should have the right to refuse vaccines that are required for school for any reason Strongly disagree 15 (12.5) 15 (12.5) 17 (14.2) 14 (11.7)

PAGE 113

103 Survey Question Baseline N (%) N = 120 Child 2 mos. N (%) N=110 Child 4 mos. N (%) N=113 Child 6 mos. N (%) N=114 Disagree 47 (39.2) 47 (39.2) 42 (25.0) 47 (39.2) Neither agree or disagree 24 (20.0) 19 (15.8) 23 (19.2) 22 (18.3) Agree 23 (19.2) 25 (20.8) 23 (19.2) 24 (20.0) Strongly ag ree 10 (8.3) 4 (3.3) 9 (7.5) 5 (4.2) Missing/ No response 1 (0.8) 10 (8.3) 6 (5.0) 8 (6.7) Mean (SD) 2.71 (1.17 ) 2.60 (1.09) 2.63 (1.09 ) 2.69 (1.18) H ow confident are you that you will be able to protect your child from some types of infec tious disease by vaccinating him or her * Not at all confident 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) Slightly confident 2 (1.7) 2 (1.7) 1 (0.8) 4 (3.3) Somewhat confident 28 (23.3) 7 (5.8) 10 (8.3) 28 (23.3) Very confident 59 (49.2) 48 (40.0) 58 (48.3) 54 (45.0) Absolutely confident 31 (25.8) 53 (44.2) 45 (37.5) 27 (22.5) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 3.99 (0.75) 4.38 (0.69) 4.29 (0.66) 4.23 (0.63) H ow confident are you that you have the nec essary information to make decisions about vaccination for your child * Not at all confident 8 (6.7) 2 (1.7) 0 (0.0) 0 (0.0) Slightly confident 12 (10.0) 6 (5.0) 7 (5.8) 0 (0.0) Somewhat confident 43 (35.8) 27 (22.5) 20 (16.7) 12 (10.0) Very confident 36 (30.0) 42 (35.0) 61 (50.8) 63 (52.5) Absolutely confident 21 (17.5) 33 (27.5) 26 (21.7) 38 (31.7) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 3.42 (1.10) 3.89 (0.96) 3.93 (0.81) 3.92 (0.75) I am concerned that my childs immune system could be weakened by too many vaccines * Strongly disagree 14 (11.7) 16 (13.3) 13 (10.8) 12 (10.0) Disagree 52 (43.3) 55 (45.8) 64 (53.3) 67 (55.8) Neither agree or disagree 28 (23.3) 22 (18.3) 19 (15.8) 17 (14.2) Agree 23 (19.2) 16 (13.3) 16 (13.3) 14 (11.7) Strongly agree 3 (2.5) 1 (0.8) 2 (1.7) 2 (1.7) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 8 (6.7) Mean (SD) 2.58 (1.01 ) 2.37 (0.94 ) 2.41 (0.92 ) 2.35 (0.91 ) Vac cines strengthen the immune system Strongly disagree 1 (0.8) 1 (0.8) 2 (1.7) 0 (0.0) Disagree 9 (7.5) 10 (8.3) 3 (2.5) 11 (9.2) Neither agree or disagree 54 (45.0) 39 (32.5) 45 (37.5) 34 (28.3) Agree 45 (37.5) 45 (37.5) 50 (41.7) 5 7 (47.5)

PAGE 114

104 Survey Question Baseline N (%) N = 120 Child 2 mos. N (%) N=110 Child 4 mos. N (%) N=113 Child 6 mos. N (%) N=114 Strongly agree 10 (8.3) 15 (12.5) 14 (11.7) 11 (9.2) Missing/ No response 1 (0.8) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 3.45 (0.79) 3.57 (0.87) 3.62 (0.80) 3.60 (0.80) Getting vaccines is a good way to protect my child from infe ctious diseases Strongly disagree 0 (0.0) 1 (0.8) 0 (0.0) 0 (0.0) Disagree 0 (0.0) 0 (0.0) 1 (0.8) 0 (0.0) Neither agree or disagree 5 (4.2) 5 (4.2) 3 (2.5) 5 (4.2) Agree 77 (64.2) 61 (50.8) 65 (54.2) 68 (56.7) Strongly agree 3 7 (30.8) 43 (35.8) 45 (37.0) 40 (33.3) Missing/ No response 1 (0.8) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 4.27 (0.53) 4.32 (0.65) 4.35 (0.58) 4.31 (0.55) I believe there has not been enough research on the safety of vaccines Strongly disagree 11 (9.2) 18 (16.4) 16 (13.3) 14 (11.7) Disagree 50 (41.7) 51 (42.5) 49 (40.8) 50 (41.7) Neither agree or disagree 41 (34.2) 30 (25.0) 33 (27.5) 32 (26.7) Agree 15 (12.5) 8 (6.7) 11 (9.2) 13 (10.8) Strongly agree 3 (2.5) 3 (2.5 ) 5 (4.2) 4 (3.3) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 2.58 (0.91) 2.34 (0.93) 2.47 (1.00) 2.50 (0.97) I believe it is better for my child to develop immunity by getting sick than to get a shot Strongly disagree 21 (17.5) 30 (25.0) 26 (21.7) 24 (20.0) Disagree 60 (50.0) 50 (41.7) 67 (55.8) 65 (54.2) Neither agree or disagree 31 (25.8) 19 (15.8) 15 (12.5) 17 (14.2) Agree 8 (6.7) 8 (6.7) 6 (5.0) 6 (5.0) Strongly agree 0 (0.0) 2 (1.7) 0 ( 0.0) 1 (0.8) Missing/ No response 0 (0.0) 11 (9.2) 6 (5.0) 7 (5.8) Mean (SD) 2.22 (0.81 ) 2.01 (0.95 ) 2.01 (0.75) 2.05 (0.80 ) I believe it is better for my child to get the natural disease than to get a vaccine Strongly disagree 4 1 (34.2) 56 (46.7) 46 (38.3) 44 (36.7) Disagree 62 (51.7) 49 (40.8) 57 (47.5) 57 (47.5) Neither agree or disagree 16 (13.3) 4 (3.3) 9 (7.5) 9 (7.5) Agree 1 (0.8) 1 (0.8) 1 (0.8) 2 (1.7) Strongly agree 0 (0.0) 0 (0.0) 1 (0.8) 1 (0.8) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 1.81 (0.69) 1.55 (0.62) 1.72 (0.72) 1.75 (0.75) I am concerned that it would be

PAGE 115

105 Survey Question Baseline N (%) N = 120 Child 2 mos. N (%) N=110 Child 4 mos. N (%) N=113 Child 6 mos. N (%) N=114 painful for my child to receive so many shots during one doctors visit Strongly disagree 11 (9.2) 4 (3.3) 3 (2.5) 44 (36.7) Disagree 37 (30.8) 29 (24.2) 32 (26.7) 57 (47.5) Neither agree or disagree 23 (19.2) 23 (19.2) 28 (23.3) 9 (7.5) Agree 47 (39.2) 50 (41.7) 46 (38.3) 2 (1.7) Strongly agree 2 (1.7) 4 (3.3) 5 (4.2) 1 (0 .8) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 2.93 (1.07) 3.19 (0.99) 3.16 (0.97) 3.16 (0.95) I believe many of the illnesse s vaccines prevent are serious* Strongly disagree 1 (0.8) 0 (0.0) 0 (0.0) 1 (0.8) Disagree 0 (0.00) 1 (0.8) 0 (0.0) 0 (0.0) Neither agree or disagree 5 (4.2) 3 (2.5) 4 (3.3) 3 (2.5) Agree 57 (47.5) 40 (33.3) 48 (48.0) 51 (42.5) Strongly agree 57 (47.5) 66 (55.0) 62 (51.7) 58 (48.3) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 4.41 (0.65) 4.55 (0.60) 4.51 (0.57) 4.46 (0.64) I believe my child could get a serious disease if he or she were not vaccinated Strongly disagree 0 (0.0) 0 (0.0) 1 (0.8) 2 (1.7) Disagree 3 (2.5) 4 (3.3) 2 (1.7) 3 (2.5) Neither agree or disagree 16 (13.3) 13 (10.8) 11 (9.2) 7 (5.8) Agree 68 (56.7) 46 (38.3) 60 (50.0) 60 (50.0) Strongly agree 33 (27.5) 47 (39.2) 40 (33.3) 40 (33.3) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 8 (6 .7) Mean (SD) 4.09 (0.71) 4.24 (0.80) 4.19 (0.75) 4.19 (0.81) I believe vaccines are generally safe Strongly disagree 1 (0.8) 1 (0.8) 1 (0.8) 2 (1.7) Disagree 1 (0.8) 2 (1.7) 1 (0.8) 1 (0.8) Neither agree or disagree 10 (8.3) 7 (5.8) 8 (6.7) 6 (5.0) Agree 91 (75.8) 72 (60.0) 82 (68.3) 79 (65.8) Strongly agree 16 (13.3) 28 (23.3) 22 (18.3) 25 (20.8) Missing/ No response 1 (0.8) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 4.01 (0.57) 4.13 (0.68) 4.08 (0.61) 4.10 (0.68 ) Getting vaccinated will greatly reduce the chance of getting infectious diseases like polio or measles, but will not eliminate the chance of gett ing these diseases completely Strongly disagree 4 (3.3) 2 (1.7) 1 (0.8) 1 (0.8) Disagr ee 11 (9.2) 7 (5.8) 7 (5.8) 7 (5.8)

PAGE 116

106 Survey Question Baseline N (%) N = 120 Child 2 mos. N (%) N=110 Child 4 mos. N (%) N=113 Child 6 mos. N (%) N=114 Neither agree or disagree 14 (11.7) 11 (9.2) 9 (7.5) 13 (10.8) Agree 68 (56.7) 59 (49.2) 71 (59.2) 66 (55.0) Strongly agree 22 (18.3) 31 (25.8) 26 (21.7) 25 (20.8) Missing/ No response 1 (0.8) 10 (8.3 ) 6 (5.0) 8 (6.7) Mean (SD) 3.78 (0.97) 4.00 (0.90) 4.00 (0.80) 3.96 (0.82) Children get more vaccines than are good for them Strongly disagree 9 (7.5) 12 (10.0) 8 (6.7) 10 (8.3) Disagree 50 (41.7) 49 (40.8) 52 (43.3) 49 (40.8) Neither agree or disagree 41 (34.2) 34 (28.3) 41 (34.2) 41 (34.2) Agree 14 (11.7) 12 (10.0) 10 (8.3) 11 (9.2) Strongly agree 5 (4.2) 3 (2.5) 3 (2.5) 2 (1.7) Missing/ No response 1 (0.8) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 2. 63 (0.94) 2.50 (0.93) 2.54 (0.85) 2.52 (0.86) My child does not need vaccines for diseases that are not common anymore, like polio Strongly disagree 33 (27.5) 44 (36.7) 44 (36.7) 42 (35.0) Disagree 56 (46.7) 46 (38.3) 51 (42.5) 50 (41.7) Neithe r agree or disagree 26 (21.7) 15 (12.5) 14 (11.7) 16 (13.3) Agree 4 (3.3) 5 (4.2) 4 (3.3) 4 (3.3) Strongly agree 0 (0.0) 0 (0.0) 1 (0.8) 1 (0.8) Missing/ No response 1 (0.8) 10 (8.3) 6 (5.0) 7 (5.8) Mean (SD) 2.01 (0.80) 1.83 (0.83) 1.8 3 (0.84) 1.87 (0.85) I am concerned that there are serious side effects of vaccines Strongly disagree 8 (6.7) 9 (7.5) 5 (4.2) 5 (4.2) Disagree 39 (32.5) 42 (35.0) 42 (35.0) 45 (37.5) Neither agree or disagree 33 (27.5) 30 (25.0) 4 3 (35.8) 34 (28.3) Agree 34 (28.3) 23 (19.2) 22 (18.3) 23 (19.2) Strongly agree 6 (5.0) 4 (4.2) 2 (1.7) 5 (4.2) Missing/ No response 0 (0.0) 11 (9.2) 6 (5.0) 8 (6.7) Mean (SD) 2.93 (1.04) 2.75 (1.03) 2.77 (0.87) 2.80 (0.97) I am concerned that some vaccines cause autism in healthy children Strongly disagree 23 (19.2) 27 (22.5) 27 (22.5) 25 (20.8) Disagree 42 (35.0) 37 (30.8) 37 (30.8) 34 (28.3) Neither agree or disagree 36 (30.0) 29 (24.2) 32 (26.7) 38 (31.7) Agree 15 (12.5) 14 (11.7) 15 (12.5) 11 (9.2) Strongly agree 4 (3.3) 3 (2.5) 3 (2.5) 4 (3.3) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 8 (6.7) Mean (SD) 2 .45 (1.0 4) 2 .36 ( 1.07 ) 2.39 (1.07 ) 2.41 (1.05 )

PAGE 117

107 Survey Question Baseline N (%) N = 120 Child 2 mos. N (%) N=110 Child 4 mos. N (%) N=113 Child 6 mos. N (%) N=114 I am concerned that the ing redients in vaccines are unsafe Strongly disagree 8 (6.7) 11 (9.2) 11 (9.2) 10 (8.3) Disagree 64 (53.3) 60 (50.0) 64 (53.3) 58 (48.3) Neither agree or disagree 28 (23.3) 20 (16.7) 18 (15.0) 23 (19.2) Agree 15 (12.5) 12 (10.0) 15 (12 .5) 15 (12.5) Strongly agree 5 (4.2) 5 (4.2) 4 (3.3) 6 (5.0) Missing/ No response 0 (0.0) 12 (10.0) 8 (6.7) 8 (6.7) Mean (SD) 2.54 (0.94) 2.44 (0.97) 2.44 (0.97) 2.54 (1.01) I am concerned about serious infectious diseases like whooping cough or measles* Strongly disagree 0 (0.0) 0 (0.0) 0 (0.0) 1 (0.8) Disagree 5 (4.2) 2 (1.7) 3 (2.5) 3 (2.5) Neither agree or disagree 12 (10.0) 4 (3.3) 8 (6.7) 10 (8.3) Agree 70 (58.3) 57 (47.5) 67 (55.8) 60 (50.0) Strongl y agree 33 (27.5) 47 (39.2) 36 (30.0) 37 (30.8) Missing/ No response 0 (0.0) 10 (8.3) 6 (5.0) 9 (7.5) Mean (SD) 4.09 (0.73) 4.35 (0.64) 4.19 (0.68) 4.16 (0.77) No te: 1= Strongly disagree to 5= S trongly agree *=p<.0 01; **=p<.05 Column perce ntages Finally, to further describe the cohort survey items that were categorical Cochran Armitage test of trends and Friedman tests were conducted to see if there were significant differences in responses over time (Tables 4.3 and 4.4). Parents self r eported endorsement of intention to vaccinate their child increases over time. Seventy four percent of the cohort endorsed intentions to fully vaccinate their child at baseline. When the child was six months of age, this increased to 92% endorsement. It was interesting that approximately 24% of the cohort endorsed knowing someone who had a bad reaction to vaccines. There were no significant changes over time in this survey item. Another significant change over

PAGE 118

108 time was the influence of the physician on vaccine decision. This influence of the physician increased steadily over time (Table 4.3) Table 4.3 Trends Over T ime of C ohort Using CochranArmitage, N= 120. Baseline N(%) a 2 month N(%) a 4 month N(%) a 6 month N(%) a z sore p value Self reported intentions about the decision to vaccinate their child Fully vaccinate 89 (74.17) 98 (89.09) 105 (92.11) 104 (92.04) 4.07 < .001 Some vaccines 15 (12.50) 7(6.36) 7 (6.14) 6 (5.31) 2.01 <0.05 Delay vaccines 4 (3.33) 3 (2.73) 2 (1.75) 2 (1.77) 0.89 0.37 U ndecided 12 (10.00) 2 (1.82) 0 (0.00) 1 (0.88) 4.01 < .001 Self reported knowing someone who had a bad reaction to vaccines 16 (23.88) 21 (31.34) 16 (23.88) 14 (20.90) 4.50 0.62 Self reported influence on the decision about vaccinating their child S pouse/partner 57 (28.50) 52 (26.00) 50 (25.00) 41 (20.50) 1.78 0.07 Family member 4 (19.05) 5 (23.81) 7 (33.33) 5 (23.81) 0.62 0.53 Friend 1 ((25.00) 1 (25.00) 1 (25.00) 1 (25.00) 0.03 0.97 Medical provider 22 (15.60) 34 (24.11) 37 (26.24) 48 (34.04) 3.88 < .001 Alternative medical provider 1 (100.00) 0 (0.0) 0 (0.0) 0 (0.0) 1. 32 0.19 No one 31 (43.06) 13 (18.06) 14 (19.44) 14 (19.44) 2.70 0.007 Other b 4 (23.53) 4 (23.53) 5 (29.41) 4 (23.53) 0.18 0.86 aT0 = 120; T1 = 110; T2 = 11 4; T3 = 113 bOther categories: T0: childs illness (2), book (1), own medical training (1); T1: book (1), own medical training (1), own research (2), physician friend (1); T2: own research (3), physician friend (1), own medical training (1); T3: on cologist (1), own research (2). The use of the internet to seek out health information decreased over time while use of the internet to search for vaccine information increased w hen the child was born through four months and then decreased to baseline use at six months of age of child (Table 4.4) Parents preference of vaccine decision making changed over time. At baseline, parents endorsed more intended collaboration with their childs physician in making decisions about vaccines. This changed to endorsement of making their own decision about vacc ines for their child over time (Table 4.4)

PAGE 119

109 Table 4.4 Trends Over T ime of C ohort Using CochranArmitage, N= 120. Baseline N(mean rank) 2 month N(mean rank) 4 month N(mean rank) 6 month N(mean rank) df Chi square p valu e Use of the internet in general 105 (2.57) 105 (2.47) 105 (2.52) 105 (2.44) 3 10.02 .018 Use of internet for health information 104 (2.73) 104 (2.64) 104 (2.45) 104 (2.18) 3 21.27 < .001 Use of the internet to search for vaccine information 104 (2.24) 104 (2.66) 104 (2.75) 104 (2.35) 3 18.65 < .001 Use of social media in general 105 (2.46) 105 (2.61) 105 (2.50) 105 (2.43) 3 3.44 .329 Use of the social media to search for health information 106 (2.49) 106 (2.45) 106 (2.68) 106 (2.38) 3 4.88 181 Use of the social media to search for vaccine information 105 (2.38) 105 (2.49) 105 (2.61) 105 (2.52) 3 4.60 .204 Preference of vaccine decision making 106 (2.83) 106 (2.43) 106 (2.39) 106 (2.34) 3 29.50 < .001 Note: scale ranges from 1=not at all to 5=every day for internet and social media questions; 1=prefer to make all vaccine decisions to 5=prefer to leave all vaccine decisions to my childs doctor. Summary of Longitudinal Findings These descriptive results provide valuable information on changes over tim e of parents concerns, attitudes, beliefs and intentions about childhood vaccines. This feasibility study involved administration of a survey and measured childhood vaccine concerns from pregnancy through 6 months of age of the child. By captur ing the vac cine attitudes and beliefs prospectively, recall bias was minimized. This p rovided data that was near real time to the decision about childhood vaccines. In this cohort, 74% of the pregnant mothers intended to get all v accines on time for their child The intention to vaccinate their child increased significantly from baseline to after the child was born. Confidence in vaccine knowledge increased in the parent cohort over time. This investigation also describes the uncertainty parents experience about vacc ines when they were pregnant and its

PAGE 120

110 subsequent decrease after the birth of their child These data suggest that the time to implement interventions is during pregnancy, when uncertainty is at its highest. Using trends over time to analyze the survey item s, concerns about vaccine preventative disease increased after the birth of the child This is expected, as a higher proportion of parents in the cohort were firsttime parents without previous experience with making decisions about childhood vaccines In addition, the influence of a physician on the parents vaccine decision making increased over time. These results suggest that pregnant mothers would benefit from intervention prior to the birth of the child, when concerns are elevated. Parents use of th e internet for vaccine information increased from pregnancy to two through four months of age of the child and then decreased at six months age of the child. This information is valuable when designing interventions for parents concerns about vaccines for their child. Providing faceto face and online interventions during pregnancy through the first year of life of the child may be optimal in timing. However, these pilot results suggest that parents significantly decrease use of the internet by 6 months of age of the child that suggests early intervention is optimal. Interestingly, concerns about vaccine ingredients, side effects and autism related to childhood vaccines stayed statistically insignificant over time. This may reflect parents ongoing concern s about these issues. This information is useful to interventions that are designed to reduce vaccine concerns and increase overall vaccination rates. These enduring concerns are target content areas in which

PAGE 121

111 researchers can develop tailored information fo r the parent in an effort to reduce these concerns. By designing well planned interventions that target pregnant women and their partners, uncertainty about vaccines can be reduced prior to the birth of the child. This would increase the parents confidenc e in their vaccine decisions and help parents to access information on topics such as immunology and the childhood vaccine schedule. By reducing parents concerns and increasing knowledge about childhood vaccines during pregnancy and weeks after the child s birth, more time could be focused on other developmental screening s in the pediatricians office and less time on v accine education of the parent. One of the most significant process related learning was that it is important to continuously screen for f etal demise and abortion during recruitment and survey administration. Although rare, it is necessary to stop research activities when adverse events occur. By conducting this feasibility study in a setting that has a robust electronic health record and data system, there was opportunity to develop protocol for screening that will benefit future studies. By knowing more about the vaccine decision making of parents a stronger evidence base for policy and practice is established. Studies in parents knowled ge, attitudes and beliefs about childhood vaccines have been characterized as using parent recall after the act of vaccination happens. Oftentimes surveys ask parents to recall events, feelings, and reasons months and even years after vaccination has occu rred Using a prospective longitudinal design provided new information on how parents concerns, attitudes, beliefs and intentions about childhood vaccines change

PAGE 122

112 over time. Although the sample was small, these data provide a description of mothers decision making about vaccines for their child from pregnancy to six months after the childs birth. This current investigation has limitations. First, the sample size was small and located in one geographical area. However, as it was a feasibility study, the small sample size allowed for the ca reful design of recruitment and survey administration that can be implemented in larger research projects. Second, there is the possibility of response bias in taking multiple surveys over a short period of time (Litwin, 1995) However, longitudinal surveys are a paradox. They serve as strong indicators of change over time by using the same survey items at each time. A t the same time, longitudinal desig n is criticized for bias, as individuals may recall previous questions and the answers they selected. This potential bias was reduced by designing at least two months time between each survey. Future studies may benefit from the use of survey item randomization, where survey questions or response sets are produced in a rando m order for each new administration of the survey. This further reduces bias from repeated survey measurement (Litwin, 1995). There are several next steps from the findings of this feasibility study. First, surveying parents at the twelfth month of the c hild s life would provide additional insight into the vaccine decisionmaking process of parents. Second, expanding both the diversity of the population and geographical diversity to assess the operab ility of the longitudinal survey outside of an integrated he alth c are system is important. B y testing the survey design on other target populations, the survey instrument and

PAGE 123

113 its administration can be further evaluated for use. Finally, larger sample size would permit additional statistical procedures and provide additional information on the decision making process of parents. In this chapter, the Pilot cohort data was analyzed for trends over time. It was reassuring that concerns about childhood vaccines decreased over time, especially after the birth of the child. However, there were enduring concerns about vaccines that can inform the content development of interventions. These concerns were around the safety of childhood vaccines: vaccines cause developmental conditions in ch ildren such as autism; ingredients in vaccines are unsafe; and vaccines have harmful side effects. In the next chapter the data was analyzed using three different measures of vaccine hesitancy. The Pilot cohort was used in the primary analysis. As a confirmation, the Primary cohort data, which used the finalized version of the survey, was analyzed using two measures of vaccine hesitancy. These analyses were conducted to establish criterion validity of the instrument.

PAGE 124

114 CHAPTER V COMPARING THREE MEASURES OF VACCINE HESITANCY I n the preceding chapter, the Pilot cohort survey data was analyzed for trends over time. As parents transition from pregnancy to parenthood, concerns about vaccines decrease and parents rely more on their childs physician for vaccine information. However, there were three areas of concerns about vaccines that did not change over time: concerns about autism, ingredients, and side effects from vaccines. In this chapter, the Pilot and Primary cohort survey items were compared across different measures of vacc ine hesitancy. This step was taken to validate the results of the survey using the CA BI V with different measures of hesitancy to compare if and how they differ in detecting hesitancy in the CABI V. Vaccine Hesitancy Parents are expressing concerns about c hildhood vaccines to pediatricians, policymakers, and through media outlets (Benin, et al 2006; Freed, et al 2010) In particular, parents have concerns about the childhood vaccine schedule. Although specific concerns shift over time, the main categories of concern focus on the timing and number of doses of vaccines given in the first two years of life, the ingredients in vaccines and potential cumulative effects of ingredients, and that the childs immune system may be overwhelmed by too many vaccines (Gust, et al., 2005; Freed, et al., 2010) There have been numerous cross sectional survey investigations that have measured parents concerns (Gellin, et al 2000; Freed, et al 2010; 2011; Dempsey,

PAGE 125

115 et al 2011; Kennedy, et al 2011) Parents who do vaccinate their child according to the recommended vaccine schedule differ from parents who deviate from the schedule or do not vaccinate at all. These differences include information sources to make vaccine decisions, trust of institutions and the information they provide, and education level of the parent (Downs, et al 2008; Freed, et al ., 2010) However, parents who vaccinate their child per the recommended vaccine schedule also have similar concerns. Vaccine hesitancy has been defined in various ways ( Barrows, Coddington, Richards, & Aaltonen, 2015; MacDonald, 2015 ; Witteman, 2015) The broadest definition uses subjective attitudes about vaccines as the metric. This is measured through qualitative measures (i.e. themes from focus groups) or quantitative measures (i.e. survey items). The most restrictive use is through measure of vaccine behavior. This is measured through objective measures (i.e. metric of how many vaccine doses were administered or ICD 9 diagnostic code of hesitancy). The International Vaccine Hesitancy Working Group defines vaccine hesitancy a delay in acceptance or refusal of vaccines despite availability of vaccine services. Vaccine hesitancy is complex and context specific, varying across time, place and vaccines. It is influenced by factors such as complacency, convenience and confidence (WHO SAGE Working Group, 2014, p. 7 ) Although there have been many single application survey questions that measure a pointintime study question, there are few validated measures of vaccine he sitancy available. Opel, et al (2011a,b; 2013b) have validated a brief screener that can be easily administered to parents in the clinical setting The Parent Attitudes

PAGE 126

116 about Childhood Vaccines (PAC V) is a fifteen item screening instrument. However, its use in multiple settings, larger samples and with diverse populations has been limited and results of these limited uses are not yet published. Survey instruments that can be administered across time and in multiple studies are needed. Given the increasing concerns of parents about vaccines, c omprehensive examination of current measures of hesitancy provides additional information on survey items and other factors of decision making about vaccines The p urpose of this investigation was twofold. T he first p urpose was to characterize a cohort of pregnant mother s through three hesitancy measures using demographic and descriptive survey items The second purpose was to compare the performance of the hesitancy measures on concerns, attitudes, beliefs and intentions about vaccines survey items. This study used the baseline survey data from the Pilot cohort of 120 pregnant women and Primary cohort of 320 pregnant women and parents of children under twelve months of age. Methods Design To address the need for further descriptive information about vaccine hesitancy measures, the ba seline data from the Pilot cohort of 120 pregnant women was used Three measures of hesitancy were used to assess the survey items : PAC V screening instrument, vaccine intention question, and average days undervaccinated (ADU). These measures are described below.

PAGE 127

117 Setting and Study Population The research was conducted between July 2012 and December 2014 at Kaiser Permanente Colorado (KPCO). The study uses the survey data from the Pilot coh ort and the Primary cohort. The cohort development, enrollment, and recruitment methods have been described in previous chapters. Analytic Methods. Three Measures of Vaccine Hesitancy There were three measures of vaccine hesitancy used to assess demo graph ics and survey items. The three measures are described below. Hesitancy can be viewed through intentions and behaviors. Behaviors are the gold standard as they represent the actual action (outcome) of the parents decisionmaking about vaccines In this analysi s, the broad term of vaccine hesitancy is used in the following ways: 1. Parents subjective responses that they intend not to vaccinate their child according to the recommended childhood schedule. 2. Parents summed score from a validated screening instr ument of vaccine hesitancy that uses numerous survey items and aggregates the score. 3. The metric of average days under vaccinated, a scale variable that measures objective vaccine behavior. Int ention to Vaccinate Survey Item A one question intention to vac cinate item was developed for the study As a parent, I intend to A sevenitem response is provided: get all vaccines on time as recommended by my childs doctor; get some vaccines recommended by my

PAGE 128

118 childs doctor with the chance I might change my m ind in the future and get all vaccines recommended; get some vaccines recommended by my childs doctor and delay other vaccines until my child is older; delay vaccines and wait until my child is older; not vaccinate my child with the chance I might ch ange my mind in the future and get vaccines recom mended by my childs doctor; n ot vaccinate my child and not get any vaccines recommended by my childs doctor; and undecided about my decision about vaccines. Appendix C contains all survey questions and response items. This was an extensive list generated for testing in the P ilot cohort. These responses were based on parents expressing the need and desire for more categories on surveys during focus groups conducted for research development. Responses we re dichotomized into 0 vaccinate according to the recommended childhood schedule and 1, not vaccinate according to the recommended childhood schedule. For the latter, this includes responses of getting some, none, delaying vaccines as well as undecided responses for pregnant participants who had not yet made their vaccine decision for their child. Parent Attitude about Childhood Vaccines (PAC V) Screener Parent Attitude ab out Childhood Vaccines (PAC V) is a fifteenitem survey develop ed by Opel, et al (2011a,b; 2013b) to screen for parental hesitancy about childhood vaccines. It has been validated in one regional clinical setting, demonstrated high reliability, and shown to predict childhood vaccine status. In collaboration with the developer o f the PACV, Douglas Opel, MD, the investigator modified this screener to operate as a prospective hesitancy tool. This was achieved by re wording the questions for future decisions and intentions, rather than using it

PAGE 129

119 in its current format of retrospective surv ey questions. The PAC V was administered over the phone to parents who consented to the study, and prior to parents taking the paper and pencil baseline CABI V survey. The PAC V uses three different response scales: Likert scale, dichotomous, and an elevenpoint scale with responses ranging from 0 not sure at all to 10 complet ely sure. It is brief, taking five minutes or less to complete and has a readability index of 6th grade. Responses are transformed to a threepoint scale (0, 1, or 2) and summed. T he range of scores is 0 to thirty Scores of fifteen or above are considered hesitant. Average Days Undervaccinated (ADU) Average days under vaccinated (ADU) is a metric developed by Glanz, et al (2013a). ADU determines under vaccination based on each vac cine dose. ADU calculates the difference between when a vaccine dose was administered and when a vaccine dose should have been administered according to the recommended vaccine schedule. In this study, these differences are combined across all vaccines tha t are due from birth to 200 days, and form the metric of ADU. Analyses were based on participants who had continuous enro llment in KPCO allowing for a thirty one day gap. This provided access to all vaccine information available and accurately determined t he childs ADU status. For this study ADU was used as a dichotomous variable, yes or 1 for ADU value greater than 0 or no or 0 for ADU value at 0. Analyzing the Three Hesitancy Measures For categorical variables, Chisquare test of independence was us ed. This statistical test is used when categorical variables have two or more possible values.

PAGE 130

120 Chi square test shows whether the proportion for one variable differs among the values of the other variables. When expec ted cell counts were less than five, Fis hers Exact T est of significance was used. For continuous variables, i ndependent sample ttest of significance was used. The i ndependent ttest for equal means determines if tw o distributions of data are significantly different from one another. L evenes u nequal variances test of significance was used for all ttest results, as the population variance was unknown and assumed unequal or different from one to the other. P values of <.05 were considered statistically significant. Analytical Steps First chi sq uare and ttest statistics were used to compare the demographics of cohort members who were and were not vaccine hesitant as measured by each of the three hesitancy measures. Then, each of the three vaccine hesitancy measures to screen survey items develop ed for the CABI V survey instrument. For this step, independent ttests were conducted to understand if there were significant differences between survey response averages in hesitant and nonhesitant groups across the three hesitancy measures. In a n additional analysis, the Primary cohort of pregnant women and par ents of children younger than twelve months of age used two measures of hesitancy. These two measures were ADU and self reported vaccine intentions for their child. I ndependent ttests were conduc ted to examination whether there were d ifferences in the means of parents who self reported intentions not to vaccinate their child and who intended to vaccinate their child This was repeated for the ADU measure.

PAGE 131

121 Results Results from the Pilot C ohort The Pilot cohort of 120 pregnant mothers was used for these analyses. The number of hesitant parents in each vaccine hesitancy measure varied. For the PACV screener, there were fifteen (12.5%) hesitant scores out of 120. For the self reported vaccine intention question, there were thirty one (25.8%) hesitant responses out of 120. Finally, for the average days undervaccinated metric, eight (10%) children were undervac cinated before 200 days of age For the ADU metric, eighty children from the 120 participants had continuous enrollment in KPCO through age 200 days, and only these children were included in the ADU metric analyses. This represents approximately onethird attrition for determining undervaccinated status. T he characteristics of cohort members who were and were not vaccine hesitant were measured by each of the three measures used in the study (Tables 5.1, 5.2, 5.3) Descriptively, the one survey item that was statistically significant across all three measures was personally knowing someone who had a bad reaction to a vaccine. There were significantly higher numbers of parents who knew someone who had a bad reaction in the undervaccinated group across all three measures of vaccine hesitancy.

PAGE 132

122 Table 5.1. Characteristics of the Cohort Using PAC V Screener Grouping V ariable, N=120 Opel PAC V screener hesitancy group Characteristic Not hesitant N=105 Hesitant N=15 p value Age M (SD) 29.19 (4.00) 30.60 (3.27) 0.196 Race (white) N(%) 97 (92.4) 13 (86.7) 0.612 College education N(%) 92 (89.3) 14 ( 93.3) 0.710 First time parent N(%) 71 (67.6) 9 (60.0) 0.558 Subjective vaccine decision to not vaccinate N(%) 17 (16.2) 14 (93.3) <.001 Prefer to make vaccine decision myself N(%) 9 (8.6) 5 (33.3) 0.016 Physician influences vaccine decision N(%) 20 (19 .0) 2 (13.3) .736 Average days undervaccinated > 0 N(%) 5 (6.8) 3 (20.0) 0.019 Personally know someone who had a vaccine reaction N(%) 8 (7.6) 8 (53.3) <.001 Table 5.2 Characteristics of the Cohort Using Self Reported Vaccine I nt ention Grouping Variable, N=120 Self reported vaccine intention Characteristic Fully vaccinate N=89 Delay, Decline or Uncertain N=31 p value Age M (SD) 29.71 (3.22) 29.25 (4.17) 0.575 Race (white) N(%) 82 (92.1) 28 (90.3) 0.717 College education N(%) 76 (87.4) 30 (96.8) 0.180 First time parent N(%) 55 (61.8) 25 (80.6) 0.055 Opel PAC V hesitancy screener as hesitant N(%) 1 (1.1) 14 (45.2) <.001 Prefer to make vaccine decision myself N(%) 9 (10.1) 5 (16.1) .351 Physician influences vaccine decision N(%) 18 (20.2) 4 (12 .9) .364 Average days undervaccinated > 0 N(%) 5 (6.8) 3 (9.68) .405 Personally know someone who had a vaccine reaction N(%) 7 (7.9) 9 (29.0) .005

PAGE 133

123 Table 5.3. Characteristics of the Cohort Using Average Days Undervacc inated G roupin g Variable, N=80 A verage days undervaccinated Characteristic ADU=0 N=72 ADU>0 N=8 p value Age M (SD) 29.64 (3.94 ) 31.00 (2.51 ) .343 Race (white) N(%) 67 (93.1) 7 (87.5) .489 College education N(%) 65 (91.5) 8 (100) .232 First time parent N(%) 55 (61.8) 25 (80.6) .055 Opel PAC V hesitancy screener as hesitant N(%) 4 (5.6) 3 (37.5) .019 Prefer to make vaccine decision myself N(%) 7 (9.7) 3 (37.5) .057 Physician influences vaccine decision N(%) 17 (23.6) 0 (0.0) .192 Subjective vaccine decision to not vaccinate N(%) 1 7 (23.6) 3 (37.5) .405 Personally know someone who had a vaccine reaction N(%) 5 (6.9) 4 (50.0) .004 M=Mean; SDStandard Deviation Note: column percentages Next, the data from the three vaccine hesitancy measures and their significance to distinguish differences in mean scores for each CABI V survey item was compared Independent ttest tables were scanned for significance across all three vaccine hesitancy measures. Then, the data was scanned for significance across two measures: PAC V screener and subj ective vaccine decision. For this se condary scan, the objective measure of undervaccination was not considered as it had a smaller cohort size after applying the criteria of enrollment and a corresponding smaller number of those with undervaccination. The results are presented in a summary table (Table 5.4). As can be seen, there were six questions th at were significant across all three vaccine hesitancy measures (shaded grey) For the secondary scan, an additional thirteen questio ns were significant across the two vaccine hesitancy measures (PAC V and self reported intentions). These are shaded lighter grey. This resulted in a total of nineteen out of twenty three questions

PAGE 134

124 Table 5.4. Three Me asures of Vaccine Hesitancy by Survey I tems PAC V N=120 Self reported intention N=120 Average days undervaccinated N=80 Hesitant N=15 Not hesitant N=105 Not fully vaccinate N=31 Fully Vaccinate N=89 ADU >0 N=8 ADU= 0 N=72 p value Survey item M M p M M p M M p SUBJECTIVE NORMS ABOUT VACCINATING b Generally I int end to do what my childs doctor recommends about vaccines for my child 3.47 4.53 <.001 3.74 4.63 <.001 3.88 4.51 <.001 I have given a lot of thought about vaccinations for my child 4.27 3.67 0.045 3.68 3.76 0.704 4.63 3.60 <.001 Most of the parents I kn ow vaccinate their children 4.20 4.17 0.890 4.00 4.24 0.127 3.38 4.51 0.004 PERCEIVED CONTROL OVER VACCINE DECISION b How confident are you that you have the necessary information to make decisions about vaccination for your child 2.67 3.52 0.004 2.61 3.7 0 <.001 3.88 3.40 0.226 Parents should have the right to refuse vaccines that are required for school for any reason 3.93 2.56 <.001 3.27 2.53 0.002 2.88 2.63 0. 562 How confident are you that you will be able to protect your child from some types of infe ctious disease by vaccinating him or her 3.53 4.06 0.065 3.48 4.17 <.001 3.50 4.11 0.020 EVALUATION OF VPD/VAE b I am concerned that there are serious side effects of vaccines 4.13 2.75 <.001 3.77 2.63 <.001 4.00 2.71 0.001 I am concerned about serious i nfectious diseases that are not common anymore, like polio 3.73 4.14 0.043 3.74 4.21 0.002 4.11 4.12 0.958 I believe my child could get a serious disease if he or she 3.67 4.15 0.013 3.68 4.24 <.001 3.88 4.17 0.232

PAGE 135

125 PAC V N=120 Self reported intention N=120 Average days undervaccinated N=80 Hesitant N=15 Not hesitant N=105 Not fully vaccinate N=31 Fully Vaccinate N=89 ADU >0 N=8 ADU= 0 N=72 p value Survey item M M p M M p M M p were not vaccinated I am concerned that some vaccines cause autism in healthy children 3.67 2.29 <.001 3.29 2.17 <.001 3.00 2.33 0.100 My child does not need vaccines for diseases that are not common anymore, like polio 2.50 1.94 0.013 2.43 1.87 0.001 1.88 1.93 0.849 I believe vaccines are ge nerally safe 3.33 4.11 0.009 3.55 4.17 <.001 3.13 4.14 0.849 I believe many of the illnesses vaccines prevent are serious 4.00 4.47 0.009 4.10 4.52 0.001 4.25 4.43 0.485 Children get more vaccines than are good for them 4.07 2.44 <.001 3.50 2.34 <.001 3. 38 2.49 0.120 I am concerned that the ingredients in vaccines are unsafe 3.93 2.34 <.001 3.45 2.22 <.001 3.50 2.42 0. 068 Getting vaccinated will greatly reduce the chance of getting infectious diseases like polio or measles, but will not eliminate the ch ance of getting these diseases completely 3.64 3.80 0.570 3.70 3.81 0.596 4.00 3.72 0.241 BELIEFS ABOUT VACCINES b I believe it is better for my child to develop immunity by getting sick than to get a shot 3.13 2.09 <.001 2.90 1.98 <.001 2.88 2.10 0.009 I am concerned that my childs immune system could be weakened by too many vaccines 3.87 2.39 <.001 3.55 2.24 <.001 3.50 2.33 0.040 Vaccines strengthen the immune system 2.57 3.57 <.001 2.93 3.63 <.001 2.88 3.58 0.014 Getting vaccines is a good way to pr otect my child from infectious diseases 3.79 4.33 <.001 3.93 4.38 <.001 3.87 4.33 0.026 I believe there has not been enough research on the 3.40 2.46 <.001 3.03 2.42 0.001 3.13 2.50 0.297

PAGE 136

126 PAC V N=120 Self reported intention N=120 Average days undervaccinated N=80 Hesitant N=15 Not hesitant N=105 Not fully vaccinate N=31 Fully Vaccinate N=89 ADU >0 N=8 ADU= 0 N=72 p value Survey item M M p M M p M M p safety of vaccines I believe it is better for my child to get the natural disease than to get a vaccine 2.60 1.70 <.001 2.39 1.61 <.001 2.13 1.75 0.134 I am concerned that it would be painful for my child to receive so many shots during one doctors visit 3.20 2.90 0.303 3.23 2.83 0.076 2.88 2.93 0.883 M=Mean; p=p valu e a1=strongly disagree to 5= strongly agree bTheoretical constructs confirmed with Cronbach Alpha Note: Column percent Additional Analyses The purpose of the additional analyses was to determine if the two different measures of hesitancy used in the Primary cohort were significant for each survey question used in the survey constructs. This would give additional evaluation evidence that the survey items perform as expected in their respective constructs. In the additional analyses using the Primary cohort of 239 pregnant women and parents of young children, eigh teen of the twenty three questions had significant mean differences in hesitancy in both vaccine hesitancy measures. In the analysis using ADU, twenty seven were undervaccinated. Using the self reported vaccine intention response, there were fifty two parents who intended not to follow the recommended childhood vaccine schedule. T he analysis was limited to only those who had ADU value to provide comparable results acros s the two measures of hesitancy.

PAGE 137

127 Table 5.5. Two Me asures of Vaccine Hesitancy by Survey Items, n=23 9 Average Days Undervaccinated Self reported vaccine intention Survey itemb ADU=0 N=212 ADU>0 N=27 Intend to get vaccines N=187 No intention to vaccinate N=52 p value M M p M M p SUB JECTIVE NORMS ABOUT VACCINATING a In general, most of my close friends have similar beliefs about vaccines as me 3.60 3.11 .001 3.65 3.17 <.001 In general, my family (e.g. sisters, brothers and cousins) have similar beliefs about vaccines as me 3.88 3.11 .004 3.96 3.19 <.001 In general, my parents have similar beliefs about vaccines as me 4.00 3.11 .001 4.10 3.17 <.001 In general, my obstetrician/childs pediatrician has similar beliefs about vaccines as me 3.82 2.70 <.001 3.96 2.75 <.001 In general, my spouse or partner has similar beliefs about vaccines as me 4.23 4.07 .284 4.28 3.96 <.001 PERCEIVED CONTROL OVER VACCINE DECISION a How confident ar e you about your knowledge of how vaccines work 3.19 3.04 .448 3.25 2.90 0.03 How confident are you that you have the necessary information to make decisions about vaccination for your child 3.34 3.37 .883 3.45 2.94 <.001 How confident are you that you can express your vaccine views to your childs pediatrician 3.75 3.52 .197 3.81 3.42 0.01 How confident ar e you about your knowledge of infectious diseases 3.10 3.11 .951 3.10 3.10 0.97 EVALUATION OF VPD/VAE a Vaccines are safe 3.86 3.00 <.001 3.95 3.10 <.001 Children get more vaccines than they need 2.71 4.04 <.001 2.61 3.75 <.001 I am concerned that t he i ngredients in vaccines are unsafe 2.70 3.96 <.001 2.57 3.81 <.001 I am concerned that s ome vaccines cause autism in healthy children 2.49 3.30 .001 2.36 3.38 <.001 I am concerned that vaccines have serious side effects 2.68 3.89 <.001 2.55 3.79 <.001 I believe t here has not been enough research on the safety of vaccines 2.69 3.89 <.001 2.57 3.77 <.001 I believe it is better for my child to get the natural disease than to get a vaccine 1.79 2.67 <.001 1.67 2.65 <.001 I am concerned that m y childs 2.48 3.74 <.001 2.38 3.48 <.001

PAGE 138

128 Average Days Undervaccinated Self reported vaccine intention Survey itemb ADU=0 N=212 ADU>0 N=27 Intend to get vaccines N=187 No intention to vaccinate N=52 p value M M p M M p immun e system could be weakened by too many vaccines BELIEFS ABOUT VACCINES a How confident are you that you will be able to protect your child from some types of infectious disease by vaccinating him or her 3.80 3.00 <.001 3.88 3.10 <.001 My child will not need vaccines for diseases that are not common anymore, like polio 1.95 2.52 .013 1.91 2.37 <.001 My child could get a serious disease if he or she were not vaccinated 4.38 3.89 .029 4.41 4.02 <.001 Vaccines strengthen the immune system 3.62 2.96 <.001 3.66 3.13 <.001 Many of the illnesses vaccines prevent are serious 4.14 3.70 .003 4.18 3.75 <.001 Getting vaccines is a good way to protect my child from infectious diseases 4.27 3.59 .003 4.30 3.81 <.001 aEFA constructs co nfirmed with Conbach Alpha and CFA b1=strongly disagree to 5= strongly agree Note: Column percent Discussion This study used des criptive statistics to examine three different measures of hesitancy in a cohort of pregnant women. An additional analysis inclu ded a larger cohort of pregnant women and parents of children less than twelve months of age. Using survey items, dichotomous measures of hesitancy assessed differences between groups. The survey items in both cohorts performed well in producing means that were significantly different between those in the hesitancy group and those in the nonhes itant group. The survey items acted as expected ; those survey items that a parent who was hesitant was endorsed as such. By ass essing the survey items across three d ifferent hesitancy measures, this assessment ensured that the

PAGE 139

129 survey items are assessing what they are intended to measure. The PAC V hesitancy measure is a gold standard validated measure (Opel, 2013b) and performed similarly to the oneitem intention q uestion. The PAC V is undergoing reduction in items, but it may be more efficient and less burdensome to use the oneitem intention question. Larger scale evaluations of these measures are needed. The survey item Do you personally know someone who had a bad vaccine reaction was significant across all three hesitancy measures. This is suggests that parents are influenced by their social networks. Knowing if parents have had previous experience with vaccine adverse events through others is a relatively eas y inquiry within the clinical setting and may help physicians communicate risk associated with vaccine doses into perspective for the parent. It is important to develop measures of hesitancy that can be used across a variety of settings. Not every study w ill have access to objective data on vaccines to determine undervaccination. Therefore, different measures of hesitancy need to be compared to see how they differ in detecting hesitancy in the CABI V. There are limitations to this investigation. First, th e sample sizes in each cohort were small. In the Primary cohort, the PAC V validated screener for hesitancy was not administered, thus limiting the comparisons. In addition, this analysis used pairwise comparisons The purpose of the current evaluation was li mited to assessing each hesitancy measure and its association to survey items. In future survey validation, the survey constructs developed for the CABI V should be analyzed with different measures of hesitancy.

PAGE 140

130 The term vaccine hesitancy is relatively new. There is the potential that each measure of hesitancy is different from the other and therefore measures different aspects of hesitancy. It is important to continue to develop measures of hesitancy and simultaneously evaluate their use in research and practice settings. Such evaluation provides information on use and guides use of the most brief yet most accurate measure. In the final chapter, the conclusions of the investigation are presented.

PAGE 141

131 CHAPTER V I CONCLUSIONS The Role of Vaccine Pol icy in Vaccine Decision Making Vaccine policy has sparked heated discussions in the last few decades, some of which have been divis ive. Vaccine policymakers engage in an enduring balancing act to meet the needs of their stakeholders, understand all aspects of the vaccine debate, and influence vaccine rates thro ugh policymaking to protect the publics health While policymakers benefit from data about parents concerns ab out vaccines in order to develop policy design researchers are working in tandem to de sign interventions to reduce parents concerns about vaccines and increase vaccine rates in communities. It is advantageous for both policymakers and practitioners to know what parents concerns are about vaccines and how these may change over time. Key Findings of the Dissertation The purpose of this dissertation was to develop and evaluate a theory driven survey instrument to measure parents concerns, attitudes, beliefs and intentions about childhood vaccines. Although there are numerous single study su rveys used for research on attitudes and beliefs about vaccines there are few measures desig ned to be used over time and across multiple interventions. Developing the survey took time; the expert panel and cognitive interviews with pregnant women and parents guided the refinement and further elimination of survey questions.

PAGE 142

132 Piloting the survey revealed the need to significantly revise two constructs with new items and administer to a new cohort of pregnant women and parents of children under twelve months of age. This new baseline data had adequate sample size to conduct factor analyses and internal consistency. Through these methods, a finalized survey was ready for use in intervention studies. This study found that three of the four survey constructs wer e associated with the objective outcome of hesitancy. The construct that performed poorly did so in both univariate and multivariate analyses. This construct, Perceived Control of Vaccinating Decision, is likely a core aspect of parents attitudes and beliefs about vaccines regardless of vaccine decision to vaccinate or not vaccinate. Further survey item development to measure this latent construct is needed. The longitudinal data across four time points presented interesting results. Parents concerns ab out vaccines decrease after the birth of their child. However, there were three areas that did not decrease significantly: concerns about vaccines causing autism, concerns about ingredients in vaccines, and concerns about side effects of vaccines. These th ree areas of concern could be the focus of interventions aimed to reduce parental concerns about vaccines. Measures of hesitancy performed similarly in distinguishing hesitant parents from nonhesitant parents on survey items. Lessons Learned from Survey D evelopment By conducting a focused, indepth study that developed and then implemented a survey instrument, several lessons learned occurred. First, systematic organization of the survey items was essential to deciding which ones

PAGE 143

133 stayed in the eligibility pool and which ones were removed. Second, there is both an art and science in d esigning surveys. T he look and feel, design, ease of use of the survey instrument is important. The instrument should not fatigue, confuse or frustrate the respondent. It shoul d be appealing, well organized and not cluttered. W ellwritten questions that accurately measure attitudes and beliefs are needed. Writing good survey questions guards against bias and ambiguity. The over all survey should answer relevant questions important to policymakers, practitioners, parents, and researchers. Survey development is a collaborative and iterative process. The researcher should be the decision maker over and above the statistical output. The researcher needs to determine what survey item s stay in the construct and what items to take out. The statistical m ethods guide the researcher on these decisions, but should not depend on them for decisions. Otherwise, important conceptual information may be lost through removal of survey items. The u se of theoretical frameworks and models also guide the researcher in decision making about survey items by providing context and meaning as to what the items should be measuring By using an organized and sequential process to develop the survey, it was c lear what required revision and reevaluation. Pilot testing the instrument was essential to understanding what survey items did and did not work. The outcome variable of average days undervaccinated is very accurate, however, at the cost of data loss. Du e to the need for continuous enrollment to determine all vaccine doses across time, those with gaps of enr ollment more than thirty one days are removed from the cohort. Finally, the survey instrument was

PAGE 144

134 developed and piloted in one Denver Metro integrated health delivery system. Although this allowed for rigorous evaluation of vaccine beliefs and vaccine behaviors, the CABI V now needs to be used in other clinical settings. Currently, the CABI V is being used in two intervention studies. The first is a randomized controlled trial (RCT) to reduce parental concerns and increase vaccine behavior through use of a website. The survey is administered to parents online at baseline, three to five months of age, and then at twelve months of age of the child. The sec ond intervention study uses the CABI V instrument to tailor content, based on parents responses to survey items. Below, some the challenges involved with the development and evaluation of the CABI V are described The CABI V survey has potential to be us ed in several ways. First, it can be used as a stand alone survey administered cross sectionally or longitudinally to describe the population. Second, the CABI V can be used in interventions that aim to reduce vaccine hesitancy in different populations and age groups. These interventions would span across geographical populations and provide comprehensive information about vaccine hesitancy Third, the survey instrument can be tested in combination with other survey instruments or limit the use to one or two of the survey constructs to supplement other surveys. Conclusions and Next Steps The survey instrument was associated with the objective outcome of undervaccination. It is now ready to be used on diverse target populations across a variety of geograph ical locations.

PAGE 145

135 REFERENCES American Academy of Pediatrics, Committee on Standards of Child Health Care. (1977). AAP National Immunization Policy. AAP News and Commentary, 28:78. Atwell, J. E., Van Otterloo, J., Zipprich, J., Winter, K., Harriman, K., Salmon, D. A., ... & Omer, S. B. (2013). Nonmedical vaccine exemptions and pertussis in California, 2010. Pediatrics, 132(4), 624630. Baggs, J., Gee, J., Lewis, E., Fowler, G., Benson, P., Lieu, T., ... & Weintraub, E. (2011). The Vaccine Safety Datalink: a model for monitoring immunization safety. Pediatrics, 127 (Supplement 1), S45 S53. B arnes, John (1954). Class and c o mmittees in a Norwegian Island p arish. Human Relations, (7), 3958. Barnes, J. A. (1969). Graph theory and social networks: A technical comment on connectedness and connectivity. Sociology, 3 (2), 215232. Barrows, M. A., Coddington, J. A., Richards, E. A., & Aa ltonen, P. M. (2015). Parental vaccine hesitancy: Clinical implications for pediatric p roviders. Journal of Pediatric Health Care, 29 (4), 385394. Bell, C. A., Simmonds, K. A., & MacDonald, S. E. (2015). Exploring the heterogeneity among partially vaccinated children in a populationbased cohort. Vaccine, 33(36), 45724578. Benin, A.L., Wisler Scher, D.J., Colson, E., Shapir o E.D. & Holmboe, E.S. (2 006). Qualitative analysis of mothers decisionmaking abo ut vaccines for infants: the importance of trust. Pediatrics 117 (5), 15321541. Bennett, G.G. & Glasgow, R.E. (2009). The delivery of public health interventions via the internet: Actualizing their potential. Annual Review of Public Health, 30 273292.

PAGE 146

136 Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological bulletin, 107 (2), 238. Blair, J., Czaja, R. F., & Blair, E. A. (2013). Designing sur veys: A guide to decisions and procedures Sage Publications. Blank, N. R., Caplan, A. L., & Constable, C. (2013). Exempting schoolchildren from immunizations: states with few barriers had highest rates of nonmedical exemptions. Health Affairs, 32(7), 12821290. Brewer, N.T., Weinstein, N.D., Cuite, C.L., & Herrington, J. (2004). Risk perceptions and their relation to risk behavior. Annals of Behavioral Medicine, 27 125 130. Brewer, N. T. Chapman, G. B. Gibbons, F. X. Gerrard, M. McCaul, K. D. & Weinstein, N. D. (2007). Meta Analysis of the relationship between risk perception and health behavior: The example of vaccination. Health Psychology, 26 (2), 136145. Brown, K. F., Long, S. J., Ramsay, M., Hudson, M. J., Green, J., Vincent, C. A., ... & Sevd alis, N. (2012). UK parents decisionmaking about measles mumps rubella (MMR) vaccine 10 years after the MMR autism controversy: A qualitative analysis. Vaccine, 30 (10), 1855 1864. Brunson, E. K. (2013). The impact of social networks on parents vaccination decisions. Pediatrics, 131 (5), e1397 e1404. Centers for Disease Control and Prevention (CDC). (1999). Ten great public health achievements, 1990 1999: Impact of vaccines universally recommended for children. MMWR Morbidity and Mortality Weekly Report 48, 241243. Centers for Disease Control and Prevention (CDC ) (2011). Ten great public health achievements --United States, 2001 2010. MMWR. Morbidity and Mortality Weekly Report 60(19), 619.

PAGE 147

137 Champion, V.L., Skinner, C.S. (2008). The Health Belief Model, in Glanz, K., Rimer, B.K., Viswanath, K. Health Behavior and Health Education: Theory, Research, and Practice San Francisco, CA: Jossey Bass, 45 62. Chen, R.T. & Hibbs, B. (1998). Vaccine safety: current and future challenges. Pediatric Annals, 27 445455. Chew, L. D., Griffin, J. M., Partin, M. R., Noorbaloochi, S., Grill, J. P., Snyder, A., ... & VanRyn, M. (2008). Validation of screening questions for limited health literacy in a large VA outpatient population. Journal of General Internal M edicine, 23 (5), 561566. Colgrove, J. (2006). State of Immunity: The p olitics of vaccination in t wentieth c entury America Berkeley, CA: University of California Press. Comrey, A. L. (1988). Factoranalytic methods of scale development in personality and clinical psychology. Journal of Consulting and Clinical P sychology, 56( 5), 754. Cook, R. D., & Weisberg, S. (1982). Criticism and influence analys is in regression. Sociological Methodology 313361. Cooper, L. Z., Larson, H. J., & Katz, S. L. (2008 ). Protecting public trust in immunization. Pediatrics, 122 (1), 149153. Conner, M. & Norman, P. (2005). Predicting health b ehavior 2nd edition. New York, NY: Open University Press. Cronbach, L.J. (1951). Coefficient alpha and the internal structure o f tests. Psychometrika 16(3), 297334 D iekema, D. S. (2014). Personal belief exemptions from school vaccination r equirements. Annual Review of Public Health, 35 (1), 275292.

PAGE 148

138 D empsey, A.F., Zimet, G.D., Davis, R.L. & Koutsky, L. (2006). Factors associ ated with parental acceptance of human papillomavirus vaccines: a randomized intervention study of written information about HPV. Pediatrics 17 1486 93. Dempsey, A.F., Schaffer, S., Singer, D., Butchart, A., Davis, M. & Freed, G.L. (2011). Alternative vaccination schedule preferences among parents of young children. Pediatrics 128, 848856. DeStefano F. (2007). Vaccines and autism: evidence does not support a causal association. Clinical Pharmacology and Therapy 82, 756759. DeVellis, R. F. (1991). Scale development: Theory and a pplication Applied Social Research Methods Series, 26, Newbury Park: Sage. Dietz, T., Ostrom, E., & Stern, P. C. (2003). The struggle to govern the commons. Science 302(5652), 19071912. Downs, J.S., Bruine de Bruin, W. & Fischhoff, B. (2008). Parents vaccination comprehension and decisions. Vaccine, 26 15951607. Durbach N. (2000). They might as well brand us: working class resistance to compulsory vaccination in Victorian England. Social History of Medicine, 13, 4 5 62. Elam Evans, L. D., Yankey, D., Singleton, J. A., & Kolasa, M. (2014). National, State, and selected local area vaccination coverage mong children aged 19 35 m onths United States, 2013. MMWR. Morbidity and Mortality Weekly Report 63(34), 741748. Evans, S. J. (1991). Good surveys guide. British Medical Journal, 302(6772), 302303. Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological M ethods 4(3), 272.

PAGE 149

139 Fenner, F., Henderson, D. A., Arita, I., Jezek, Z., & Ladnyi, I. D. (1988). Smallpox and its eradication. In Smallpox and its eradication World Health Organization. Feudtner, C. & Marcus, E. (2000). Ethics and immunization policy: Promoting dialogue to sustain consensus. Pediatrics, 107(5), 11581164. Fine, P. (1993). Herd immunity: history, theory, practice. Epidemiological Review, 15 (2), 265 302. Fowler Jr, F. J. (2013). Survey research methods Thousand Oaks, CA: Sage publi cations. Fox, S., & Duggan, M. (2013). The social life of health information 2012. Washington, DC. : Pew Research Centers Internet and A merican Life Project Accessed on 4 26 2013 at http://pewinternet.org/Reports/2013/Health online.aspx Freed, G.L., Clark, S.J., Butchart, A.T., Singer, D.C. & Davis, M.M. (2010). Parental vaccine safety concerns in 2009. Pediatrics 125 654659. Freed, G. L., Clark, S. J., Butchart, A. T., Singer, D. C., & Davis, M. M. (2011). Sources and perceived credibility of vaccine safety information for parents. Pediatrics, 127 (Supplement 1), S107 S112. Gellin, B.G., Mailbach, E.W., & Marcuse, E.K. (2000). Do parents understand immunizations? A national telephone survey. Pediatrics, 106(5), 10971102. Glanz, J.M., McClure, D.L., Magid, D.L., Daley, M.F., France, E.K., Salmon, D.A. & Hambidge, S.J. (2009). Parental refusal of pertussis vaccination is associated with an increased risk of pertussis infection in children. Pediatrics 123(6), 14461451. Glanz, J.M., McClure, D.L., Magid, D.L., Daley, M.F., France, E.K. & Hambidge, S.J. (2010). Parental refusal of varicella vaccination is associated with an increased risk of varicella infection in children. Archives of Pediatric and Adolescent Medicine, 164 (1), 6670.

PAGE 150

140 aGlanz, J .M., Newcomer, S.R., Narwaney, K.J., Hambidge, S.J., Daley, M.F., Wagner, N.M., ... & Weintraub, E. S. (2013). A population based cohort study of undervaccination in 8 managed care organizations across the United States. Journal of the American Medical As sociation P ediatrics 167(3), 274281. bGlanz, J. M., Wagner, N. M., Narwaney, K. J., Shoup, J. A., McClure, D. L., McCormick, E. V., & Daley, M. F. (2013). A mixed methods study of parental vaccine decision making and parentp rovi der t rust. Academic P ed iatrics 13(5), 481488. Glanz, K., Rimer, B.K., & Viswanath, K. (2008). Health b ehavior and health e ducation: Theory, research, and p ractice San Francisco, CA: Jossey Bass. Gonzalez, E.R. (1982). TV report on DTP galvanizes US pediatricians. Journal of the American Medical Association, 248 1222. Griffin, J.P (ed). (2009). The t extbook of pharmaceutical m edicine, 6th edition. Wiley Blackwell Publishing Ltd: Hoboken, NJ. Gust, D.A., Kennedy, A, Shui, I., Smith, P.J., Nowak, G., & Pickering, L.K. (2005). Parent attitudes toward immunizations and healthcare providers: the role of information. American Journal of Preventive M edicine, 29 (2), 105112. Gust, D.A., Darling, N., Kennedy, A. & Schwartz, B. (2008). Parents with doubts about vaccines: which vaccines and why. Pediatrics 122 718725. Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., & Tatham, R. L. (2006). Multivariate data analysis (Vol. 6). Upper Saddle River, NJ: Pearson Prentice Hall. Halsey, N.A. & Salmon, D.A. (2015). Mea sles at Disneyland, a problem for all ages. Annals of Internal Medicine, 162 655656.

PAGE 151

141 Hambidge, S. J., Newcomer, S. R., Narwaney, K. J., Glanz, J. M., Daley, M. F., Xu, S., ... & DeStefano, F. (2014). Timely versus delayed early childhood vaccination and seizures. Pediatrics 133 (6), e1492e1499. Hardin, G. (1968). The tragedy of the commons. Science, 162 (3859), 12431248. Hardin, G. (1998). Extensions of the tragedy of the commons. Science, 280 (5364), 682683. Heaney, C.A., & Israel, B.A. (2008). Social n etworks and social s upports, in Glanz, K., Rimer, B.K., Viswanath, K. (2008). Health Behavior and Health Education: Theory, Research, and Practice. San Francisco, CA: Jossey Bass, 189 210. Hodge J & Gostin L. (2002). School vaccination requ irements : historical social, and legal perspectives Kentucky Law Journal, 90 831890. Hosmer Jr, D. W., Lemeshow, S., & Sturdivant, R. X. (2013). Appli ed logistic regression, 3rd edition Hoboken, NJ: John Wiley & Sons. Jackson, C., Cheater, F.M., Peacock, R., Leask, J., & Trevena, L. (2010). Evaluating a web based MMR decision aid to support informed decisionmaking by UK parents: A beforeand after feasibility study. Health Education Journal, 69 (1), 7483. Jackson, C., Cheater, F.M., Harrison, W., Peacock, Bekker, H., West, R, & Leese, B. (2011). Randomised cluster trial to support informed parental decisionmaking for the MMR vaccine, BMC Public Health, 11 (1), 475486. Kaufman, J., Synnot, A., Ryan, R., Hill, S., Horey, D., Willis, N., ... & Robinson, P. (2013). Face to face interventions for informing or educating parents about early childhood vaccination. Cochrane Database Syst ematic Rev iew CD010038. doi:10.1002/14651858.CD010038.pub2. Kennedy, A.M., Brown, C.J. & Gust, D.A. (2005). Vacci ne beliefs of parents who oppose compulsory vaccination. Public Health Report 120 252258.

PAGE 152

142 Kennedy, A, LaVail, K., Nowak, G, Basket, M. & Landry, S. (2011). Confidence about vaccines in the United States: Understanding parents perceptions Health Affairs, 30(6), 1151 1159. Kingdon, J. W., & Thurber, J. A. (1984). Agendas, alternatives, and public policies (Vol. 45). Boston: Little, Brown. Koppenjan, J. F. M., & Klijn, E. H. (2004). Managing uncertainties in networks: a network approach to problem solving and decision making Psychology Press. Lantos, J. D., Jackson, M. A., Opel, D. J., Marcuse, E. K., Myers, A. L., & Connelly, B. L. (2010). Controversies in vaccine mandates. Current Problems in Pediatric and Adolescent Health C are 40(3), 3858. Larson, H. J., Cooper, L. Z., Eskola, J., Katz, S. L., & Ratzan, S. (2011). Addressing the vaccine confidence gap. The Lancet, 378 (9790), 526535. Leech, N.L., Barrett, K.C, & Morgan, G.A. (2008). SPSS for intermediate statistics: Use and interpretatio n, 5th edition Routledge: NY, New York. Litwin, M. S. (1995). How to measure survey reliability and validity (Vol. 7). Thousand Oaks, CA: Sage Publications. Longest, B. B. (2002). Health policymaking in the United States Chicago, IL: Health Administr ation Press. Luman, E. T., McCauley, M. M., Shefer, A., & Chu, S. Y. (2003). Maternal characteristics associated with vaccination of young children. Pediatrics, 111 (Supplement 1), 12151218. MacDonald, N. E., Smith, J., & Appleton, M. (2012). Risk per ception, risk management and safety assessment: what can governments do to increase public confidence in their vaccine system?. Biologicals, 40 (5), 384388.

PAGE 153

143 MacDonald, N. E. (2015). Vaccine hesitancy: Definition, scope and determinants. Vaccine, 33 (34), 41614164. Maglione, M. A., Das, L., Raaen, L., Smith, A., Chari, R., Newberry, S., ... & Gidengil, C. (2014). Safety of vaccines used for routine immunization of US children: a systematic review. Pediatrics 134 (2), 325337. Malone, K.M. & Hi nman, A.R (2003). Vaccination mandates: The public health imperative and individual r ights. In: Goodman RA, Rothstein MA, Hoffman RE, et al., eds. Law in Public Health Practice Ne w York: Oxford University Press, 262284. March, J. G. (1994). Primer on decision making: How decisions happen Simon and Schuster. Marmot, M., & Wilkinson, R. (Eds.). (2005). Social determinants of h ealth Oxford: Oxford University Press. McCauley, M. M., Kennedy, A., Basket, M., & Sheedy, K. (2012). Exploring the choice to refuse or delay vaccines: a national survey of parents of 6 through 23month olds. Academic Pediatrics 12(5), 375383. McNaughton, C. D., Cavanaugh, K. L., Kripalani, S., Rothman, R. L., & Wallston, K. A. (2015). Validation of a short, 3 item v ersion of the Su bjective Numeracy Scale. Medical Decision Making 35(8), 932936. McNeil, M. M., Gee, J., Weintraub, E. S., Belongia, E. A., Lee, G. M., Glanz, J. M., ... & DeStefano, F. (2014). The Vaccine Safety Datalink: successes and challenges monitoring vaccine s afety. Vaccine, 32 (42), 53905398. Mergler, M. J., Omer, S. B., Pan, W. K., Navar Boggan, A. M., Orenstein, W., Marcuse, E. K., ... & Salmon, D. A. (2013). Association of vaccinerelated attitudes and beliefs between parents and health care providers. Va ccine 31(41), 45914595.

PAGE 154

144 Miller, D. (1992). Deliberative democracy and social choice. Political studies 40(s1), 5467. Miller, M.A., Viboud, C., Balinska, M. & Simonsen, L. (2009). The signature features of influenza pandemics implications for polic y. The New England Journal of Medicine 360:25952598. Montano, D.E., & Kasprzyk, D. (2008). Theory of reasoned action, theory of planned behaviors, and the integrated b ehavioral m odel, in Glanz, K., Rimer, B.K., Viswanath, K. (2008). Health Behavior an d Health Education: Theory, Research, and Practice. San Francisco, CA: Jossey Bass, 6796. Moser, B. K., & Stevens, G. R. (1992). Homogeneity of variance in the two sample means test. The American Statistician 46(1), 19 21. National Vaccine Advisory C ommittee. (1999). Strategies to sustain success in Childhood Immunizations. Journal of t he American Medical Association, 282, 363370. National Vaccine Advisory Committee. (2013). Protecting the Public's Health: Critical f unctions of the Section 3 17 Immu nization Program A r eport of the National Vaccine Advisory Committee. Public Health R eports (Washington, DC: 1974) 128(2), 78. Nunnally, J. C., Bernstein, I. H., & Berge, J. M. T. (1967). Psychometric theory New York: McGrawHill. Nunnaly, J. (1978). Psychometric theory New York: McGrawHill. Nyhan, B., Reifler, J., Richey, S., & Freed, G. L. (2014). Effective messages in vaccine promotion: A Randomized Trial. Pediatrics 133(4), e835e842. Offit, P. A., & Moser, C. A. (2009). The problem with D r Bob's alternative vaccine schedule. Pediatrics 123 (1), e164e169.

PAGE 155

145 Offit, P.A. (2011). Deadly c hoices: H ow the Anti Vaccine movement threatens us a ll. Philadelphia, PA: Basic Books. Omer, S.B., Pan, K.Y., Halsey, N.A., Stokley, S., Moulton, L.H., Nav ar, A.M., Pierce, M. & Salmon, D.A. (2006). Nonmedical exemptions to school immunization requirements; Secular trends and association of state policies with pertussis incidence. Journal of the American Medical Association 296 (14), 17571763. Omer, S.B., Enger, K.S., Moulton, L.H., Halsey, N.A., Stokley, S. & Salmon, D.A. (2008). Geographic clustering of nonmedical exemptions to school immunization requirements and associations with geographic clustering of pertussis. American Journal of Epidemiology 16 8 (12), 13891396. Omer, S. B., Richards, J. L., Ward, M., & Bednarczyk, R. A. (2012). Vaccination policies and rates of exemption from immunization, 2005 2011. New England Journal of Medicine, 367 (12), 11701171. aOpel, D.J., MangioneSmith, R., Taylor J.A., Korfiatis, C., Wiese, C., Catz, S., & Martin, D.P. (2011). Development of a survey to identify vaccinehesitant parents: The parent attitudes about childhood vaccines survey. Human Vaccines, 7 (4), 419425. bOpel, D. J., Taylor, J. A., Mangione S mith, R., Solomon, C., Zhao, C., Catz, S., & Martin, D. (2011). Validity and reliability of a survey to identify vaccinehesitant parents. Vaccine 29(38), 65986605. aOpel, D. J., Heritage, J., Taylor, J. A., Mangione Smith, R., Salas, H. S., DeVere, V. ... & Robinson, J. D. (2013). The architecture of provider parent vaccine discussions at health supervision visits. Pediatrics, 132 (6), 10371046. bOpel, D. J., Taylor, J. A., Zhou, C., Catz, S., Myaing, M., & Mangione Smith, R. (2013). The relationship between parent attitudes about childhood vaccines survey scores and future child immunization status: a validation study. JAMA Pediatrics, 167 (11), 10651071. Orenstein, W.A. & Hinman, A.R. (1999). The immunization system in the United States the role of school immunization laws. Vaccine, 17(3), S19S24.

PAGE 156

146 Orenstein, W.A., Douglas, R.G., Rodewald, L.E., & Hinman, A.R. (2005). Immunizations in the United States: success, structure, and stress. Health Affairs, 24(3), 599610. Osborn, C.Y., Cavanaugh, K., Wallston, K.A., White, R.O., & Rothman, R.L. (2009). Diabetes numeracy: An overlooked factor in understanding racial disparities in glycemic control. Diabetes Care, 32 (9), 16141619. Patton, M. Q. (2005). Qualitative research John Wiley & Sons, Ltd. Pickering, L.K. & Orenstein, W.A. (2002). Development of pediatric vaccine recommendations and policies. Seminars in Pediatric Infectious Disease 13(3), 148154. Plotkin, S.L & Plotkin S.A. (2004). A short history of vaccination. In: Plotkin SA, Or enstein WA, eds. Vaccines. 4th ed. Philadelphia, PA: Saunders, 1 15. Pluye, P., Potvin, L., & Denis, J. L. (2004). Making public health programs last: Conceptualizing sustainability. Evaluation and Program Planning, 27 (2):121133. Price, C.S., Thompson W.W., Goodson, B., Weintraub, E.S., Croen, L.A., Hinrichsen,V.L., Marcy, M., Robertson, A., Eriksen, E., Lewis, E., Bernal, P., Shay, D., Davis, R.L., & DeStefano, F. (2010). Prenatal and infant exposure to thimerosal from vaccines and immunoglobulins and risk of autism. Pediatrics 126 (4), 656664. Prislin, R., Dyer, J. A., Blakely, C. H., & Johnson, C. D. (1998). Immunization status and sociodemographic characteristics: the mediating role of beliefs, attitudes, and perceived control. American Journal of Public Health, 88 (12), 18211826. Robison SG, Groom H, & Young C. (2012). Frequency of alternative immunization schedule use in a metropolitan area. Pediatrics, 130, 3238.

PAGE 157

147 Rochefort, D. A., & Cobb, R. W. (1994). The politics of problem definition: Shaping the policy agenda Univ ersity Pr inting of Kansas. Rosenstock, I. M., Derryberry, M., & Carriger, B. K. (1959). Why people fail to seek poliomyelitis vaccination. Public Health Reports, 74 (2), 98. Rota, J.S., Salmon, D.A., Rodewald, L.E., Chen, R.T., Hibbs, B.F., Gangarosa, E.J. (2001). Processes for obtaining nonmedical exemptions to state immunization laws. American Journal Public Health, 91:645648. Roush, S.W. & Murphy, T.V. (2007). Historical comparisons of morbidity and mortality for vacc inepreventable diseases in the United States. Journal of t he American Medical Association, 298 (18), 21552163. Rousseau, J. J. (1920). The social contract: & discourses (No. 660). JM Dent & Sons, Limited Sadaf, A., Richards, J. L., Glanz, J., Salmon D. A., & Omer, S. B. (2013). A systematic review of interventions for reducing parental vaccine refusal and vaccine hesitancy. Vaccine 31(40), 42934304. Salmon, D.A, Teret, S., Raina Madnytre, C., Salisbury, D., Burgess, M. & Halsey, N. (2006). Compulsory vaccination and conscientious or philosophical exemptions: past, present, and future. Lance t 367 436442. Salmon, D.A., Moulton, L.H., Omer, S.B., DeHart, M.P., Stokley, S. & Halsey, N.A. (2005). Factors associated with refusal of childhood vaccines among parents of schoolaged children. Archives of Pediatric Adolescence Medicine 159 470476. Sandstrom, A. (2015). Nearly all states allow religious exemptions for vaccinations. Pew Research Center, http://www.pewresearch.org/fact tank/2015/07/16/nearlyallstates allow religious exemptions for vaccinations/ Accessed 9 22 2015.

PAGE 158

148 Seither, R., Calhoun, K., Knighton, C. L., Mellerson, J., Meador, S., Tippins, A., ... & Dietz, V. (2015). Vaccination coverage among children in k indergarten United States, 2014 15 school y ear. MMWR Morbidity and M orta lity Weekly Report 64(33), 897 904. Serpell, L., & Green, J. (2006). Parental decisionmaking in childhood vaccination. Vaccine, 24 (19), 40414046. Sheeran, P., Trafimow, D., Finlay, K.A., & Norman, P. (2002). Evidence that the type of person affects the strength of the perceived behavioral control intention relationship. British Journal of Social Psychology, 41 (2), 253270. Shi, L., & Singh, D. A. (2011). The Nation's h ealth, eighth ed Sudbury, MA: Jones & Bartlett Publishers. Shonkoff, J. P., Boyce, W. T., & McEwen, B. S. (2009). Neuroscience, molecular biology, and the childhood roots of health disparities: building a new framework for health promotion and disease prevention. Journal of the American Medical Association, 301 (21), 22522259. Shoup, J. A., Wagner, N. M., Kraus, C. R., Narwaney, K. J., Goddard, K. S., & Gl anz, J M. (2015). Development of an interactive social media tool for parents with concerns about v accines. Health Education & Behavior 42(3), 302312. Simon, H. A. (1959). Theories of decisionmaking in economics and behavioral science. The American E conomic R eview, 253283. Slovic, P. editor. (2010). The feeling of risk: New perspectives on risk perception Earthscan: Washington, DC. Smith, J. C. (2010). The structure, role, and procedures of the US Advisory Committee on Immunization Practices ( ACIP). Vaccine, 28, A68 A75.

PAGE 159

149 Smith, M.J., Ellenberg, S .S, Bell, L.M., & Rubin, D.M. (2008). Media coverage of the measles mumps rubella vaccine and autism controversy and its relationship to MMR immunization rates in the United States. Pediatrics, 121, e 836e843. Smith, P. J., Chu, S. Y., & Barker, L. E. (2004). Children who have received no vaccines: who are they and where do they live?. Pediatrics, 114 (1), 187 195. Smith, P.J., Humiston, S.G., Marcuse, E.K., Zhao, Z., Dorell, C.G., Howes, C., & Hibb s, B. (2011). Parental delay or refusal of vaccine doses, childhood vaccination coverage at 24 months of age, and the health belief model. Public Health Reports, 126 (S2), 135146. Streiner, D. L., Norman, G. R., & Cairney, J. (2014). Health measurement s cales: a practical guide to their development and use. Oxford university press. Sturm, L. A., Mays, R. M., & Zimet, G. D. (2005). Parental beliefs and decision making about child and adolescent immunization: from polio to sexually transmitted infections. Journal of Developmental & Behavioral Pediatrics, 26 (6), 441452. Thompson, B., & Daniel, L. G. (1996). Factor analytic evidence for the construct validity of scores: A historical overview and some guildelines. Educational and Psychological M easurement, 56(2), 197208. Thompson, B. (2004). Exploratory and confirmatory factor analysis: Understanding concepts and applications Washington, DC: American Psychological Association. Thompson, J.W., Tyson, S., Card Higginson, P., Jacobs, J., Wheeler, J., Sim pson, P., Bost, J., Ryan, K. & Salmon, D. (2007). Impact of addition of philosophical exemptions on childhood immunization rates. American Journal of Preventive Medicine 32(3), 194201. Tinsley, H. E., & Tinsley, D. J. (1987). Uses of factor analysis in counseling psychology research. Journal of Counseling P sychology, 34 (4), 414.

PAGE 160

150 Webb, T.L., & Sherran, P. (2004). Does changing behavioral intentions engender behavioral change? A metaanalysis of the experimental evidence. Psychological Bulletin, 132, 2 49168. Wei, F., Mullooly, J. & Goodman, G., McCarty, M.C., Hanson, A.M., Crane, B. & Nordin, J.D. (2009). Identification and characteristics of vaccine refusers. BMC Pediatrics 9 (18), 1 9. Weinstein ND, Kwitel, A, McCaul KD, Magnan, R.E., Gerrard M, & Gibbons FX. (2007). Risk perceptions: assessment and relationship to influenza vaccination. Health Psychology, 26 (2), 146151. Willis, G. B. (2005). Cognitive interviewing: A tool for improving questionnaire design Thousand Oaks, CA: Sage. Witteman, H. O. (2015). Addressing vaccine h esitancy with v alues. Pediatrics, 136(2), 215217. Wood, D.L. (2003). Increasing immunization coverage. Pediatrics 112 (4), 978981. World Health Organization (WHO). Report of the SAGE working group on vaccine hesi tancy. November 12, 2014. Accessed on 5292015 at http://www.who.int/immunization/programmes_systems/vaccine_hesitanc y/en/ Wroe, A.L., Turner, N., & Salkovskis, P.M. (2004). Understanding and predicting parental decisions about early childhood immunizations. Health Psychology, 23(1), 3341. Wroe, A.L. Turner, N., & Owens, G.R. (2005). Evaluation of a decisionmaking aid for parents regarding childhood immunizations. Health Psychology, 24 (6), 539547.

PAGE 161

151 APPENDIX A Survey Item Additions and Deletions in Development of Pilot Survey 1. There were 152 survey items identified in the literature. 2. Of these, 109 survey items were removed due to duplicates, lack of specificity to childhood vaccines, not pertaining to vaccine choice (some other barrier), or not relevant to the concept model 3. There were now forty three (43) survey items. The following thirteen (13) items were removed by a collaborative team process. This process involved discussion of the literature, context of the question, scenarios about how it would be used, and assessment of its theoretical contribution. 1. Vaccination requirements protect my child from getting disease from unvaccinated children 2. I might change my mind in the future about my vaccine decision I made for my child 3. I am currently no t planning to change my vaccine decision I made for my child 4. My child is more likely to get sick than other children 5. My child does not seem to have resistance to disease 6. How important do you think vaccines are to the health of your child 7. My child is at low risk to get a vaccine preventable infectious disease 8. Parents, not the government, should made decisions about vaccinating their child 9. The internet is an important source of information when I make decisions about vaccination for my child 10. I am concerned that my child will not be vaccinated on time because there is not enough of some vaccines 11. How safe do you think vaccines are to the health of your child? 12. I have religious concerns about vaccinating my child 13. Ive heard negative things about vaccines in the me dia 4. Now, thirty (30) survey items remained for evaluation by the Expert Panel. Of these 30 items, seven (7) were removed after the SMEs evaluation and cognitive interviews with the target population 1. Vaccines are always proven safe before they are approv ed for use 2. Diseases had already begun to disappear before vaccines were introduced, because of better hygiene and sanitation 3. So many children are vaccinated that my child is safe from infectious disease even if I do not vaccinate them 4. The benefits out weigh the risks of vaccines 5. Government vaccine requirements help protect everyone against infectious diseases 6. Vaccines are getting safer all the time as a result of medical research 7. New vaccines are recommended only if they are as safe as older vaccines 5. Th e survey now had twenty three (23) survey items. This was the pilot survey administered to the pregnant mother cohort. 6. After the survey was administered and evaluated using internal consistency methods, an additional eight (8) items were removed. 1. As a pa rent, I have given a lot of thought about vaccinations for my child

PAGE 162

152 2. Parents should have the right to refuse vaccines that are required for school for any reason So many children are vaccinated that my child is safe from infectious disease even if I do n ot vaccinate them 3. Getting vaccinated will greatly reduce the chance of getting infectious diseases like polio or measles, but will not eliminate the chance of getting these diseases completely 4. That it would be painful for my child to receive so many shots during one doctors visit 5. It is better for my child to develop immunity by getting sick than to get a shot 6. Generally I intend to do what my childs doctor recommends about vaccines for my child 7. Most of the parents I know vaccinate their child 8. I am concern ed about serious infectious diseases like whooping cough or measles 7. After additional literature review, an additional fourteen (14) survey items were added to the fifteen existing items for a total of twenty nine items. 1. In general, most of my close frien ds have similar beliefs about vaccines as me 2. In general, my family (e.g. sisters, brothers and cousins) have similar beliefs about vaccines as me 3. In general, my parents have similar beliefs about vaccines as me 4. In general, my spouse or partner has similar beliefs about vaccines as me 5. In general, my obstetrician has similar beliefs about vaccines as me 6. Information on the internet about vaccines helps me to make decisions about vaccinating my child 7. I think that there are other parents, like me, struggling with the decision about vaccines for their child 8. School laws requiring that children have up to date vaccines to enter daycare or public school influence my decisions about vaccinating my child 9. Allowing parents to delay vaccine doses or skip some vaccines lets parents be more in charge of their childrens health care 10. Parents who skip or delay certain vaccines are relying on other people in the community being vaccinated to protect their unvaccinated children from getting sick. 11. How confident are you about your knowledge about how vaccines work? 12. How confident are you about your knowledge about infectious diseases? 13. How confident are you that you can express your views about vaccines to your obstetrician? 14. The risks from getting a vaccine outweigh the risks from getting a disease 8. The revised survey of twenty nine survey items was administered to the Primary cohort of pregnant women and parents of children under twelve months of age 9. Through exploratory factor analyses, six (6) survey items were deleted. 1. Informatio n on the internet about vaccines helps me to make decisions about vaccinating my child 2. I think that there are other parents, like me, struggling with the decision about vaccines for their child 3. School laws requiring that children have up to date vaccines t o enter daycare or public school influence my decisions about vaccinating my child 4. Allowing parents to delay vaccine doses or skip some vaccines lets parents be more in charge of their childrens health care 5. Parents who skip or delay certain vaccines are r elying on other people in the community being vaccinated to protect their unvaccinated children from getting sick. 6. The risks from getting a vaccine outweigh the risks from getting a disease 10. The finalized survey had twenty three it ems and is located in Ap pendix C

PAGE 163

153 APPENDIX B Justification of the Dichotomous Split of the Variable Covariate Bivariate description Justification Survey constructs (Beliefs about Vaccinating, Evaluation of VPD/VAE, Perceived Control about Vaccination Decision, Subjective N orms about Vaccinating Converted from responses scale of 1 5 to bi nary where 1 3.49 = hesitant response and 3.505 = not hesitant response The literature supports that parents who fully vaccinate have concerns (Freed, 2010, 2011; Kennedy, 2011). Using vari ous testing scenarios, it was determined that these cut points allowed a minimal amount of uncertainty for some responses without compromising misclassification. Employment status Converted from 6 c ategorical response to binary where full time = 1 and pa rt time, student, stay at home parent, unemployed, retired = 0. The variable was split based on the assumption that full time employed parents may have less time to limit vaccines, have additional vaccine visits, or time for vaccine related illnesses (i.e. varicella). Classifying those who work part time with other categories is justified as part time employment can be seasonal or limited hours. Therefore, full time is quantified as 40 hours or more a week and is compared to the other categories. Number o f children Two or more children = 1 and less than two children = 0. In recent literature, it was reported that larger families have more undervaccinated children by parental choice (Bell, 2015; Luman, 2003) Health Literacy Responses of extremely or qui te a bit = 0; responses of somewhat, a little bit, or not at all = 1. This cut point is established by validity testing of the question by Chew, et al., 2008. Influence of doctor on vaccine decision Converted from a 7 it em response scale to a bin ary variable. Medical provider = 1 and all others (spouse, family, friend, alternative medical provider, none, other) = 0. Numerous survey studies have reported the influence of medical providers (pediatrician or family physician) on trust and vaccine deci sion making (Benin, 2006; Freed, 2010; 2011). Social Media use for health information Converted from a 5 point response to a binary variable. 1=once week or more and 0=less than weekly Pew (2013) data finds that social media use for health information is an upward trend. Also, the literature indicates that parents who do not vaccinate use the internet as a primary source of information (Freed, 2011). By setting the level of use at a weekly basis, this justifies the bivariate split. Parent controls vacci ne decision Converted this 5 item response scale to binary 1 = Parents controls vaccine decision and 0 = shared or physician controls vaccine decision. Parents who are hesitant or do not vaccinate according to the recommended schedule want control over th e decision (Prislin, 1998).

PAGE 164

154 Covariate Bivariate description Justification Race Converted from a 6 ite m response scale to a binary variable of white = 1 and all other races = 0. Although the literature is not clear or consistent, race was determined to be a meaningful variable. White compared to all o ther categories provided enough response counts in the other category. Education Converted from a 6 it em response scale to a binary variable of some college or greater = 1 and high school or less = 0. The literature suggests that more highly educated pa rents have more undervaccinated children (Gust, 2008). Age Converted from a scale variable to a binary variable of 30 years of age or greater = 1 and less than 30 years= 0 The literature suggests older parents have more undervaccinated children (Gust, 200 8). Income Converted from a 5 item scale to a binary variable. 80K

PAGE 165

155 APPENDIX C Pilot Su rvey Administered to Cohort of 120 Pregnant Women Survey item Response scale Theoretical construct Source of item Survey Items Constructs Generally I intend to do what my childs doctor recommends about vaccines for my child b Strongly disagree to stron gly agree Subjective Norms Freed, et al., 2010 Most of the parents I know vaccinate their children b Strongly disagree to strongly agree Subjective Norms New item developed As a parent, I have given a lot of thought about vaccinations for my child a ,b S trongly disagree to strongly agree Subjective Norm New item developed How confident are you that you have the necessary information to make decisions about vaccination for your child b Very confident to not confident at all Perceived Control New item deve loped How confident are you that you will be able to protect your child from some types of infectious disease by vaccinating him or her b Very confident to not confident at all Perceived Control New item developed Parents should have the right to refuse vaccines that are required for school for any reason b Strongly disagree to strongly agree Perceived Control Freed, et al., 2010 I believe many of the illnesses vaccines prevent are serious a ,b Strongly disagree to strongly agree Evaluation of VPD/VAE O pel, et al., 2011 a,b Getting vaccinated will greatly reduce the chance of getting infectious diseases like polio or measles, but will not eliminate the chance of getting these diseases completely b Strongly disagree to strongly agree Evaluation of VPD/VAE Song, et al., 2014 I believe vaccines are generally safe b Strongly disagree to strongly agree Evaluation of VPD/VAE Modified from Gellin, et al., 2000 I believe my child could get a serious disease if he or she were not vaccinated b Strongly disagree to strongly agree Evaluation of VPD/VAE Kennedy, et al., 2011 Children get more vaccines than are good for them a ,b Strongly disagree to strongly agree Evaluation of VPD/VAE Gellin, et al., 2000 My child will not need vaccines for Strongly disagree Evaluation of Adapted from

PAGE 166

156 Survey item Response scale Theoretical construct Source of item diseases that are not common anymore, like polio a ,b to strongly agree VPD/VAE Freed, et al., 2010 I am concerned about serious infectious diseases like whooping cough or measles b Strongly disagree to strongly agree Evaluation of VPD/VAE Kennedy, et al., 2011 I am concerned that the ingredients in vaccines are unsafe a ,b Strongly disagree to strongly agree Evaluation of VPD/VAE Gust, et al., 2005 I am concerned that some vaccines cause autism in healthy children a ,b Strongly di sagree to strongly agree Evaluation of VPD/VAE Freed, et al., 2010 I am concerned that there are serious side effects of vaccines a ,b Strongly disagree to strongly agree Evaluation of VPD/VAE Freed, et al., 2010 I believe it is better for my child to de velop immunity by getting sick than to get a shot a ,b Strongly disagree to strongly agree Beliefs about vaccines Salmon, et al., 2005 I believe there has not been enough research on the safety of vaccines a ,b Strongly disagree to strongly agree Beliefs a bout vaccines Freed, et al., 2010 I believe it is better for my child to get the natural disease than to get a vaccine a ,b Strongly disagree to strongly agree Beliefs about vaccines Modified from Salmon, et al., 2005 Vaccines strengthen the immune syste m Strongly disagree to strongly agree Beliefs about vaccines Salmon, et al., 2005 Getting vaccines is a good way to protect my child from infectious diseases b Strongly disagree to strongly agree Beliefs about vaccines Freed, et al., 2010 I am concern ed that my childs immune system could be weakened by too many vaccines a ,b Strongly disagree to strongly agree Beliefs about vaccines Gellin, et al., 2000 I am concerned that it would be painful for my child to receive so many shots during one doctors v isit a ,b Strongly disagree to strongly agree Beliefs about vaccines Gust, et al., 2005 Survey Items as Descriptive Variables As an expecting parent, I intend to: 5 choices that are reduced to intend to vaccinate or intend not to vaccinate, including undec ided c Intentions about vaccinating New item developed

PAGE 167

157 Survey item Response scale Theoretical construct Source of item Which person had the most influence on your decision about vaccination for the child you are pregnant with? 7 choices d Functions of social networks New item developed Do you personally know anyone wh o has had a bad reaction to a vaccine? Yes; No Subjective Norms Opel, 2011 When it comes to future decisions about vaccination for the child I am expecting 5 point preference scale e Perceived control about vaccination decision New item developed How ofte n do you have someone like a family member, friend, or clinic worker help you read hospital, pharmacy, or medical materials, like instructions or information on a medical condition Always; Often; Sometimes; Occasionally; Never Health Literacy Chew, 2008 H ow confident are you filling out forms by yourself Extremely; Quite a bit; Somewhat; A little bit; Not at all Health Literacy Chew, 2008 How often do you h ave problems learning about your medical condition because of difficulty understanding written infor mation Always; Often; Sometimes; Occasionally; Never Health Literacy Chew, 2008 How good are you with fractions 1: not at all good to 6: Extremely good Health Numeracy McNaughton, 2015 How good are you at figuring out how much a shirt will cost if it is 25% off 1: not at all good to 6: Extremely good Health Numeracy McNaughton, 2015 How often do you gind numerical information to be useful 1: not at all good to 6: Extremely good Health Numeracy McNaughton, 2015 In the past 6 months, how often have you us ed the Internet, in general Not at all; Less than once a month; About once a month; Every week; Every day Informational support New item developed I n the past 6 months, how often have you used the Internet to look for health information Not at all; Less t han once a month; About once a month; Every week; Every day Informational support New item developed In the past 6 months, how often have you used the Internet to help get information about vaccines for your child Not at all; Less than once a month; About once a month; Every week; Every day Informational support New item developed If I had access to vaccine information, blogs and discussion boards to chat with other parents on a Kaiser Permanente vaccine information website, I think my level of concern ab out vaccines would Increase; Stay the same; Decrease Informational support New item developed In the past 6 months, how often have you Not at all; Less Informational New item

PAGE 168

158 Survey item Response scale Theoretical construct Source of item used social media (such as blogging online, Facebook, online discussion boards), in general than once a month; About once a month; Every week; Every day support developed In the past 6 months, how often have you used social media (such as blogging online, Facebook, online discussion boards) to look for and talk about health informati on Not at all; Less than once a month; About once a month; Every week; Every day Informational support New item developed In the past 6 months, how often have you used social media (such as blogging online, Facebook, online discussion boards), to look for an talk about vaccines for your child Not at all; Less than once a month; About once a month; Every week; Every day Informational support New item developed How many children do you have 1 child; 2 children; 3 or more children; None, I am pregnant with m y first child Demographic information N/A In general, what kind of schooling do you anticipate your child who you are pregnant with will participate in Public school; Private school; Charter school; Home school; Dont know yet Demographic information N/A What is the highest level of school that you completed Grade school; Less than high school; High school; some college; College; Graduate school Demographic information N/A Write your age in the box below Write in Demographic information N/A What is your gender Male; Female Demographic information N/A Are you pregnant Yes; No Demographic information N/A What is your current marital status Married; Separated; Divorced; Not married; Single; Living with a partner; Widowed Demographic information N/A Pleas e indicate which of the following best describes you White; Black/African American; Asian; American Indian/Alaskan Native; Native Hawaiian or Other Pacific Islander; Demographic information N/A

PAGE 169

159 Survey item Response scale Theoretical construct Source of item Other Do you consider yourself to be Latino, Spanish or Hispa nic Yes, No, Dont know Demographic information N/A Please indicate your current employment status Employed full time; Employed part time; Unemployed; Stay at home parent; Retired; Student Demographic information N/A What is the total yearly income for y our household Less than 40,000 dollars; 40,000 to 80,000 dollars; 81,000 to 120,000 dollars; 121,000 to 150,000 dollars; Over 150,000 dollars; Prefer not to answer Demographic information N/A PARENT ATTITUDE ABOUT CHILDHOOD VACCINES (PAC V) f hinking abou t decisions you will make about vaccines for your child, will you ever delay having your child get a shot (not including seasonal flu or swine flu (H1N1) shots) for reasons other than illness or allergy Yes; No; Dont know Vaccine hesitancy Opel, 2011 Th inking about decisions you will make about vaccines for your child, will you ever decide not to have your child get a shot (not including seasonal flu or swine flu (H1N1) shots) for reasons other than illness or allergy Yes; No; Dont know Vaccine hesitanc y Opel, 2011 Do you want the baby you are expecting to get all the recommended shots Yes; No; Dont know Vaccine hesitancy Opel, 2011 How sure are you that following the recommended shot schedule will be a good idea for your child? Please answer on a scale of 0 to 10, where 0 = "not at all sure" and 10 = "completely sure". Vaccine hesitancy Opel, 2011 Children get more shots than are good for them. Strongly agree to strongly dis agree g Vaccine hesitancy Opel, 2011 I believe that many of the illness es that shots prevent are severe Strongly agree to strongly disagree g Vaccine hesitancy Opel, 2011 It is better for my child to develop immunity by getting sick than to get a shot Strongly agree to strongly disagree g Vaccine hesitancy Opel, 2011 It is better for children to get fewer vaccines at the same time Strongly agree to strongly disagree g Vaccine hesitancy Opel, 2011 I will trust the information I will receive about shots Strongly agree to strongly disagree g Vaccine hesitancy Opel, 2011 I am able to openly discuss my concerns about shots with my obstetrician or midwife Strongly agree to strongly disagree g Vaccine hesitancy Opel, 2011

PAGE 170

160 Survey item Response scale Theoretical construct Source of item How concerned are you that your child might have a serious side effect from a shot Not concerned at all to Ve ry concerned h Vaccine hesitancy Opel, 2011 How concerned are you that any one of the childhood shots might not be safe Not concerned at all to Very concerned h Vaccine hesitancy Opel, 2011 How concerned are you that a shot might not prevent the disease Not concerned at all to Very concerned h Vaccine hesitancy Opel, 2011 Overall, how hesitant about childhood shots would you consider yourself to be j Not at all hesitant to Very hesitant i Vaccine hesitancy Opel, 2011 All things considered, how much do yo u trust your obstetrician or midwife? Please answer on a scale of 0 to 10, where 0 = "do not trust at all" and 10 = "completely trust" Vaccine hesitancy Opel, 2011 aReverse coded bStron gly disagree; Disagree; Neither agree or disagree ; Agree; Strongly a gree cGet all vaccines on time as recommended by my Kaiser provider; Get some of the vaccines recommended by my Kaiser provider; Delay vaccines and wait until my child is older; Not vaccinate my child and not get any vaccines recommended by my Kaiser provi der; Undecided about my decision about vaccines d Spouse or partner; Family member; Friend; Medical provider; Alternative medicine provider; none; other e Get all vaccines on time as recommended by my child's doctor; Get some of the vaccines recommended by my child's doctor with the chance I might change my mind in the future and get all vaccines recommended; Get some of the vaccines recommended by my child's doctor and delay other vaccines until my child is older; Delay vaccines and wait until my child is older; Not vaccinate my child with the chance I might change my mind in the future and get vaccines recommended by my child's doctor; Not vaccinate my child and not get any vaccines recommended by my child's doctor; Undecided about my decision about vaccin es. fAdministered by telephone prior to baseline CABI V survey gStrongly agree; Agree, Not sure; Disagree; Strongly disagree hNot concerned at all; Not too concerned; Not sure; Somewhat concerned; Very concerned iNot at all hesitant; Not too hesitant; Not sure; Somewhat hesitant; Very hesitant

PAGE 171

161 APPENDIX D Finalized survey items organized by theoretical construct and source of survey item Survey item a.b Response scale Theoretical construct Source of item Survey Items Constructs How confident are you that you have the necessary information to make decisions about vaccination for your child Absolutely confident to Not at all confident c Perceived Control New item developed How confident are you about your knowledge about how vaccines work Absolu tely confident to Not at all confident c Perceived Control New item developed How confident are you about your knowledge about infectious diseases Absolutely confident to Not at all confident c Perceived Control New item developed How confident are you tha t you can express your views about vaccines to your obstetrician Absolutely confident to Not at all confident c Perceived Control New item developed In general, most of my close friends have similar beliefs about vaccines as me Strongly disagree to strongl y agree d Subjective Norms New item developed In general, my family (e.g. sisters, brothers and cousins) have similar beliefs about vaccines as me Strongly disagree to strongly agree d Subjective Norms New item developed In general, my parents have similar beliefs about vaccines as me Strongly disagree to strongly agree d Subjective Norms New item developed In general, my spouse or partner has similar beliefs about vaccines as me Strongly disagree to strongly agree d Subjective Norms New item developed In g eneral, my obstetrician has similar beliefs about vaccines as me Strongly disagree to strongly agree d Subjective Norms New item developed How confident are you that you will be able to protect your child from some types of infectious disease by vaccinatin g him or her Absolutely confident to Not at all confident c Beliefs about vaccines New item developed Many of the illnesses vaccines prevent are serious Strongly disagree to strongly agree d Beliefs about vaccines Opel, et al., 2011 My child will not nee d vaccines for diseases that are not common anymore, like polio a Strongly disagree to strongly agree d Beliefs about vaccines Adapted from Freed, et al., 2010 My child could get a serious disease if he or she were not vaccinated Strongly disagree to str ongly agree d Beliefs about vaccines Kennedy, et al., 2011 Vaccines strengthen the immune system Strongly disagree to strongly agree d Beliefs about vaccines Salmon, et al., 2005 Getting vaccines is a good way to protect Strongly disagree Beliefs about Freed, et al.,

PAGE 172

162 Survey item a.b Response scale Theoretical construct Source of item my child from infectious disease s to strongly agree d vaccines 2010 Vaccines are safe Strongly disagree to strongly agree d Evaluation of VPD/VAE Modified from Gellin, et al., 2000 Children get more vaccines than they need Strongly disa gree to strongly agree d Evaluation of VPD/VAE Modified from Gellin, et al., 2000 I am concerned that the ingredients in vaccines are unsafe a Strongly disagree to strongly agree d Evaluation of VPD/VAE Gust, et al., 2005 I am concerned that some vaccines cause autism in healthy children a Strongly disagree to strongly agree d Evaluation of VPD/VAE Freed, et al., 2010 I am concerned that vaccines have serious side effects Strongly disagree to strongly agree d Evaluation of VPD/VAE Freed, et al., 2010 I believe there has not been enough research on the safety of vaccines a Strongly disagree to strongly agree d Evaluation of VPD/VAE Freed, et al., 2010 I believe it is better for my child to get the natural disease than to get a vaccine a Strongly disagree to strongly agree d Evaluation of VPD/VAE Modified from Salmon, et al., 2005 I am concerned that my childs immune system could be weakened by too many vaccines a Strongly disagree to strongly agree d Evaluation of VPD/VAE Gellin, et al., 2000 Survey Ite ms as Descriptive Variables As an expecting parent, I intend to: 7 choices that are reduced to intend to vaccinate or intend not to vaccinate, including undecided e Intentions about vaccinating New item developed Which person had the most influence on yo ur decision about vaccination for the child you are pregnant with? 7 choices f Functions of social networks New item developed When it comes to future decisions about vaccination for the child I am expecting 5 point preference scale g Perceived control abou t vaccination decision New item developed How often do you have someone like a family member, friend, or clinic worker help you read hospital, pharmacy, or Always; Often; Sometimes ; Health Literacy Chew, 2008

PAGE 173

163 Survey item a.b Response scale Theoretical construct Source of item medical materials, like instructions or information on a medical condition Occasionally; Never Do you prefer to receive medical information In words; In numbers; Both words and numbers Preference of health information presentation New item developed In the past 2 months, how often have you used the Internet to look for health information Not at all; Less than once a month; About once a month; Every week; Every day Informational support New item developed In the past 2 months, how often have you used the Internet to help get information about vaccine s for your child Not at all; Less than once a month; About once a month; Every week; Every day Informational support New item developed In the past 2 months, how often have you used social media (such as blogging online, Facebook, online discussion boards ), in general Not at all; Less than once a month; About once a month; Every week; Every day Informational support New item developed In the past 2 months, how often have you used social media (such as blogging online, Facebook, online discussion boards) t o look for and talk about health information Not at all; Less than once a month; About once a month; Every week; Every day Informational support New item developed How many children do you have 1 child; 2 children; 3 or more children; None, I am pregnant with my first child Demographic information N/A What is the highest level of school that you completed Grade school; Less than high school; High school; some college; College; Graduate school Demographic information N/A Write your age in the box below Wr ite in Demographic information N/A What is your gender Male; Female Demographic information N/A What is your current marital status Married; Separated; Divorced; Not married; Single; Living with a partner; Widowed Demographic information N/A Please indi cate which of the following best describes you White; Black/African Demographic information N/A

PAGE 174

164 Survey item a.b Response scale Theoretical construct Source of item American; Asian; American Indian/Alaskan Native; Native Hawaiian or Other Pacific Islander; Other Do you consider yourself to be Latino, Spanish or Hispanic Ye s, No, Dont know Demographic information N/A Please indicate your current employment status Employed full time; Employed part time; Unemployed; Stay at home parent; Retired; Student Demographic information N/A What is the total yearly income for your ho usehold Less than 40,000 dollars; 40,000 to 80,000 dollars; 81,000 to 120,000 dollars; 121,000 to 150,000 dollars; Over 150,000 dollars; Prefer not to answer Demographic information N/A aReverse score for constructs bLeading phrase depended on whether par ent was pregnant or parent: As a parent expecting a child or As a parent. cAbsolutely confident; Very confident; Somewhat confident; Slightly confident; Not at all confident dStrongly disagree; Disagree; Not sure; Agree; Strongly agree e Get all vacci nes on time as recommended by my child's doctor; Get some of the vaccines recommended by my child's doctor with the chance I might change my mind in the future and get all vaccines recommended; Get some of the vaccines recommended by my child's doctor and delay other vaccines until my child is older; Delay vaccines and wait until my child is older; Not vaccinate my child with the chance I might change my mind in the future and get vaccines recommended by my child's doctor; Not vaccinate my child and not get any vaccines recommended by my child's doctor; Undecided about my decision about vaccines. f Spouse or partner; Family member; Friend; Medical provider; Alternative medicine provider; none; other g I prefer to make the final decision about which vaccines my child receives or does not receive; I prefer to make the final decision about which vaccines my child receives or does not receive after seriously considering my childs doctor's opinion; I prefer that my childs doctor and I share responsibility for de ciding which vaccines my child receives or does not receive; I prefer that my childs doctor make the final decision about which vaccines my child receives or does not receive, but seriously considers my opinion; I prefer to leave all decisions about which vaccines my child receives or does not receive to my childs doctor.