Maternal efforts to prevent type 1 diabetes in genetically screened infants:

MISSING IMAGE

Material Information

Title:
Maternal efforts to prevent type 1 diabetes in genetically screened infants:
Physical Description:
iv, 170 leaves : ill. ; 29 cm.
Language:
English
Creator:
Baughcum, Amy E
Publication Date:

Subjects

Subjects / Keywords:
Clinical and Health Psychology thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Clinical and Health Psychology -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2004.
Bibliography:
Includes bibliographical references.
Statement of Responsibility:
by Amy E. Baughcum.
General Note:
Printout.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 003103653
System ID:
AA00003583:00001

Full Text











MATERNAL EFFORTS TO PREVENT TYPE 1 DIABETES
IN GENETICALLY SCREENED INFANTS














By

AMY E. BAUGHCUM


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2004
































Copyright 2003

by

Amy Baughcum
































This dissertation is dedicated to my parents and grandparents who instilled in me the
value of life-long learning.














ACKNOWLEDGMENTS

First, I want to thank the families that willingly gave their time to participate in this

study. Second, I would like to thank my collaborators at the University of Florida and

Medical College of Georgia as well as the Children's Miracle Network and the American

Diabetes Association for funding this work.

On a more personal note, I am grateful to my family for their constant love,

encouragement, and support. I appreciate their sacrifices that allowed me to pursue my

education. My friends and labmates were also huge helps to me in this process by

providing feedback, caring, and welcomed distraction. I am grateful to my research

assistants (Adam Lewin, MS; and Jennifer Walsh, BS) who assisted me with this study.

Without them, this project could not have happened.

I am forever indebted to all of my mentors who helped me along the way, including

Scott Powers, PhD; and Robert Whitaker, MD, MPH, with whom I worked before

entering graduate school. They both gave me a strong foundation of skills and

knowledge that continue to serve me well as I pursue my career.

Most importantly, I appreciate the incredible opportunity I had at the University of

Florida to be mentored by two brilliant, inspiring women (Suzanne Bennett Johnson,

PhD; and Dr. Alexandra Quittner, PhD) both of whom I consider to be exemplary

academicians. Several other faculty members (including Desmond Schatz, MD; Samuel

Sears, PhD; and Fonda Eyler, PhD) have also provided me with support, mentorship, and

sound guidance, which have proven influential in my graduate education.














TABLE OF CONTENTS
Page

ACKNOWLEDGMENTS......................................................................................... iv

A B STR A C T ..................................... .............................................. vii

CHAPTER

1 IN TR O D U CTIO N ................................................................................................. 1

2 REVIEW OF THE LITERATURE.............................................................................. 5

Psychological Impact of Genetic Testing .............................................. .............. 5
Behavioral Impact of Genetic Testing ................................................. ................ 9
Theoretical Models of Genetic Screening and Behavioral Change ....................... 12
Newborn Genetic Screening .............................................................................. 17
Type 1 Diabetes: Etiology and Prevention .......................................... ............ .. 21
Prediction and Pre-Symptomatic Screening for Type 1 Diabetes............................ 25

3 RATIONALE AND PURPOSES ................................................................... 37

Objective 1: To Investigate the Extent of Reported Maternal Behavior Change as a
Result of Genetic Screening for Type 1 Diabetes............................................ 38
Objective 2: To Assess Predictors of Maternal Behavior Change as a Result of
Genetic Screening for Type 1 Diabetes....................................................... 39
Objective 3: To Assess Psychological Effects (i.e., Anxiety) of Maternal Behavior
Change O ver Tim e .......................................................... .............................. 42
Objective 4: To Compare Reported Behavior Change between Mothers of Children
Genetically at Risk for Developing Type 1 Diabetes with Mothers of Children in
the Diabetes Prevention Trial Who Were ICA+, and Therefore, at Even Greater
Risk for Diabetes Onset................................................... .............................. 43

4 METHODS AND MATERIALS ............................................................................... 45

Prospective Assessment of Newborn Diabetes Autoimmunity (PANDA) Study
Procedures ....................................................................................................... 45
Participants........................................................................................................... 48
Procedures............................................................................................................ 49
M measures .............................................................................................................. 51
Statistical A nalyses ............................................................................................... 61








5 RESU LTS .................................................................................................................. 75

Sam ple Characteristics ......................................................................................... 75
O objective 1 ........................................................................................................... 76
O objective 2 ........................................................................................................... 81
O objective 3 ................................................................................................................. 88
O objective 4 ........................................................................................................... 89

6 D ISCU SSIO N .......................................................................................................... 117

H ypotheses ......................................................................................................... 118
Strengths and Lim stations ........................................................................................ 130
Im plications and D directions for Future Research .................................................. 134

APPENDIX STRUCTURED TELEPHONE INTERVIEW...................................... 137

LIST O F REFEREN CES ................................................................................................ 154

BIO G RA PH ICA L SKETCH .................................................................................... 170

































vi















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

MATERNAL EFFORTS TO PREVENT TYPE 1 DIABETES IN GENETICALLY
SCREENED INFANTS

By

Amy E. Baughcum

August 2004

Chair: Suzanne Bennett Johnson
Cochair: Alexandra Quittner
Major Department: Clinical and Health Psychology

Currently, research programs exist to screen newborns in the general population for

genetic risk of developing Type 1 diabetes, including the Florida Prospective Assessment

of Newborn Diabetes Autoimmunity (PANDA) study. These screening programs are part

of longitudinal studies addressing the etiology of type 1 diabetes, with the ultimate goal

of developing preventative interventions. However, little is currently known about the

impact of newborn genetic screening on maternal behaviors of newborns found to be at

increased risk for the disease. Additionally, since we do not presently know how to

effectively prevent type 1 diabetes, health care professionals are not able to offer

definitive recommendations to mothers regarding specific behaviors to prevent diabetes

in their at risk children. In the absence of this information, mothers may take their own

actions in an effort to prevent the disease in their children. The purpose of this

exploratory study was to examine maternal reported behavior changes associated with

identifying at risk infants via genetic screening.








Structured telephone interviews were conducted with 192 mothers of children

between the ages of 2 and 7 years who were previously identified as at increased risk

through genetic screening. Interview questions elicited qualitative and quantitative

information regarding maternal behavior changes affecting the child's diet, physical

activity, stress level, environment, and health surveillance. Additional questions assessed

mothers' anxiety, perceived control, perceived risk, information-seeking behaviors, and

sources of information regarding their children's risk for diabetes.

Results indicated that most mothers reported engaging in behavior change (67%)

and typically these behaviors involved increased health surveillance and healthy lifestyle

changes. Significant predictors of behavior change included family history of diabetes,

anxiety, coping, perceived risk, and information seeking. Overall, these findings suggest

that genetic screening for type 1 diabetes has minimal negative impact on maternal

behavior. Despite the positive nature of subsequent behavior modifications, such

behavior changes that may occur in individuals' everyday lives in response to a health

risk could threaten the internal validity of natural history studies and prevention trials if

not carefully monitored.













CHAPTER 1
INTRODUCTION

Advances in the field of human genetics are rapidly changing the practice of

medicine. The Human Genome Project (HGP), completed in 2000, has played a key role

in this revolution by identifying and sequencing the genes that constitute the entire

human genome. Genetic mutations account for an estimated 5,000 diseases and influence

the development of thousands of others. Estimates suggest that 20 diseases account for

80% of the deaths in the Western world; and these diseases are due to the influence of

100 to 200 individual genes which will be identified in the next few years (Roberts, 2000;

Patenaude, Guttmacher, & Collins, 2002). Tests are currently available to identify

specific genetic markers that may lead to a disease in those who are at risk for developing

the disease at some point during their lives. There are two types of tests. One type is

known as "genetic testing," which involves using "specific assays to determine the

genetic status of an individual already suspected to be at high risk for a particular

inherited condition because of family history or clinical symptoms." The other, genetic

screening, involves the use of "various genetic tests to evaluate populations or groups of

individuals independent of a family history of a disorder" (Committee on Assessing

Genetic Risks, Institutes of Medicine, 1994, p. 4). However, the terms "genetic testing"

and "genetic screening" are often used interchangeably; and thus, these words are used

interchangeably in this paper as well.

As we come to better understand human genetics, we have the opportunity to learn

more about the role that specific genes play in the etiology of disease. It is estimated that








each person's genetic make-up contains 5 to 30 alterations in DNA that could predispose

the development or transmission of a genetically based disease. It is apparent that

advances in genetics will continue to expand, and screening for genetic susceptibility for

diseases will become more commonplace in the coming decades (Juengst, 1995).

Longitudinal research studies of genetically at risk individuals are necessary to learn

about the natural progression of disease in order to develop effective prevention

strategies. These studies will provide a better understanding of the interactions between

multiple genes in disease development as well as interactions between genes and the

environment. While there are diseases that are determined by a single gene mutation,

such as Huntington's disease, many conditions are genetically more complex, involving

multiple genes and environmental factors. Thus, as we learn more about genetic

predispositions, it will be increasingly important to examine environmental factors

(including individuals' health behaviors) to complete our scientific understanding of

disease etiology. The exciting new opportunities in genetics are accompanied by many

unanswered questions about how the public at large will accept and understand these new

techniques and the risk information they provide.

Unfortunately, new advances in medical technology have outpaced the rate at

which psychological research has proceeded. Genetic medicine can have a huge

influence in life and death issues, raising ethical, social and legal concerns. It is ethically

imperative to consider how the new genetic revolution impacts, both positively and

negatively, the quality of life for individuals and their families. By identifying an

individual's genetic risk for disease, there is potential for early treatment or disease

prevention, or in the case of an incurable disease, the ability to initiate health surveillance








and/or plan for the future. Current literature suggests that certain health behaviors (i.e.,

diet and exercise) can moderate risk for several diseases, such as heart disease, cancer or

type 2 diabetes. Therefore, awareness of one's genetic risk may directly affect behavior

change, and consequently disease progression. Genetic counseling is now shifting to

providing information regarding personal risk reduction; and allowing individuals to

make better-informed medical decisions (Lerman et al., 2002). However, there are also

many diseases for which there is no known method of prevention or cure. In these

instances, it may not be possible to make behavioral recommendations regarding health

behavior change, other than increased medical surveillance. Despite this, individuals on

their own may engage in behaviors they perceive to be beneficial or preventative.

Recently, there has been a large push for clinical psychologists to become more

involved in genetics; and to lend their expertise as clinicians, researchers, and educators

to advance our understanding of the psychosocial costs and benefits of genetic screening

(Fisher et al., 2002, Gallier, 2002; Lerman et al., 2002; Patenaude et al., 2002; Patenaude

et al., 2003). A number of agencies have made the psychosocial implications of genetic

advances a funding priority, including the Human Genome Project which designated 5%

of the total budget to ethical, legal, and social issues (Jeffords & Daschle, 2001; cited in

Patenaude et al., 2002).

Clinical psychologists can play an important role in answering how genetic risk

information impacts individuals in cognitive, affective, and behavioral realms.

Psychologists can assess the role that different personal, social, and cultural factors

contribute to the development or prevention of disease. As clinicians, psychologists can

help individuals and families understand risks, make informed behavioral and








reproductive choices, provide psychosocial support, and evaluate outcomes. A recent

article by Patenaude et al. (2003) highlights the many important roles pediatric

psychologists can play in the research and public policy arenas to inform ethical debates

on the merits of genetic testing; and to provide competent clinical care to affected

families (Patenaude et al., 2003).

While there has been some research addressing attitudes toward genetic testing,

comprehension of genetic information, and the psychological impact of genetic testing,

there is still much to be learned about the impact of genetic screening on individuals'

behavior. This exploratory study examined the behavioral impact of newborn genetic

screening for mothers whose children were found to be at risk for type 1 diabetes.

Currently, little is understood about the specific behavior changes that may result from

knowing one's child is genetically predisposed to a condition for which there is currently

no known prevention method or cure. Additionally, our present understanding of the

etiology of type 1 diabetes suggests that it develops from a combination of both genetic

and environmental influences, which are not well-defined. In the absence of definitive

recommendations from the health care community, mothers of newborns identified as "at

risk" may take actions they believe are effective in preventing type 1 diabetes in their

children. This study assessed the extent of mothers' self-reported behavior changes; and

assessed associations between reported behavior change and maternal psychological

(i.e., anxiety, perceived control, coping), and sociodemographic variables. The upcoming

sections discuss existing literature on the psychological and behavioral impact of genetic

testing, including newborn genetic screening; and our current knowledge of the etiology

and prevention of type 1 diabetes.













CHAPTER 2
REVIEW OF THE LITERATURE

Psychological Impact of Genetic Testing

Literature assessing the psychological impact of genetic testing has largely focused

on predictive testing for Huntington's disease and breast and ovarian cancer susceptibility

(BRCA1/BRCA2 genes). Many studies have examined other uses of risk screening,

including prenatal screening and carrier screening. This review focuses on predictive

genetic testing. Generally, studies of predictive testing have focused on both the short-

and longer-term effects of genetic testing on affective outcomes for screening participants

and their family members. Overall, voiced criticisms that genetic testing leads to poor

psychological adjustment appear unfounded based on published literature; and such

claims may create unnecessary panic (Horowitz et al., 2001; Palmer et al., 2002).

Longitudinal prospective studies examined levels of anxiety and depression before

genetic testing; and provided new information regarding the potential for poor

psychological adjustment. Evidence suggests that contrary to earlier concerns, making

the choice to be tested signals psychological preparedness for the outcome and ability to

handle the news well. Most people who choose to participate in population-based

screening programs do not have a family history and therefore, will most likely expect

and receive a negative result. In screening members from high-risk families (those with a

family history of the disease), patients tend to overestimate (not underestimate) their risk;

and expect to receive positive results (Croyle & Lerman, 1999; Lynch et al., 1999; Lynch

et al., 1993). While one might expect that receiving positive results would result in








clinically significant distress and increased mood symptomatology, studies have found

that a positive test result is usually not associated with clinical levels of anxiety or

depression (Broadstock et al., 2000; Lerman et al., 2002; Schwartz et al., 2002). This

may be partly because reducing one's uncertainty regarding risk may actually decrease

stress by providing relief from what was previously unknown (Marteau & Michie, 1995;

Baum et al., 1997). While some studies suggest elevated scores on measures of distress,

such as depression or anxiety (Shaw, Abrams & Marteau, 1999), scores generally return

to baseline levels after 3 to 12 months (e.g., Croyle et al., 1997; Lerman et al., 1996;

Wiggins et al., 1992). While there may be some immediate distress upon risk

notification, it appears to be neither inevitable nor long-lasting.

A literature review conducted by Broadstock et al. (2000) examined existing

prospective studies of the impact of genetic testing for Huntington's disease, ovarian and

breast cancer, and familial adenomatous polypsosis (FAP). To be included in the review,

studies had to contain both pre and post measures of psychological distress. Examples of

measures of distress used in these types of studies include the Impact of Events Scale

(IES; Horowitz, 1979), State-Trait Anxiety Inventory (STAI; Speilberger et al., 1970),

and the Beck Depression Inventory (BDI; Beck, 1961). The authors' extensive search

uncovered 11 studies, none of which found an increase in distress (defined as general or

test-specific anxiety or depression) at any point in the 12 months after testing. After

notification, distress decreased in individuals who received either a positive or negative

test result. However, this decline was greater and more rapid in those who received

negative test results. Furthermore, in regression models, the actual test result rarely

predicted psychological outcomes beyond the first month post-risk notification.








Congruent with the Stress Disease Coping Model proposed by Baum et al. (1997), the

individual's pretest emotional state, social support, and expectations were the most

predictive of subsequent distress (Marteau & Croyle, 1998), suggesting that personal

variables may play a role in how one handles genetic risk information.

Taken together, these studies generate little empirical support for the notion that

genetic testing is associated with adverse psychological outcomes (Lerman et al., 2002).

However, most of these studies involved participants in research registries and these

results may not generalize to the broader population. It should be noted that those in

clinical settings may be self-referred and more naive, and thus, less equipped to cope with

knowledge of their risk status (Broadstock et al., 2000).

Previous research suggests that psychological distress may be associated with

specific personal (e.g., optimism) or demographic characteristics (e.g., race, education).

Audrain et al. (1998) studied women with a first degree relative with breast or ovarian

cancer before testing and found that pre-test distress was predicted by age, ethnicity,

marital status, optimism, perceived control, and overestimated risk perception. Those

who perceived less control, were younger, not Caucasian, married, and less optimistic

were more likely to experience greater distress before risk screening (Audrain et al.,

1998). Hughes et al. (1997) studied ethnic differences in knowledge and attitudes

regarding testing for BRCA1 gene in at risk women; and found that African American

women had lower levels of knowledge, but more positive attitudes toward genetic testing

than Caucasian women. Risk perception appears to vary by ethnic status, with African

American women who have a family history of breast cancer having greater concerns








about their own personal risk of breast cancer and appearing more likely to avoid breast

cancer-related thoughts and feelings (Hughes et al., 1996).

Studies have examined coping strategies associated with receiving genetic risk

information to determine if coping mediates distress in genetically at risk individuals.

For these types of studies, coping has been conceptualized as the degree to which one

either seeks more information (monitoring) or avoids or distracts oneself from the

situation (blunting/avoidance) (Miller, 1987). Studies have used the Miller Behavioral

Style Scale (MBSS; Miller, 1987), to determine the style in which individuals deal with

risk information given. It has been hypothesized that individuals cope with health threats

in one of two ways. In general, an interaction has been reported between the amount of

information provided and whether monitoring or blunting characterizes the individual's

coping style.

Studies have found that matching the amount of information received to the amount

the individual desires, lowers distress (Ludwick-Rosenthal & Neufeld, 1993; Miller,

1980; Miller & Managan, 1983). There does not appear to be consensus regarding

whether coping mediates distress in those notified of increased genetic risk status. For

example, in a study of patients from high-risk families screened for BRCA1/BRCA2

genes, coping efforts (both active and avoidant) were associated with higher levels of

distress prior to notification; whereas post-notification distress was associated with the

test result, not coping (Tercyak et al., 2001a). Lerman et al. (1993) found that a high

level of monitoring in women at risk for breast cancer predicted an increase in distress

over a 3-month follow-up period; whereas Vernon and colleagues' (1997) study of FAP

screening found the opposite to be true. Anxiety appears to be influenced by whether or








not the event is controllable and by the amount of information given to an individual

(Miller et al., 1989). Sex differences may also play a role in risk appraisal and coping.

Marteau et al. (1997) found that women have a greater fear of threat, worry more about

negative outcomes, and perceive greater risks from technology than men; whereas men

show higher threat minimization after positive carrier testing for cystic fibrosis (CF).

Behavioral Impact of Genetic Testing

A major question of interest to researchers is whether results of predictive genetic

testing lead to increases or decreases in health behaviors and medical surveillance. Does

informing people of genetic susceptibility to disease motivate them to take action to

reduce their risk? Or does knowing that one is genetically predisposed suggest a sense of

pre-determined destiny and perceived immutability (Senior, Marteau, & Peters, 1999;

Senior, Marteau, & Weinman, 1999)? Marteau & Lerman (2001) reviewed literature

related to cancer, smoking, and heart disease and espoused that providing genetic

information may not increase motivation to change behaviors and may even result in

reducing motivation. However, these authors also suggested that genetic information

might better facilitate change if individuals are offered effective risk-reducing

interventions tailored to their genetic risk. Most existing research studies in this area

focused on cancer, particularly, breast cancer screening; and physician recommended

behaviors, such as mammography and breast self-examinations.

Studies have examined the impact of distress caused by risk notification to

determine if distress predicts health-protective or preventative behaviors. Croyle &

Lerman (1999) reviewed studies on how coping and distress influenced the processing of

genetic risk information and subsequent decision-making. Studies have found that risk

information can be too anxiety provoking for some; and therefore, anxiety acts as a








barrier to following through with screening recommendations (Kash et al., 1992; Lerman

et al., 1994; Lerman et al., 1993). Other studies have suggested that increased distress or

worry actually increases health behaviors (Burnett et al., 1999; Diefenbach et al., 1999)

and even leads to excessive health practices (i.e., breast self-examinations) (Epstein et al.,

1997; Epstein & Lerman, 1997; Lerman et al., 1994; Lerman & Schwartz, 1993). Epstein

et al. (1997) found that those who were excessive in their protective behaviors were more

likely to be African American, older, and less educated. These findings may be

explained by results from Audrain et al. (1998) suggesting that African American women

undergoing genetic screening experience greater distress and have lowers levels of

cancer-related knowledge. Others have found no significant relationship between distress

and adherence to recommended medical surveillance (Lerman et al., 2000). Taken

together, these studies suggest that an inverted U-shape curve may explain the

relationship between distress and screening behavior, with highest rates of adherence

predicted by a moderate level of anxiety (Hailey, 1991; Lerman et al., 1991; Lerman &

Rimer, 1993).

There have been conflicting reports of how perceived versus actual risk impacts

screening behaviors (Hailey, 1991). Overall, some cancer studies found an increase in

screening rates in those informed they are genetically at higher risk (Meiser et al., 2000;

Ritvo et al., 2002; Schwartz, Taylor, et al., 1999), while other studies have found

adherence rates similar to those of the general population (Bratt et al., 2000; cited in

Marteau & Lerman, 2001; Clavel-Chapelon et al., 1999). In a study of colon cancer

screening and behavior intentions, half of respondents indicated that they would decrease

their use of screening tests and make fewer attempts to reduce their dietary fat intake if








their test results indicated that they were at low risk (Lerman et al., 1996). Women with a

family history of breast cancer were more likely to perceive higher genetic risk and

engage in appropriate screening behaviors (Hailey et al., 2000; McCaul et al., 1996).

Perceived risk has been found to predict screening compliance above and beyond the

actual risk associated with family history of a disease (Aiken et al., 1994). Women who

perceived themselves to be at greater risk were more likely to engage in initial as well as

repeated screenings (Lerman et al., 1990; McCaul et al., 1996). However, the perception

of risk does not appear to be necessarily related to the accuracy of risk. It has been

suggested that accurate recall of risk information does not necessarily lead to risk-

reducing behavior. Therefore, many have begun to examine the links between risk

perception and risk-reducing behavior, particularly the potentially mediating variable of

disease-specific worry or anxiety.

Many studies have been conducted examining the role of information-seeking and

health behavior change, particularly as it relates to public health issues (i.e,. HIV/AIDS

prevention). Rakowski (1990) conducted a randomized survey among adults in the

general population and found a positive association between more frequent information

seeking and personal health-related practices. However, hardly any studies have

examined this issue in the context of genetic screening. One such study examined

women at genetic risk of ovarian cancer; and found that monitors (information-seekers)

demonstrated greater adherence to behavioral recommendations, such as attending cancer

screenings (Wardle, 1995).

Demographics factors are also important in predicting health behaviors. Schwartz

et al. (1999b) found that women with less education who were at risk for breast cancer








and screened, reduced their use of mammography after breast cancer risk counseling.

Additionally, studies have found that women from ethnic minority groups and women

with lower levels of education reported greater disease-specific worry (Aiken et al., 1994;

Audrian et al., 1998) and retained less information about screening programs in general

(Browner et al., 1996, Donovan & Tucker, 2000; Hughes et al., 1997)

Lerman et al. (1993) found that reproductive behaviors were also impacted by

cancer screening. In a study of women under age 49, 22% reported that they would be

less likely to have children if they tested positive; and 17% reported being uncertain

whether they would continue a pregnancy. Other studies that assessed the reproductive

impact of genetic testing, found that 46-83% of subjects within reproductive age in the

general population would not have children or would limit further reproduction if they

tested positive for a disease gene (Kessler et al., 1989; Schoenfeld et al., 1984; Zerres et

al., 1986). Unfortunately, little is known about cancer screening and other lifestyle

changes involving smoking, physical activity, or diet (Marteau & Lerman, 2001).

In the current study, both lifestyle and health surveillance behaviors were included

as outcomes. Similar to other studies, relationships between reported behavioral

outcomes and psychological variables (such as anxiety, perceived control, coping, and

risk perception) were assessed.

Theoretical Models of Genetic Screening and Behavioral Change

Most studies in the area of genetic testing fail to use a theoretical model to conduct

or interpret the findings. However, theoretical models are important as they can serve as

a contextual framework for interpreting complex results. Generally, most models of

health behaviors assume that the motivation for health-protective behavior comes from








both the anticipation of a negative health outcome and the hope of avoiding it (Weinstein,

1993).

Tercyak (2000) advocated for conceptualizing the impact of genetic testing as a

family systems issue (Tercyak, 2000). The rationale for this was that genetic risk

information impacts multiple family members and the family's dynamics as a whole.

Additionally, Tercyak (2000) reasoned that families with no history or experience with a

particular illness would fare differently than those with a family member who is already

ill. Pre-existing illness in one member of the family provides personal experience and an

increased knowledge base about that condition for other family members that families

without a history of that particular disease would not have. Therefore, the meaning and

implications of genetic test results would be different depending on family history. By

translation, subsequent behavior change may also differ.

Rolland (1999) advocated the use of a specific model, the Family Systems-Illness

Model, when examining the psychological impact of predictive testing. Rolland (1999)

stated it to be a "useful guide," as it emphasizes the psychosocial demands of different

disorders over time and emphasizes the key components of family functioning (i.e.,

multigenerational patterns), illness life cycles, and belief systems. Rolland recognized

that psychosocial challenges varied according to biological variables, including the

degree to which a disease is influenced by both the environmental and genetic factors and

the degree to which prevention is possible (Rolland, 1999). Rolland advocated for

longitudinal follow-up of families after genetic testing, as psychosocial strains related to

knowledge of future risk do not just present themselves upon result notification, but will

tend to surface at major life-cycle transitions (Rolland, 1999). These challenges








influence family decision making and health behaviors. In his descriptions, Rolland did

not address specific mechanisms of his model or how they apply to screening results

(Rolland, 1999). It is apparent that Rolland was using his model as a conceptual guide

rather than as a testable model. To date, although genetic testing is recognized as

impacting the daily lives of entire family units, there has been no formal testing of a

family systems model pertaining to the psychosocial impact of genetic testing.

The Health Belief Model (HBM) (Rosenstock, 1974) has been used most

frequently in previous genetic testing research. Investigators have used this model to

explain preventative health care behavior in the context of one's perceived susceptibility

to an illness, the perceived severity of that illness, and the potential benefits and costs of

performing a specific behavior to reduce the risk (i.e., Aiken et al., 1994; Becker &

Maiman, 1975; Frets et al., 1990; Rowley et al., 1991; Sagi et al., 1992; Shiloh & Saxe,

1989; Sorenson et al., 1987). These factors are hypothesized to be predictive of the

decision to engage in health behavior change or to increase surveillance. Preventative

action is most likely taken when individuals perceive themselves to be at risk for a

serious disease and when the benefits to action outweigh the costs of not engaging in the

specific health behavior. For the purposes of genetic screening studies, most have

focused more on the perceived susceptibility component rather than the perceived

severity.

The HBM has several significant weaknesses. The perceived severity component

has not gained strong empirical support as a major predictor for preventative behavior

(Leventhal et al., 1983). Additionally, the HBM has proven influential for health

attitudes, but not consistently for health behaviors. HBM also assumes that health








behaviors arise from a single rational decision based on cost-benefit analysis, which may

be an oversimplification (Home & Weinman, 1998). Finally, this model does not specify

underlying beliefs, how to change beliefs (Home & Weinman, 1998) or what beliefs need

to be changed in order to change behavior. Other social variables and personal factors, as

reviewed in later sections, may have more importance in influencing health behavior than

this model would suggest.

While many studies have used the HBM, Lerman et al. (1997) endorsed a different

model, the Self-Regulation Model of Health Behavior (SRM; Leventhal, 1965), to more

specifically address why women at risk for breast cancer experiencing too great or too

little worry were less likely to practice risk-reduction behavior. AccordiAg to Leventhal

(1970), a health threat results in both cognitive and affective responses, which occur in

parallel. The SRM suggests that moderate levels of perceived health threat (e.g.,

diagnosis of cancer in a relative) engender a moderate level of concern/worry, which in

turn, leads individuals to take actions that will reduce the anxiety caused by a health

threat. Fear arousal coupled with an action plan leads to a "cognitive representation" of

the threat (Home & Weinman, 1998). Excessive cancer-related anxiety might produce

avoidance of screening, and at least a minimal level of anxiety is necessary to motivate

these behaviors. Similar to this model is the Fear Arousing Communications Theory

(Janis and Feshback, 1953), which states some degree of fear arousal is needed to predict

adoption of health care behaviors. If individuals are not concerned, they may deny the

threat; and if they are overconcerned, they may come to avoid preventative health

practices (Kash et al., 1992).








However, contrary to the aforementioned models, communication of a threat alone

may be insufficient to change one's behavior (Home & Weinman, 1998). Relatively few

studies have used the SRM, perhaps because of its complexity, which makes it difficult to

operationalize. However, there does appear to be some empirical support for this model

in studies of medication adherence in hypertension (Meyer et al., 1985) and regimen

adherence in diabetes (Gonder-Frederick & Cox, 1991; cited in Home & Weinman,

1998).

The model that best informs the current study is a transactional model of stress and

coping, known as the Stress-Disease Risk-Coping Model, which is a comprehensive

model specifically designed for studies of genetic testing (Baum et al., 1997). Baum and

colleagues' (1997) model is based on the concept of risk appraisal espoused by Lazarus

& Folkman (1984), in which primary appraisal involves the judgment of the threat of a

stressor; and secondary appraisal consists of a judgment regarding available resources to

deal with the threat. This model is particularly concerned with the relationship between

uncertainty and risk perception, which influence one's stress response, and consequently

affect one's behavior. Proponents believe this model is useful in predicting

psychological and behavioral responses to genetic testing results (Lerman, 1997). This

model hypothesizes that distress and behavior changes will be affected by the interaction

between personal factors (perceived risk influenced by family history, optimism), actual

test results, characteristics of disease, and degree of uncertainty remaining after testing

(Figure 1). The central component of the model involves the appraisal process regarding

the test results. In this step, appraisal of increased certainty regarding future outcomes is

coupled with perceived available options for action. This appraisal process is influenced








by the degree to which one perceives him/herself to be at risk; and is influenced by

surveillance and prevention options, along with other variables (such as social support,

optimism, perceived control, etc.). This appraisal process is associated with the stress

response to the information. The more resources available, the better one may be able to

cope with the stressor (Wallston, 2000). This stress response and these coping

mechanisms in turn, relate to behavioral consequences. This model suggests that the

adoption of health behaviors is influenced by personal factors, perceived risk, perceived

control, distress, and coping resources. In describing the model, Baum et al. (1997)

review the studies that influenced the design of the model, indicating that this model was

originally informed by both theoretical and empirical evidence.

Baum and colleagues' (1997) model is fairly new; and presently, no known

published studies have tested this model. Despite a lack of available empirical support,

this appears to be the most comprehensive and relevant model to use when examining the

behavioral impact of genetic screening. The model incorporates many variables that have

been examined in the context of genetic screening studies (i.e., perceived risk distress). It

should be noted that this model applies to individuals and unfortunately, does not directly

incorporate the family unit, which is undoubtedly affected by results of genetic testing.

Despite this limitation, for the purposes of the current study, this model was applied to

maternal behavior change in response to risk identification in their children.

Newborn Genetic Screening

The current study examined the impact of genetic screening of infants, which is

ethically more complicated than testing within adult populations. The Institutes of

Medicine (IOM) reports that 3% of children have an illness or disorder of probable

genetic origin (IOM, 1994). Understandably, while one would want to extend the








benefits of biomedical advances to children, additional considerations are involved.

Consequently, organizations have generated ethical guidelines for performing genetic

tests in pediatric populations, including the American Academy of Pediatrics (AAP),

American Society of Human Genetics and American College of Medical Genetics,

Clinical Genetics Society, and Institutes of Medicine (American Society of Human

Genetics and American College of Medical Genetics, Boards of Directors, 1995; Clarke

et al., 1994; Wertz et al., 1994). These guidelines are especially important for testing for

diseases for which there are no known cures or modes of prevention. In the absence of

clearly beneficial treatments or effective methods of prevention, it is difficult to justify

the genetic testing of children and adolescents, including newborn screening. Because

young children are unable to understand the value of genetic information for their own

lives, particular care must be exercised by parents and pediatricians when making

decisions about genetic testing for children (AAP Committee on Bioethics, 2001). Other

important factors to consider include the psychological and economic impact on the

family, time of disease onset, degree of risk, and possible medical benefits.

For these reasons, newborn genetic screening is controversial, especially for those

diseases with no known cure. The Institute of Medicine (IOM) report recommended

three principles (IOM 1994) to govern the maintenance of existing screening tests and the

introduction of new newborn tests:

identification of the genetic condition must provide a clear benefit to the child
a system must be in place to confirm the diagnosis
treatment and follow-up must be available for affected newborns

In other words, newborn genetic screening is supported only if the infants would benefit

from early identification and prevention/treatment. Other guidelines exist that allow









regulated research protocols to test children when no immediate medical benefit exists

but the contribution to scientific knowledge is great (American Society of Human

Genetics/American College of Medical Genetics Board of Directors, 1995; Clark, 1994).

Newborn screening is the most widely used type of genetic screening, with nearly

all states in the U.S. mandating newborn screening for phenylketonuria (PKU) and

congenital hypothyroidism, in which early diagnosis leads to treatment and better medical

outcomes (IOM, 1994). Recently, in some states, newborn screening has expanded to

include testing for congenital adrenal hyperplasia and cystic fibrosis (CF) (in WI, CO,

and WY). As a point of comparison for CF screening, only 6% of newborn U.S. children

are screened versus 92% of newborns in Australia (Wilcken & Travert, 1999). In the past

decade, newborn screening has been implemented in research settings to test for risk of

type 1 diabetes (discussed in a later section).

Currently, there has been relatively little research on the psychological implications

of screening newborns. Much of the research has occurred in other countries. Studies

from Wales on newborn screening for Duchenne muscular-dystrophy, an incurable X-

linked condition eventually leading to death during early adulthood (Fenton-May et al.,

1994), suggest that the screening has been well-received, with few adverse psychological

outcomes reported and a participation rate of 90% for eligible families (Bradley et al.,

1993; Parsons et al., 2002). Such a favorable outcome is not always the case. A

screening program for alpha-1-antitrypsin (lung disease) in newborns in Sweden had to

be terminated prematurely because of adverse effects. These included negative changes

in family dynamics and parental nonadherence to medical recommendations, including

increased smoking behavior (McNeil et al., 1989).








Most of the newborn screening literature has been dedicated to screening for cystic

fibrosis (Kerem et al., 1989). CF screening remains controversial (especially for those

without a family history) as some view the psychological costs as outweighing the

medical benefits of early diagnosis (Wald & Morris, 1998). Adverse psychological

outcomes have included greater parenting stress (Baroni et al., 1997) and a small

percentage of mothers experiencing short-lived feelings of rejection toward their child

(Al-Jader et al., 1990). It should be noted that these effects might also be present when

diagnosis is made through conventional means when children are a little older (Al-Jader

et al., 1990; Boland & Thompson, 1990; Wilcken et al., 1983). Boland & Thompson

(1990) found newborn screening versus traditional screening did not produce greater

overprotectiveness in mothers. The delay in diagnosis that occurred when screening was

not conducted resulted in greater maternal distress and anger. Therefore, these

psychological risks do not appear significant when the potential benefits of newborn

screening include better health outcomes due to earlier initiation of treatment (Waters et

al., 1999).

Further exploration of the psychological effects of newborn screening is an

important area of research as new genetic tests become available; and decisions will need

to be made regarding the appropriateness of their use. Whether testing is conducted in

the general population or in research settings only; and whether it is conducted with all

families or just those with a family history of the disease, are important questions to be

answered. How risk information is understood and used by families; and whether it then

translates into emotional and/or behavioral changes are key areas for future research.








Type 1 Diabetes: Etiology and Prevention

In the U.S., the prevalence of insulin dependent diabetes mellitus (IDDM; type 1

diabetes) is approximately 2-3/1,000 children, which makes it one of the most prevalent

childhood chronic illnesses (Arslanian et al., 1997; LaPorte et al., 1995). Annual

incidence is estimated to be over 12,000 children each year, with peak incidence of

diagnosis occurring between five and six years of age and again between the ages of

eleven and thirteen. The prevalence of type 1 diabetes is higher among Caucasians

(National Diabetes Data Group, 1995). In type 1 diabetes, the body produces little or no

insulin due to the autoimmune destruction of islet cells in the pancreas. This leads to

high blood glucose levels. Type 1 diabetes is thought to be the endpoint of an

immunologically mediated attack on pancreatic beta cells. It is an autoimmune disorder

where islet cells are destroyed by an immune response, or more simply, destroyed by

cells within one's own body that normally protect a person from germs. Complications

of type 1 diabetes can include retinopathy, blindness, renal disease, neuropathy, lower

extremity ulcers, digestive disorders, heart disease, and vascular disease (National

Diabetes Data Group, 1995). The average life span for those with diabetes is generally

shortened due to vascular complications. With no cure available, type 1 diabetes is

currently medically managed by administering insulin on a daily basis and adhering to a

specialized diet and exercise program. These daily treatment demands can greatly affect

an individual and their family's lifestyle.

In addition to the impact on the family, type 1 diabetes is a substantial societal and

economic burden. Therefore, an obvious need exists for diabetes prevention. Currently,

diabetes (including treatment, prevention, and research) consumes one in every seven

dollars spent on health care in the U.S. (Schatz et al., 2002). Often diabetes is not








diagnosed until a patient is having a crisis episode (ketoacidosis), which can lead to

increased medical complications and longer hospitalizations (Beisswinger, 2000; cited in

Schatz, 2002).

Unfortunately, we do not fully understand the etiology of type 1 diabetes. Nearly

90% of type 1 diabetes occurs in families with no history of the disease (Dalquist et al.,

1985) and there is only a 30-50% concordance rate among monozygotic twins (National

Diabetes Data Group, 1995; Kyvik et al., 1995; LaPorte et al., 1995). However,

approximately 3-6% of first-degree relatives with type 1 will develop the disease as well

(Tillil & Kobberling, 1987). The chance of developing diabetes for the general

population is about 1 in 300 while, for those with first-degree relatives with diabetes, the

chances increase to 1 in 20 (National Diabetes Data Group, 1995). These data suggest

IDDM is caused by a combination of genetic and environmental factors.

It is generally thought that environmental triggers initiate an autoimmune process

that leads to the destruction of pancreatic beta-cells and consequently, type 1 diabetes. It

is still unclear the degree to which these environmental factors play a role. In order to

determine the interactions between genetics and the environment, longitudinal studies are

needed to follow at risk individuals over time. To date, research studies have suggested

viral illness enteroviruss and rotovirus) may be one class of environmental triggers

(Akerblom & Knip, 1998, Couper, 2001, Dorman et al., 1995). Additionally, Classen &

Classen (2001) argue that timing of vaccines increases the risk of type 1 diabetes. The

risk of type 1 diabetes decreases when children receive vaccinations after at least two

months of age, arguing for the benefits of delayed immunization schedules. A recent

study found increased social mixing in young children (i.e., attendance of daycare) in








early infancy was protective against the development of type 1 diabetes because it

increased exposure to infections and strengthened immunity (McKinney et al., 2000).

However, there has been no other direct evidence in favor of such an association

(Akerblom & Knip, 1998). Finally, early emotional stress may also be a contributing

factor (Thernlund et al., 1995).

Dietary factors have been implicated as important environmental contributors to the

development of type 1 diabetes. Such dietary factors included not breastfeeding

(Akerblom & Knip, 1998), early introduction of cow's milk (Akerblom, et al, 1993;

Gerstein, 1994; Virtanen et al., 2000), high intake of nitrites/nitrates (Virtanen & Aro,

1994), accelerated prenatal growth (Dahlquist et al., 1996), high intake of proteins

(Akerblom & Knip, 1998), high intake of carbohydrates (Akerblom & Knip, 1998) and

increased weight gain in infancy (Hypponen et al., 1999). Although, based on both

animal and human studies, the most likely putative dietary factors are hypothesized to be

cow's milk, proteins, and nitrates/nitrites (Akerblom & Knip, 1998

The greatest amount of research regarding environmental factors related to type 1

diabetes has examined whether breastfeeding is protective and how this interacts with

exposure to cow's milk in infancy. Cow's milk is implicated because it has a higher

protein content, specifically the protein casein, than that found in human breastmilk.

Many studies have been conducted to address this issue with no firm consensus reached

(Akerblom & Knip, 1998; Couper, 2001). To examine the role of cow's milk, the multi-

national Trial to Reduce IDDM in the Genetically at Risk (TRIGR) is ongoing to

determine if delayed exposure to cow's milk until after 6 months of age will have an

effect on the subsequent development of diabetes (Karges et al., 1997; Schatz, 2002;








Virtanen et al., 1997). Schatz & Maclaren (1996) warn it is premature to recommend

eliminating cow's milk from an at risk child's diet as there is no convincing evidence to

suggest the nutritional benefits of milk for young children outweigh the potential dangers.

To answer questions regarding the prevention of type 1 diabetes, The Diabetes

Prevention Trial (DPT-1) was initiated in 1994 to determine whether subcutaneous or

oral insulin could prevent or delay the onset of diabetes in at risk relatives (DPT-1 Study

Group, 1995, 2002). Within this large-scale randomized, nonblind study, there were two

separate trials for the two types of insulin administration. Three hundred and thirty nine

participants, who were between 3 and 45 years of age and had a first degree relative with

type 1 diabetes were randomized in the subcutaneous insulin trial (out of 84,228 screened

first degree relatives). To be eligible, participants had to be determined as "high risk,"

defined as a 50% chance of developing type 1 diabetes over the next five years. This was

determined by the absence of protective genetic markers, positive antibody testing, and a

low first-phase insulin response in glucose tolerance testing. Participants were

randomized to either the intervention group, which received low dose subcutaneous

insulin, or the close observation group, and all of whom were followed for an average of

3.7 years. Results from the subcutaneous insulin trial were recently published. Results

suggested that injected insulin does not delay or prevent type 1 diabetes (DPT-1 Study

Group, 2002). The oral insulin trial is ongoing and results are not currently available.

In contrast to type 1 diabetes, type 2 diabetes (non-insulin dependent diabetes) is a

different form of diabetes that is considered a metabolic disorder, rather than an

autoimmune disease. It is usually diagnosed in adulthood, although it can develop in

childhood. In type 2 diabetes, the body is unable to make enough or properly use insulin;








however, beta cells are preserved. This is in contrast to type 1 diabetes, in which beta

cells are destroyed, leading to insulin deficiency. Type 2 diabetes accounts for 90-95%

of diabetes and researchers have found obesity and a sedentary lifestyle to be contributing

factors, as well as genetic predispotion (Fletcher et al., 2002). Because the prevalence of

type 2 diabetes is rapidly increasing to epidemic proportions, the health care community

and the media have recently focused significant attention on type 2 diabetes, advocating

for healthy lifestyle changes. Recent research has indicated moderate diet and exercise

reduces risk for type 2 diabetes more effectively than even oral insulin (Tuomilehto, et

al., 2001). A healthy diet is effective because it reduces the insulin load and exercise is

effective because physical inactivity reduces tissue glucose tolerance and is associated

with insulin resistance. Scientific evidence is not clear as to whether these same

behaviors have an impact on the development of type 1 diabetes; however, at the present

time, it seems unlikely (Schatz, personal communication). People who do not understand

the distinction between type 1 and 2 diabetes may apply recommendations for type 2 to

their children at risk for type 1. The current study explored whether this hypothesis was

true for our sample population of mothers of at risk young children.

Prediction and Pre-Symptomatic Screening for Type 1 Diabetes

While we do not fully comprehend the natural history of the development of

diabetes, we do know that the destruction of pancreatic cells is a precursor to type 1

diabetes and begins long before overt symptoms. It is currently possible to detect

pancreatic cell destruction and identify those at risk for developing Type 1 diabetes. Riley

et al. (1990) found the determination of islet-cell antibodies in relatives of probands with

Type 1 diabetes increased an individuals' risk for developing the disease in the future.








Currently there are two types of screening for diabetes, autoantibody screening and

genetic screening. The most recently developed test, genetic screening, is typically done

in newborns to determine the present of high-risk genetic markers (DR 3/4, DR 4/4, DR

3/3) in the Human Leukocyte Antigen (HLA) region (the Major Histocompatability

Complex (MHC)) on chromosome 6. This is an area that helps control immune response,

and such markers are known to confer 50% of the genetic risk for Type 1 diabetes (Yu et

al., 1999) (Table 1). The second type of testing, antibody screening, is a process that has

been in existence for longer and detects islet-related autoantibodies, including

autoantibodies to insulin (Christie et al., 1994; Landin-Olsson et al., 1992), GAD or islet

antigen-2 (IA-2), as well as islet cell antibodies (ICA) (Riley et al., 1990; Schatz et al.,

1994) present in blood serum. It has been shown the presence and number of these

antibodies is directly related to risk for type 1 diabetes (Knip, 1998). An ICA positive

result signifies that the process of beta cell destruction has begun and therefore, those

who are ICA positive are farther along in the process of developing type 1 diabetes. For

example, individuals who test positive for ICA have approximately a 45% chance of

developing diabetes in the next ten years. Antibody screening has been conducted with

children and adults and used as a primary screening method and as follow-up to newborn

genetic screening.

While critics oppose screening for risk of developing type 1 diabetes before

symptoms appear, Schatz, et al. (2002) argue it is very important to the future of diabetes

prevention research. The authors assert screening helps us in a number of ways: it allows

us to better understand the prediabetic period and diabetes pathogenesis, assists in

identification of individuals for prevention trials, facilitates earlier diagnoses which








reduces the mortality and morbidity associated with type 1 diabetes (Schatz et al., 2002).

Currently, genetic screening is only conducted within research settings since widespread

screening of the general population, when there is no available effective intervention, is

considered unethical. Many longitudinal studies are now ongoing to follow newborns

found to be genetically at risk for Type 1 to better study the development of diabetes.

These trials, taking place in Germany (BABYDIAB), Finland (DIPP), Denver, CO

(DAISY), and Gainesville, FL (PANDA), include studies of the participants from the

general public and at risk families (e.g., Nejentsev et al., 1999; Rewers et al., 1996;

Schatz et al., 2000; Schenker et al., 1999; Ziegler et al., 1999).

Opponents of screening argue that without a prevention strategy, studies should

avoid disclosing results to participating families and that if disclosure is necessary then

research should only be conducted with infants who have a first degree relative with type

1 diabetes (Friedman Ross, 2003). Critics argue that screening under any other

circumstances may result in harm to children and their parents. Friedman Ross (2003)

stated that genetic screening can only convey at most a susceptibility that is a 20%

probability. She claims that as a result of these tests, parents may prepare unnecessarily

and treat their child as ill (Friedman Ross, 2003). As the debate continues regarding the

merits of genetic screening of the general population and as interest in diabetes

prevention continues to rise, research on the psychological and behavioral impact of

genetic screening becomes timely and highly relevant.

Psychological Impact of Diabetes Screening

Relatively little research has been conducted examining the parents' psychological

reactions to participation in a newborn screening program for type 1 diabetes. However,

parents have indicated favorable attitudes towards risk screening and prevention trials for








type I diabetes (Lucidarme et al., 1998; Ludvigsson et al., 2002). To explore the

psychological impact of risk screening, Dr. Johnson and her research group have

conducted several studies of adults and children identified as at risk via autoantibody and

newborn genetic screening (Carmichael et al., 2003; Johnson, 2001; Johnson &

Carmichael, 2000; Johnson & Tercyack, 1995; Johnson et al., 1990). As explained

above, a determination that an individual is at risk as identified through presymptomatic

screening does not mean an individual will definitely develop diabetes. How this

information and level of uncertainty impacts individuals, particularly newborns and their

families, is an important factor to consider when evaluating the ethical nature of genetic

risk screening.

In one of the first studies in this area, Johnson et al. (1990) reported individuals

found to be at high risk (as identified through ICA screening) and their family members

exhibited clinically significant levels of anxiety subsequent to at risk notification. Those

testing ICA+ were told their chances of developing diabetes were 50%. Johnson and

Tercyak (1995) subsequently found notification of islet cell antibody positive (ICA+)

status had an emotional impact on the at risk individual (adults and children) and their

family members (i.e., spouses, parents). Initial notification was associated with

considerable situationally-specific anxiety (as measured by the state portion of the State-

Trait Anxiety Inventory (STAI; Speilberger, 1970) and the State-Trait Anxiety Inventory

for Children (STAI-C; Speilberger, 1973) in both individuals with the risk and their

family members. This was especially true in parents of ICA+ children. In addition,

parent and child anxiety was highly correlated. However, initial anxiety seemed to

decrease to normal levels over time, as measured in a 4-month follow-up interview.








In a similar study with fewer participants, Galatzer et al. (2001) examined antibody

positive children (n=10) and their parents using the Impact of Events Scale (IES;

Horowitz, 1979) and found that high levels of distress reported by parents upon results

notification decreased by the 3-month interview. Galatzer et al. compared their results

with a study of parents of children newly diagnosed with type 1 diabetes (Kovacs, 1985)

and found similarly strong emotional reactions, but more so in the group of parents of

children with diabetes. Another small-scale study conducted by Yu et al. (1999) (n= 88)

found notification of high-risk genetic status in newborns was not associated with

increased parenting stress as measured by total stress score (TSS) of the Parenting Stress

Index (PSI; Abidin, 1990) more than three months after notification.

A follow-up study to Johnson & Tercyak (1995) examined how individuals found

to be at risk (ICA+) coped with their own or a loved one's at risk status, by administering

the Ways of Coping Checklist-Revised (WCC-R; Folkman & Lazarus, 1980) (Johnson &

Carmichael, 2000). Using this multi-dimensional measure allows for closer examination

of coping styles (i.e., problem-focused, seeking social support, wishful thinking,

avoidance, self blame) beyond the concept of monitoring vs. blunting found in previous

cancer genetic screening studies. Johnson & Carmichael (2000) found at risk children

used more avoidance coping (e.g., tried to forget the whole thing, kept your feelings to

yourself; slept more than usual) than at risk adults, mothers of at risk children, or spouses

of at risk adults. At risk children also used more wishful thinking (e.g., hoped a miracle

would happen; wished the situation would go away) than at risk adults. Initial state

anxiety in response to risk notification was related with subsequent coping as mothers

who were more anxious tended to use more wishful thinking, avoidance, and they tended








to blame themselves for their child's at risk status. Coping strategies appeared to

influence the maintenance of anxiety over time as mothers who blamed themselves

tended to remain anxious.

In the late 1990s, testing moved from biological to genetic markers, and from

family cohorts to the general population. Carmichael et al. (2003), Johnson &

Carmichael (2000) and Johnson et al. (submitted) interviewed mothers of infants at risk

for developing Type 1 diabetes as identified through participation in the longitudinal

Prospective Assessment of Newborns for Diabetes Autoimmunity (PANDA) study

(Schatz, 2000). As described earlier, PANDA involves HLA genotyping and serial

antibody screenings over time. Interviews assessing the psychological impact of

participation in PANDA were conducted approximately 4 weeks post-notification, and

again 4 and 12 months after notification. Similar to the ICA+ studies, they found

maternal anxiety levels were clinically elevated after initial notification of risk status, but

appeared to dissipate over time to normal levels (Johnson et al., submitted).

Risk understanding was examined in mothers who participated in the initial and 4-

month follow-up PANDA interviews (Carmichael et al., 2003). Almost 75% of mothers

gave a correct estimate of their child's genetic risk at the initial interview; however, over

time, mothers were less likely to be accurate and more mothers underestimated their

child's risk. Overall, very few mothers overestimated their child's risk. Mothers who

were Caucasian and who had higher levels of education were more likely to be accurate.

Mothers whose children were in the highest risk group were least accurate. Mothers of

children with a family history of a first degree relative with diabetes were more likely to

underestimate their child's risk at the initial interview. Maternal anxiety was a predictor








of risk underestimation at the 4-month interview, but was not significant in predicting to

earlier underestimation or accuracy at either time point. As one might expect, mothers

who were more anxious were less likely to underestimate their child's risk.

In studies of maternal anxiety in this population, initial anxiety levels were found to

be higher in mothers who were Hispanic, with less education, in those whose infants were

at greater risk, and in mothers who overestimated their child's actual risk (Johnson et al.,

submitted). Coping strategies also appeared to be related to anxiety as wishful thinking

and blaming one's self predicted anxiety at the 4 and 12-month follow-up interviews

(unpublished data). As explained in later chapters, participants for this study were

recruited from this larger sample.

These studies, taken together, suggest newborn screening does not have long-term

detrimental effects on parental adjustment, as measured by either anxiety or stress.

Additionally, it appears that a majority of mothers correctly recall their infant's risk with

few mothers overestimating their child's risk and consequently becoming more anxious.

These findings are congruent with other studies of genetic testing previously discussed.

It is likely parents' reactions to the news and subsequent coping style may influence an

individual's or family's decision to participate in longitudinal trials or natural history

studies, such as PANDA, that will provide the scientific bases of a prevention or cure for

type 1 diabetes. These studies can play an important role in informing debates about the

ethics of newborn screening.

Behavioral Impact of Diabetes Screening

Johnson & Tercyak's (1995) study of ICA+ children and adults, assessed after

notification of screening results, found 52% of ICA+ children and 24% of ICA+ adults

reported making a change in their behaviors and/or lifestyle in an attempt to delay or








prevent the onset of Type 1 diabetes. While details were not reported, the authors made a

general statement that these reported changes most often reportedly occurred in the areas

of diet and increased exercise. The authors also found that a higher level of anxiety was

associated with greater lifestyle/behavior modifications. Similarly, in a later study of

genetically at risk infants, mothers who continued their child's participation in the

longitudinal PANDA study tended to be more anxious with infants at higher risk.

Mothers who believed their at risk children would never get diabetes were less likely to

continue study participation (Carmichael et al., 1999b).

In a recent study of intentions for behavior change, Hendrieckx et al. (2002)

surveyed a sample of 403 adults with first-degree relatives with type 1 diabetes who were

undergoing antibody screening for type 1 diabetes and were assessed prior to results

notification. This novel study sought to better understand the relationships between

perceived control, distress, and behavioral intentions. Results indicated 73% of

participants stated they intended to make a lifestyle change if found to be at high risk,

with diet (87%) and exercise (30%) most frequently endorsed (Hendrieckx et al., 2002).

These results suggested individuals' beliefs regarding the prevention of type 1 diabetes

did not correspond well with current scientific knowledge; however, beliefs appeared

more congruent with an understanding of type 2 diabetes (Hendrieckx et al., 2002).

Hendrieckx et al. (2002) found general anxiety did not appear to be a significant predictor

of behavior change, nor were behavioral intentions predicted by education level.

However, similar to Johnson & Tercyak (1995), diabetes-specific worry was related to

intentions towards behavioral change, along with perceived internal control. Hendrieckx

et al. (2002) also found those who were female, married, and older were more likely to








report anticipating making lifestyle changes. Additionally, perceived internal control was

related to beliefs regarding the causes of diabetes. More specifically, those who believed

their relative developed diabetes largely due to heredity or chance, were more likely to

believe they were unable to do something to reduce their risk of developing diabetes.

While this study provided important exploratory data, the results were limited because

the data were collected prior to the screening results, and intentions --rather than actual

behaviors-- were assessed. Additionally, it used several new measures, which have not

been psychometrically validated.

Additional data on diabetes screening and behavioral change come from the

Participant Experience Survey, designed by Johnson, for participants who completed the

DPT-1 study (Johnson, 2002). The survey was administered anonymously across study

sites to examine subjective experiences of participants (who were at least 10 years old) in

the trial, as well their parents (of participants under the age of 18). Questions assessed a

broad range of issues, including study adherence, satisfaction, reasons for participation,

perceived need for psychological support, and efforts to prevent or delay type 1 diabetes

from developing. Items that assessed efforts to prevent/delay type 1 diabetes were

designed to reflect intentional changes in weight diet, exercise, lifestyle, stress level,

monitoring, and alternative medication use. Items were scored as either "yes" or "no" to

reflect whether or not an individual reported engaging in a specific behavior. Only data

from those who were unaware of the study's results were analyzed. Sixty-five percent of

DPT-1 participants, who were all over the age of 10, responded to the survey, with 82

from the intervention (IN) group (which received preventative insulin) and 81 from the

close observation (CO) group. Over half (54%) of all participants reported modifying at








least one behavior in an effort to delay or prevent type 1 diabetes onset, with no

significant differences between groups, with the exception of alternative medication use

(significantly greater use in IN group). Results indicated dietary changes were the most

common behaviors reported, with approximately one-third of participants stating they

reduced candy or sweets intake, reduced intake of regular soda, or increased intake of diet

soda. Twenty-eight percent of participants indicated they would increase their physical

activity. Seventeen percent stated they took alternative medications. Ten percent of

participants reported attempting weight loss. Seven participants (4 in the experimental

group and 3 in control) stated they took extra insulin in an effort to delay or prevent

diabetes onset. No significant predictors of behavior change were found in this study.

These results were congruent with Tercyak & Carmichael (1995) and Hendrieckx et al.

(2002) and indicated a substantial proportion of individuals' who are found at risk for

type 1 diabetes engage in behaviors that correspond to those found to be effective in the

treatment and prevention recommendations for type 2 diabetes (ADA, 2002b;

Tuomilehto, et al., 2001). Data are currently available from parents whose children

participated in the DPT-1 study. These data were used as a comparison group in the

current study.

These studies described herein suggest there are unanswered questions to be

explored regarding the behavioral outcomes associated with risk screening for type 1

diabetes. Existing data suggest individuals report engaging in behavior changes in

response to risk information, although it is unclear what predicts these behavioral efforts.

Based on existing literature, risk perception, perceived control, and psychological distress

appear important factors to consider. Behavior changes that result from risk notification






35


may or may not be related to scientifically validated methods of risk reduction. However,

in the case of type 1 diabetes, whether a behavior is scientifically valid is not necessarily

important, since we do not currently know what delays or prevents the onset of the

disease.


















APPRAISAL (risk perception)

*Exposure variables (uncertainty reduction,
prevention and surveillance options)

*Personal factors (social support, perceived
control, information seeking)



No stress Stress and
stress response
I


Figure 2-1. Partial representation of the Stress-Disease Risk-Coping Model adapted from
Baum et al. (1997)


STRESSOR
CHARACTERISTICS
-test results
-uncertainty
-disease characteristics













CHAPTER 3
RATIONALE AND PURPOSES

The purposes of this study were to better understand predictors of self-reported

behavior change in mothers of newborns who were identified as at-risk for type 1

diabetes through genetic screening. Currently, little is understood about the specific

behavior changes that result from knowing one's child is genetically predisposed to a

condition for which there is currently no known prevention method or cure.

Additionally, our present understanding of the etiology of type 1 diabetes suggests it

develops from a combination of both genetic and environmental influences, which are not

well-defined. In the absence of definitive recommendations from the health care

community, mothers of newborns identified as "at-risk" may take actions they believe are

effective in preventing type 1 diabetes in their children.

Based on previous studies, possible behavioral changes may include altering their

children's environment, feeding schedules, activity patterns, and/or medical surveillance

behavior (Hendrieckx, 2002; Johnson, 2001; Johnson, 2002). These efforts may represent

mothers' attempts to reduce their anxiety and better cope with the situation. However,

we have yet to document the nature and extent of such behavior changes, including the

incidence of excessive prevention efforts that may become burdensome and impact daily

functioning. Further, since the onset of diabetes is thought to be an interaction between

genetics and the environment, it is unclear to what extent certain types of behaviors could

advance or delay disease onset. Although current science does not permit us to








recommend certain behaviors as preventative, it is important for us to monitor the role of

relevant behaviors if we are to understand the natural history of this disease.

Monitoring possible behavior change associated with high-risk notification is

equally important to current and future diabetes prevention trials. In the DPT-1, for

example, behavior change efforts taken by the control group (e.g., taking insulin or

nicotinimide) could undermine the trial's internal validity. Unless these behavior

changes are monitored, interpretation of study results can become exceedingly difficult.

This is not unique to diabetes-or genetic screening-specific trials as certain behavior

changes could potentially impact other types of clinical trials as well.

This study involves both qualitative and quantitative data to examine reported

behavioral outcomes associated with participation in the Perspective Assessment of

Newborn Diabetes Autoimmunity (PANDA) study. Findings will be examined in the

context of Baum and colleagues' (1997) model of genetic testing, in which behavior

change in response to genetic test results is influenced by one's risk appraisal, affective

response to the information, and available coping resources. Objectives of the study are

listed below.

Objective 1: To Investigate the Extent of Reported Maternal Behavior Change as a
Result of Genetic Screening for Type 1 Diabetes

Hypothesis 1.1: Reported behavior changes will most likely correspond to

recommendations for the treatment of diabetes (American Diabetes Association (ADA),

2002a, 2002b) and the prevention of type 2 diabetes (Pierce et al., 1995), including

changes in diet and physical activity patterns (Forsyth et al., 1997; Pierce et al., 1995).

Rationale: There is scientific uncertainty regarding the environmental factors

associated with the development of type 1 diabetes. The health care community and








media have recently focused significant attention on type 2 diabetes, advocating for

healthy lifestyle changes. In a study of parents with type 2 diabetes, nearly half thought

they could reduce their children's risk of developing diabetes by altering their children's

diet and exercise patterns (Pierce et al., 1999). Mothers, who may not understand the

distinctions between type 1 and 2 diabetes may apply such recommendations to their

children at-risk for type 1. This hypothesis is congruent with findings from Johnson &

Tercyak (1995), Hendrieckx (2002), and unpublished data from the DPT-1 survey

(Johnson, 2002).

Objective 2: To Assess Predictors of Maternal Behavior Change as a Result of
Genetic Screening for Type 1 Diabetes

Hypothesis 2.1: Mothers who perceive they have control over their child

developing diabetes will be more likely to report engaging in behavior changes.

Rationale: Mothers may be more likely to report taking action to help prevent such

an outcome if they believe they have some control over the situation. Behavioral change

to reduce a health threat is more likely if there is a belief that change can be affected

(Diefenbach et al., 1999; Hendrieckx et al., 2002). Perceived control is related to both

uncertainty reduction and available prevention/surveillance options, which are integral to

risk perception, a key component of health behavior change (Baum et al., 1997).

Therefore, perceived control will be examined in the context of other predictors of

behavior to determine possible interaction effects.

Hypothesis 2.2:_Mothers who perceive their children to be at greater risk will be

more likely to report engaging in behavior change.

Rationale: Perceived risk, more so than infant's actual risk, will be a better

predictor of behavior. Mothers of children who perceive their children to be at higher








risk than has been identified through testing (overestimate their risk) will be more likely

to engage in behavior change. Mother of children who underestimate their child's actual

risk will be less likely to engage in behavior change. Children who are perceived to be at

high or extremely high risk may have mothers who will be more likely to try to intervene.

Studies of genetic screening for breast cancer have found that increased perceived risk

predicts likelihood of engaging in health behavior change and health surveillance

behaviors (e.g., Aiken et al., 1994; Meiser et al., 2000; Ritvo et al., 2002; Schwartz et al.,

1999).

Perceived control may also interact with risk perception as mothers who perceive

their child to be at greater risk may be more likely to engage in behavior change if they

also believe they are able to control whether their child develops diabetes.

Hypothesis 2.3: Mothers who are more anxious will be more likely to report

engaging in behavior change.

Rationale: Mothers who are more concerned and worried about their child

developing diabetes will be more likely to report taking preventative actions. Studies

indicate disease-specific worry (Hendrieckx, 2002; Johnson and Tercyak, 1995) and

beliefs regarding the effectiveness of preventative actions (Diefenbach et al., 1999)

predicted either intentions for behavior change or increased adherence to health-

protective behaviors. Maternal anxiety may also be related to the degree of maternal

perceived control, which may in turn influence behavior. Mothers who are more

anxious/worried may report more behavior change if they also believe they have some

control over the situation.








Hypothesis 2.4: Mothers who use more coping strategies, particularly active

coping (i.e., problem-focused, seeking social support), will be more likely to report

behavioral changes.

Rationale: Engaging in risk reducing behavior can be seen as a means of coping

with a health threat. Behavior change is an active coping approach and likely to be

associated with other ways of coping, particularly those that are also more active, namely

problem-focused coping and seeking social support. Mothers who use avoidant

strategies, and try not to think about the problem, may be less likely to engage in risk-

reducing behaviors.

Additionally, those who perceive greater control over the situation may be more

likely to engage in more proactive coping methods, whereas, mothers who perceive less

control may engage in more avoidant coping and be less likely to report behavior change.

Hypothesis 2.5: Mothers who report information seeking and/or report receiving

recommendations from medical professionals or other family members related to

behavior change will be more likely to report engaging in behavior change.

Rationale: Mothers who are given advice to change their behavior by those they

feel are authoritative will be more likely to follow through with recommendations.

Research findings suggest "monitors" (or information-seekers) are thought to cope more

effectively with stressful situations using more problem-focused, information-obtaining

coping strategies rather than avoiding the situation and not seeking out information

(Scheier et al., 1986; Carver et al., 1989). Participation in research studies may be

viewed as an additional form of information seeking about health status. Information

seeking may also influence one's sense of perceived control, and consequently, may








influence behavior both directly and indirectly. Those who seek and utilize information

from various sources may perceive greater control over the situation, and consequently,

be more likely to report behavior changes. Therefore, information seeking will be

measured in this study and used as an independent predictor, as well as in conjunction

with perceived control.

Hypothesis 2.6: Mothers who continue participation in the PANDA Part II study

(repeated blood testing for antibodies), will be more likely to report other behavior

changes.

Rationale: Health surveillance behaviors, such as participation in additional blood

draws, may be a likely outcome following risk notification. Increased health surveillance

may also signify increased contact with health care professionals. For those who

continue in the PANDA study, the risk of developing diabetes may be more salient and

seen as something they should address. Mothers who continue in the study have contact

with investigators and study staff over time and therefore, this contact may influence their

behavior. Mothers who are sufficiently concerned about their child's risk enough to

monitor their child's risk more closely, may be more likely to report other behavior

changes. Data on participation in PANDA Part II blood draws are available and these

data can be compared with maternal report.

Objective 3: To Assess Psychological Effects (i.e., Anxiety) of Maternal Behavior
Change Over Time

Hypothesis 3.1: Mothers who report modifying behaviors will show a greater

reduction in anxiety over time than mothers who do not report behavior change.

Rationale: Behavior changes, including health surveillance behaviors, may

represent means of coping with a health threat. Engaging in behaviors perceived as risk








reducing may help lower maternal anxiety regarding the situation. However, the

influence of reported behavior change on maternal anxiety may be influenced by

maternal perceived control (i.e., mothers who perceive control over diabetes onset and

engage in behavior change may show greater reduction in anxiety).

Objective 4: To Compare Reported Behavior Change between Mothers of Children
Genetically at Risk for Developing Type 1 Diabetes with Mothers of Children in the
Diabetes Prevention Trial Who Were ICA+, and Therefore, at Even Greater Risk
for Diabetes Onset

Hypothesis 4.1: Mothers of genetically at-risk children will be less likely to

report behavior change than mothers of ICA+ children enrolled in Diabetes Prevention

Trial-1 (DPT-1).

Rationale: Participants enrolled in the DPT-I trial were at increased risk for

diabetes as identified through positive family histories and positive ICA screening. Their

risk level was collectively higher than 98% of our original total sample population for the

PANDA Part III study (of whom 7 of 435 were ICA positive). For this reason, we

believe that mothers in the current study will report fewer behavior changes than mothers

of children in the DPT-I since children of mothers in the proposed study are at relatively

less risk than the DPT-lstudy children.

Examining this hypothesis will allow us the unique opportunity to explore

differences between maternal reports of behavior change in two at-risk groups: children

identified at birth as genetically at-risk and higher-risk children who have entered a

prevention trial.

Pierce et al's (1999) study of parents with type 2 diabetes found that those who

believed they could prevent their children from developing diabetes and who perceived





44


their child's risk to be higher, were more likely to experience greater anxiety (Pierce et

al., 1999)

For the purposes of examining this hypothesis, maternal data from the DPT-1

survey (n = 134) will be compared with mothers' reported from the PANDA study.

Based on DPT-1 participant data, over half of the sample reported at least one behavior

change, with dietary changes most often reported, followed by increased exercise, weight

loss attempts, and alternative medicine use (i.e., vitamins) use (Johnson, 2002).













CHAPTER 4
METHODS AND MATERIALS

Prospective Assessment of Newborn Diabetes Autoimmunity (PANDA) Study
Procedures

Part I

Participants were mothers whose infants were screened at birth to determine their

genetic risk for the development of Type 1 diabetes (1997-1999) through Part I of the

Prospective Assessment of Newborn Diabetes Autoimmunity (PANDA) study. This

study is a National Institutes of Health and Juvenile Diabetes Research Foundation

Internal-supported registry that uses genetic testing to identify newborns at risk for type 1

diabetes (Schatz et al., 2000). In this study, mothers were contacted at the time of their

child's birth and asked permission to screen the newborn for the presence of the high-risk

HLA-DQB1 alleles using blood spots on filter paper (obtained by heel stick at the time of

state-mandated phenylketonuria testing). Informed consent was obtained and consenting

participants were told they would only be re-contacted if their child was at increased risk

for type 1 diabetes. The majority of these women gave birth at participating locations in

Gainesville, Florida, or Pensacola, Florida, were all English speaking, and were over the

age of 18.

PANDA genetic testing results placed infants into one of six risk categories: very

low risk (1/6000), low risk (1/300), slightly increased risk (1/125), moderate risk (2/100)

high risk (5-10/100) and extremely high risk (20-25/100) (Table 4-1). Only children who








were at moderate, high or extremely risk children are followed longitudinally in the

PANDA study.

If a child was determined to be "at risk," in other words, classified as either at

"moderate," "high" or "extremely high risk," mothers were sent letters notifying them

that their children's genetic test results were available. Results were usually available

after approximately 12-20 weeks following birth. In the letters, participants were

requested to call the PANDA study coordinator to discuss the results and possible

continued study participation in PANDA Part II according to PANDA protocol. If no

response was received approximately 30 days after sending the notification letter, the

PANDA study staff attempted to contact the mother by phone to notify her of her infant's

risk.

For mothers who called for results or were contacted by phone, the study

coordinator followed a scripted presentation of the risk information, including both

categorical and numerical risk figures. Additionally, she presented available options,

including participation in the PANDA Part II study, and an opportunity to ask questions

about the study or the meaning of the results. Parents had the option at that time to

decline further participation, continue with the study, or delay their decision. Regardless

of participation status in the PANDA Part II study, all mothers were asked for their

permission to be contacted by a second individual from the Pediatric Psychology

Research lab who would ask them questions about their understanding of the study and

its psychological impact. See Figure 4.1 for procedural outline of entire PANDA study.

Part II

Part II of the PANDA study involves longitudinal follow-up of children screened at

birth. These children are periodically screened (via blood draws) starting when the child








is at least six months of age for the presence of autoantibodies, which are additional

markers of diabetes disease progression. A positive screening for autoantibodies would

suggest that the child is at even greater risk of developing type 1 diabetes. Blood draws

could be conducted either by (1) mailing out supplies to parents and to have their

pediatricians draw the blood and mail back to the PANDA staff or (2) scheduling directly

at study sites in Gainesville, Orlando, or Pensacola, Florida. Blood draws were expected

to occur every at 3, 6 ,or 12 months, depending on risk level.

Part III

Part II of the PANDA study examined the psychological impact of participation in

PANDA, including maternal affective (i.e., anxiety) and cognitive responses (i.e., risk

understanding) as well as coping response. Mothers who agreed to be contacted at the

time of notification were interviewed by telephone approximately 4 weeks following

notification (M = 3.50, SD = 1.96) and again at 4 (M = 3.93, SD = 1.96) and 12 months

(M = 12.83, SD = 2.45) post-notification (see Figure 4-1).

For the initial interview, the Part III participation rate was high. Approximately

90% (n = 435) of the mothers we were able to contact (of over 700 eligible) agreed to

complete the initial interview, 79% participated in a second interview (n = 344), and 62%

participated in the third interview (n = 269). Sixty percent (n = 262) completed all three

interviews. Of those who did not complete all three interviews, 67 declined to be

contacted beyond the first interview (no attempts were made to contact these mothers to

participate in the current interview), and 106 were unable to be contacted by phone due to

either disconnected numbers or the time that had elapsed between or study personnel

could not reach them.








Participants

To be eligible for the current study, mothers must have completed at least the initial

interview of the PANDA Part III study and at no point declined participation in either of

the subsequent two interviews (n = 368). Out of 368 eligible mothers for the current

study (i.e., those with > 1 previous interview and who did not previously decline

participation), 204 were successfully contacted (55%). Of these mothers, 192 (94%)

completed the interview, ten declined participation (5%), and two mothers were no longer

eligible because their "at risk" children had recently developed type 1 diabetes (1%). Of

the 163 mothers who could not be contacted (44%), 145 had disconnected numbers

and/or had no forwarding contact information, and 18 mothers with presumably correct

up-to-date contact information were "unable to be contacted" after multiple attempts.

Families were deemed "unable to be contacted" when there was no response after at least

fifteen attempts were made over a two-month period, with at least three messages left if a

family member or answering machine was available.

Maternal Characteristics

Mothers who completed the current interview ranged in age from 20 to 46 (M =

33.67, SD = 5.38). Eighty five percent of mothers were married and 44% had a 4-year

college degree at the time of interview (Table 4-2). Eighty-five percent of mothers where

Caucasian and therefore, minority members were under-represented compared to the

population in Alachua County, Florida, and Florida in its entirety, where among women

of childbirth age (18-44 years) approximately 25% and 23% are minorities, respectively

(Florida Office of Economic and Demographic Research, 2001). On average, mothers in

this sample had two children at the time of interview (M = 2.09, SD = 1.11). Eighty three








percent of participating mothers completed all three interviews and 62% attended at least

one blood draw.

Child Characteristics

Target at risk children of participating mothers were between the ages of 2 and 7

years (M = 4.25, SD = 0.89) and evenly split between males and females (Table 4-3).

Within this at risk sample, the majority of infants were at "moderate" risk (56%), 37%

were at "high" risk, and 7% were at "extremely high" risk. Five out of the six eligible

mothers of children who were antibody positive participated in the current interview.

Most children of participating mothers were reported as having a family history of

diabetes (72%) (type 1 or 2). Sixty-five percent of children have at least one distant (>

second degree) relative with diabetes. Thirty-seven children (19%) have at least one

first-degree family member with diabetes. Of these, 30 children have immediate family

members with type 1 diabetes (81%), including 15 participating mothers themselves,

along with seven fathers and 14 siblings. In five of these families, two immediate family

members have type 1 diabetes.

Procedures

The current interview was conducted at least one-year post PANDA Part III study

completion and therefore, two to four years post-notification (M = 3.60, SD = 0.78).

Attempts were made by the Principal Investigator or research staff to contact all eligible

mothers (n = 368) for an additional follow-up interview to measure reported maternal

behavior changes resulting from knowledge of their children's risk for type 1 diabetes.

Contact information for these mothers was kept within a computerized database with

restricted access, so telephone numbers were available only to study staff.








When participants were contacted, they were reminded of their earlier

participation in PANDA Part III interviews and asked if they would agree to participate

in an additional interview. Participants were reminded of the voluntary and confidential

nature of the study and those who agreed to participate, were given a $5 gift certificate to

Publix or Target (their choice) as a token of appreciation.

Asking mothers about their behavior might have had the potential to raise

mothers' anxiety and curiosity levels regarding what they should or should not be doing

to help their children. Therefore, at the beginning and end of the interview, there was a

disclaimer read to remind mothers that we do not currently know what causes type 1

diabetes, and that we did not have specific recommendations to offer other than

encouraging a health lifestyle, including a healthy diet, physical activity, and rest. For

mothers who asked more specific questions beyond this, we had prepared documents

from the American Medical Association (AMA) on developmentally appropriate

guidelines regarding eating, exercise, and sleep (found at www.ama.org; last accessed

6/1/02), which could be mailed to mothers upon request. Eleven mothers requested

additional information. The most frequently requested materials pertained to information

regarding signs and symptoms of type 1 diabetes.

For quality assurance purposes, data were entered twice into a computerized

database, systematically compared and cleaned, before analyses were conducted. Data

from this interview were linked to previously collected data (PANDA Part III) on these

study participants through their unique identification numbers assigned by the PANDA

study staff. This allowed for longitudinal analyses of the data. This study was approved

by the UF Health Science Center Institutional Review Board (9/1/02) and documentation








of written consent was waived. Funding for this study was obtained from the North

Central Florida's Children's Miracle Network.

Measures

Descriptive Variables

Descriptive data were collected to examine the maternal and child demographic

characteristics, overall participation rate as well as demographic differences between

mothers who agreed to participate versus those who declined or were unable to be

contacted (for further detail see section on "Predictor Variables"). These two groups of

mothers were compared across outcome and predictor variables based on data from the

initial interview.

Outcome Variable: Reported Behavior Change

A component of the structured interview was developed to assess behavioral

changes across six domains: (1) diet/eating patterns, (2) physical activity, (3) emotional

stress, (4) medical interventions, (5) medical surveillance, and (6) illness prevention

behaviors. These questions and constructs were adapted from the Participant Experiences

Survey used in the Diabetes Prevention Trial-1 (DPT-1) study (Johnson, 2002) and

constructs were classified based on the DPT-1 survey and Hendrieckx et al. (2002).

Additional questions were added to address other potential environmental triggers or

influences hypothesized in the research literature to be related to diabetes development

(Akerblom, et al, 1998) (see Appendix A).

Diet and eating patterns (a = .58). Sixteen questions addressed changes in the

frequency, amount, and types of food/drink (i.e., sweets, soda, juice, cow's milk) given to

the child as well as attempts to modify the child's weight. Also, included were questions

assessing changes in early feeding history, including timing of the introduction of solid








foods and breastfeeding. The 16 questions represented ten different types of behavior

changes, as some questions are paired to assess in which the direction changes occurred

(i.e., decrease vs. increase). Additionally, two questions referred to behaviors for which

the concepts of frequency and duration do not apply, and therefore, these detailed follow-

up questions were not asked.

Physical activity/Physical stress (a = 0.54). Four questions assessed whether

mothers increased or decreased their children's physical activity or physical exertion in

response to their child's risk for type 1 diabetes.

Emotional stress (q = 0.47). Four items were designed to assess lifestyle changes

that foster the reduction of the child's level of emotional stress.

Medications (a = 0.54). Five items addressed whether mothers provided their

children with medications, such as dietary supplements, vitamins, or insulin.

Illness Prevention (a = 0.72). Eight questions representing seven unique concepts,

assessed the degree to which mothers altered their children's environment to minimize

risk of illness or infection.

Medical surveillance (a = 0.37). Five questions were designed to assess whether

participants engaged in health- monitoring behaviors for their children, such as more

frequent doctor's visits, glucose monitoring, and reported participation in PANDA Part II

study (autoantibody screening). However, reported PANDA Part II participation was not

included in calculating the domain score, as it was used for reliability purposes and also

used as a predictor variable.

This portion of the interview began with a simple "yes/no" question assessing if,

in general, participants felt they engaged in any behavior change to prevent diabetes in








their child. This question was followed by more detailed questions regarding different

types of behaviors relevant to the six domains described above. For each section,

participants were first given an open-ended question to solicit spontaneous answers (e.g.,

Have you done anything different with your son's physical activity patterns to prevent

him from developing diabetes?) followed by more detailed forced-choice questions.

When a response was given to an open-ended question that would be later addressed by a

forced choice item, the corresponding forced choice item was also endorsed.

Within each domain, forced choice items were designed to assess a wide variety

of behavior changes. For each question, participants were reminded that these questions

apply only to behaviors initiated specifically to prevent diabetes in their children. For

forced choice questions in which the response was "yes," follow-up questions were asked

to assess duration/consistency and frequency of given behavior. Forced choice items

were scored as either "yes" or "no." Duration or consistency of the behavior was scored

as "never" (0), if the behavior never occurred, "inconsistent" (1), if the behavior was

initiated early on but stopped, or began only recently; or "consistent" (2), if the behavior

has been ongoing since results notification. Frequency of a behavior was scored as

"never" (0), "occasionally" (1) or "always/nearly everyday" (2).

Each question was scored as dichotomous "yes/no" (0 or 1) as well as given a

continuous composite score value for duration, frequency, and duration x frequency.

However, duration was relatively static as 86% of those endorsing a certain behavior

reported consistent engagement since time of notification. As for frequency of behaviors,

60% reported engaging in the behavior "always/nearly everyday" (Table 5-8).








Due to relatively low frequencies for most items and low variability in duration

scores, only the dichotomized "yes/no" scores were used for analyses. Domain scores

were calculated in two ways: (1) calculating sum of the number of behaviors endorsed

and (2) whether at least one behavior change occurred with each domain. A total score

for behavior change was similarly obtained by collapsing domains. A factor analysis of

this measure was not conducted due to low variability on the items and inadequate

sample size for the number of items in the measure. To determine the statistical strength

of the scores for the six domains and total score, coefficient alphas were calculated to

determine the reliability of each construct (Table 4-4).

Reliability was relatively strong for the total behavior score (a = 0.77) and illness

prevention domain (a = 0.72), but weaker for the other five domains, with alphas ranging

from 0.37 to 0.58. Correlations between domain scores ranged from 0.10 to 0.44 (Table

4-4). As expected given the data, the total behavior score was best correlated with diet

and health surveillance behavior scores. Due to the low frequency of behavior changes

within several domains, as well as the non-normal distribution and relatively poor

reliability of domain scores (a < 0.60), no further analyses of domain-specific behaviors

were conducted (Table 4-5). Additionally, due to the relative low frequency of

endorsement of items overall and the non-normal distribution of the total behavior score,

the total behavior score used in subsequent analyses was the dichotomous variable of

whether at least one behavior change was reported (1 = 'yes') versus no behavior change

reported (0 = 'no') (see Figure 4-2).

Reliability of self-reported behavior change. Self-reported participation in Part II

of the PANDA study was collected in the structured interview and compared with data on








actual Part II participation. These data were available through the PANDA computerized

database. Continuation in the PANDA Part II study was defined as those mothers who

brought their child in for at least one blood draw for autoantibody screening, coded as

"participated" (1) and "did not participate" (0). Actual participation in PANDA II blood

draws was the only observed behavioral data available to us. It permitted us to examine

the validity of maternal self-report data concerning this particular component of medical

surveillance. PANDA Part II participation data indicated that 61% of mothers

participated in at least one subsequent blood draw and 26% participated in two or more.

When asked in the interview, 174 mothers reported accurately whether they participated

in Part II of the study (91%), with 72 accurately reporting they had not continued

participation and 102 correctly reported they had. Three mothers reported participating

when they actually had not (1%), and 15 reported they had not participated when they

actually had (8%). These findings suggest mothers may have been open and honest when

completing the interview and that social desirability effects were not strong. If anything

it is possible mothers may have underreported efforts to prevent diabetes.

Predictor Variables

Sociodemographics

The following variables were assessed during the first PANDA Part II telephone

interview: date of interview (to calculate length of time since notification), maternal date

of birth, child date of birth, maternal and paternal education level, family income bracket,

maternal and child ethnicity, marital status, number of children and whether or not this is

her first child. The number of first-degree relatives, second-degree relatives, or greater

relatives of the child with type 1 or type 2 diabetes was also assessed, if known. In the

current interview, several non-static demographic variables were updated in this current








interview to ensure that information was current, including marital status, number of

children, family income bracket, maternal and paternal education level, and family

diabetes history. Mothers were also asked for a current address in order for gift

certificates to be sent.

Perceived control

This construct was assessed by a series of questions adapted from a questionnaire

developed by Bradley et al. (1999) and used in Hendrieckx et al. (2002). These questions

assessed whether participants believed there was anything a parent or a medical

professional could do to prevent diabetes in the children, as well as a question about

diabetes onset being determined by chance or fate. Responses were scored on a 5-point

Likert scale, anchored by "strongly disagree" (scored as 1) and "strongly agree (scored as

5)." Internal consistency of this 3-item scale was a = 0.55. This was unsatisfactory

based on the study's criteria of using an alpha score of 0.60 as the cut-off for acceptable

reliability. However, when chance was not retained as a part of this composite, internal

consistency increased ( = 0.66), therefore, only the two-item measure of perceived

control was retained as a composite measure (Table 4-7). The composite score of

perceived control was calculated by averaging the scores of the two items (Table 4-6).

Risk perception

(1) Perceived absolute risk. An absolute measure of perceived risk and its

accuracy was assessed in the previous interviews and was assessed in a similar way in the

current interview. Mothers were presented with a list of the possible risk categories (with

numerical estimates) and asked whether or not any of these were the risk group they were

told their child was in. "I don't remember" was recorded if they were unable to recall or

recognize their child's risk category or number. (2) Perceived estimated risk. Perceived








risk was considered accurate if the participant was able to recognize the infant's correct

risk status from the list. Responses were classified as "accurate" (scored 2),

"overestimates" (scored 3), "underestimates" (scored 0), or "unknown" (scored 1) based

on the relationship of the response to the child's actual risk status This component

reflected perceived absolute risk while controlling for actual risk and was included in the

composite score of risk perception, whereas, absolute risk was not. (3) Perceived

comparative risk. A question adapted from Hendrieckx et al. (2002) assessing perceived

comparative risk was included. The question was stated as follows: "How do you think

your child's risk for developing diabetes compares to other children?" The response was

rated on a 5-point Likert scale ranging from 1 to 5, anchored by "much lower" and "much

higher." (4) Expectations. A question used in all previous interviews assessed whether

participants believed their child will develop diabetes. This question was coded as "yes,

my child will develop diabetes in the near future" (scored 3), "my child will eventually

develop diabetes but not for a long from now," (scored 2) "my child will not ever develop

diabetes," (scored 0) or "I am unsure." (scored 1) This variable was previously used in

Carmichael et al. (1999).

Intercorrelations between the three risk perception variables (estimate risk,

relative risk, and expectations) were examined and a composite score was calculated. To

accomplish this, scores were transformed into z-scores and mean of the three variables

was derived as the composite. Additionally, reliability of the composite score was

assessed (.= 0.61; Table 4-7). For those whose response to an item was "unknown,"

when calculating reliability for the composite score, their score was replaced by the item








mean. When computing the individual's composite score in cases with an "unknown"

response, the individual's score was the average of the other two risk items.

Anxiety

Anxiety was measured by a 10-item short form of the state-component of the

State Trait Anxiety Inventory STAI (STAI; Spielberger, 1970). Respondents were asked

to rate the questions according to how anxious they presently felt about their child's risk

for developing type 1 diabetes on a four-point scale (i.e., Not at all, Somewhat,

Moderately, or Very much). The 10-item STAI was also administered at all previous

interviews and results were reported on in published studies (Johnson et al., submitted;

Carmichael et al., 2003).

The 10-item short form was derived from a sample of 231 mothers who

completed the full 20-item scale at the initial interview. Ten items were selected by

examining the items that most highly correlated with the full 20-item scale scores for

these participants. This form was found to be highly reliable at the initial (a = 0.93), four

month follow-up (a = 0.92) and 12 month (a = 0.90) follow-up interviews. The 10-item

short and 20-item full forms of the STAI were highly correlated (r = 0.97). The practice

of creating a short form of this measure is not unusual. The STAI-SF, a six-item Short

Form, was developed and used in a prior study related to genetic screening (Marteau &

Bekker, 1992).

A regression equation was developed which converts the short form scores into

scores compatible with STAI norms to allow for comparisons with normative data

provided in the STAI Manual. Data compiled by Carmichael et al. (2000) provides

additional comparisons to similar samples including mothers learning of their child's








increased risk status as a result of ICA testing and pregnant women undergoing

amniocentesis.

An additional question, adapted from Hendrieckx (2001) was asked to assess how

often mothers worry about their children's risk. This question was stated as "how often

do you worry about your child's risk for developing diabetes?" and rated on a 5-point

Likert scale ranging from 0 to 4 anchored by "never" and "very often".

A composite score was derived by converting both scores into z-scores and

calculating the mean z-score of the two items (a = 0.80) (Table 4-8).

Coping

The Ways of Coping Checklist-Revised (WCC-R) (Folkman & Lazarus, 1980) is a

69-item dichotic (yes/no) questionnaire used to assess the use of coping strategies and

preferred coping style. In the PANDA Part III study, the WCC-R was administered at the

4 follow-up interviews to assess maternal coping regarding their infant's genetic risk of

developing Type 1 diabetes (n = 178). This measure has also been used in similar risk

screening studies (Johnson & Tercyak, 1995; Johnson & Carmichael, 2000). Factor

scores were calculated using Vitaliano et al. (1985) factor structure, which uses 42 items.

The five factors included the following coping styles/strategies: Problem-focused Coping,

Seeking Social Support, Wishful Thinking, Self-Blame, and Avoidance. The WCC-R

was not administrated in the current interview. However, subscale scores obtained at the

second (4 month) interview were used as predictor variables. To be able to compare

across factors having a varying number of items, mean scores were calculated for each

subscale as well as for the total measure.








For this sample reliabilities for the subscales Wishful Thinking (g = 0.70), Seeks

Social Support (g = 0.73) and Problem-Focused coping (g = 0.81) were satisfactory.

However, reliabilities for the Avoidance (g = 0.36) and Self-Blame (a_= 0.53) subscales

were poorer and did not meet criteria for further analyses (a_< 0.60) (Table 4-9). The

reliability scores of the factors were consistent with previously published studies of

similar populations (Johnson & Carmichael, 2000). Correlations between variables were

significant, particularly between the total coping score and Problem Focused coping,

Seeks Social Support, and Wishful Thinking (Table 4-9). Mean factor scores were

similar to previous studies. Seeks social support was the most favored used coping style,

followed by Problem-Focused coping. Self-blame was the least used coping style (Table

4-10).

Information seeking

A self-report measure of information-seeking was given to assess participants'

sources of information regarding diabetes risk and/or behavior change. Questions were

designed to assess if participants consulted with their physicians, family members of

friends, including those who may have diabetes themselves. Follow-up questions were

asked to determine if participants were given specific advice from these sources and if

they followed the advice. Additional questions assessed behaviors such as searching the

internet, consulting written materials about diabetes, or watching diabetes-related

television news stories. This measure was scored as a continuous variable by calculating

the number of information sources reported and as a dichotomous variable denoting

whether any information seeking occurred (1 = 'Yes' and 0 = 'No') (Table 4-11).








Questions regarding the nature of the relationship and the content of the advice were used

as descriptive data.

Participation in PANDA Part II

Data on Part II participation was available through the PANDA computerized

database. Continuation in the PANDA Part II study was coded as two different variables,

(1) number of blood draws and (2) at least one blood draw (1 = 'participated') versus no

blood draws (0 = 'did not participate'). For more details, please refer to earlier section

entitled "Reliability of self-reported behavioral change".

Statistical Analyses

Data analyses were conducted using SPSS 11.0. Internal reliability of predictor

and outcome scores were calculated using Cronbach alpha and only constructs with

alphas greater than 0.60 were retained for regression analyses. Additionally, components

of risk perception and anxiety composite scores were transformed into z-scores because

they were measured on different numerical scales. Consequently these composite scores

reflect a z-transformation as well. Descriptive statistics were conducted, including

ANOVA, t-tests and chi-square analyses, to compare demographic variables between

participants and non-participants. When expected cell size was < 5, Fisher's Exact test

statistic was used instead of chi-square. Hierarchical logistic regressions were used to

predict to behavior change, as well as linear regression to predict to the continuous

outcome measure of anxiety. For further details, refer to Chapter 5 (Results).











Initial genetic screening of newborn
for diabetes risk @ birth

Scripted notification of results
to mother by telephone"
-12 weeks after birth

Initial structured telephone interview
with mother of at-risk infant to assess maternal distress
~4 weeks after notification

Second structured telephone interview
with mother of at-risk infant to assess maternal distress
-4 months after notification

Opportunity to participate in PANDA Part II Blood Draw
when infant is at least 6 months old


Third structured telephone interview
with mother of at-risk infant
to assess maternal distress-12 months after notification


CURRENT STUDY
Fourth structured telephone interview
to assess maternal behavior change
4 years post-notification


S PART I:
PANDAa
STAFF




PART III:
PPRb STAFF





PART H:
PANDA STAFF


PART III:
PPR STAFF



CURRENT:
PPR STAFF


Figure 4-1. Procedural outline of PANDA study. PANDA = Prospective Assessment of
Newborn Diabetes Autoimmunity. PPR = Pediatric Psychology Research
Lab.





'f


.00 2.00
1.00


4.00


5.00


6.00


7.00


C.


10.00


15.00


total behavior score
Figure 4-2. Frequency distribution of total behavior score (n = 192)


8.00 13.00








Table 4-1. Diabetes genetic risk factors
With first


DR/DQ
Alleles/genotypes
DR 3/4, DR 4/4
DQ 0201/0300
DQ 0300/0300


DR 3/4, DR 4/Xa
DQ 0201/0201
DQ 0300/Xb


DR 3/4 or X/X
DQ X/X


DR 3/4 or X/X
DQ 0602


degree
relative
20-25/100
Extremely
high risk


10/100
High risk


1/125
Intermediate
risk


1/15,000
Protective


Without first


degree
relative
5/100
High risk


2/100
Moderate
risk


1/600
Very low
risk


1/15,000
Protective


X is a non-defined allele, 75% of the time X= DR 4 or DQ 0301. X allele is not 0602.


% General
population
5%


10%


85%


% Type 1
patients
>50%


30-40%


10%








Table 4-2. Maternal characteristics of current sample (n = 192)
Variable


Maternal age at notification
Current maternal age


Race
Caucasian
African American
Hispanic
Asian/other


Mothers level of education
High school or less
Some college/trade school
College degree or beyond

Marital status (married)

Annual income (in $10,000 intervals)


30.49 + 5.36
33.67 + 5.38


162 (85%)
6 (3%)
13 (6%)
10 (5%)


45 (23%)
62 (32%)
85 (44%)

164 (85%)

4.97 + 2.50

2.09+ 1.11

2.79 + 0.51
9 (5%)
23 (12%)
160 (83%)

1.19+1.17
74 (39%)
47 (25%)
29 (15%)
41(21%)


Number of children


Number of previous interviews
1
2
3

Number of blood draws (Part II)
0
1
2
3


Note: Data are n (%) and means + SD.









Table 4-3. Child characteristics of current sample (n = 192)
Variable

Infant risk classification
Moderate (2/100) 108 (56%)
High (1/10) 71(37%)
Very high (1/5) 13 (7%)

Child age at notification (mo.) 7.85 + 6.24

Current child age (years) 4.25 + 0.89

Child sex (Male) 97 (51%)

Only child (Yes) 62 (33%)

Family history
No family history 50 (26%)
Third degree relative 98 (51%)
Second degree relative 82 (43%)
First degree relative 37 (19%)

Note: Data are n (%) and means + SD.





67


Table 4-4. Intercorrelations and coefficient alphas for domain scores of reported
behaviors
Domain 1 2 3 4 5 6 7
1. Health surveillance 0.37
2. Diet 0.44* 0.58
3. Physical activity 0.27** 0.32** 0.54
4. Illness prevention 0.35** 0.24** 0.36** 0.72
5. Medications 0.28** 0.28** 0.12 0.36** 0.54
5. Stress 0.16* 0.18* 0.10 0.28** 0.22** 0.47
6. Total 0.74** 0.79** 0.58** 0.65** 0.47** 0.35** 0.77

Note: Coefficient alphas are presented in boldface along the diagonal. p < 0.05 p <
0.01.





68


Table 4-5. Mean domain scores of reported behavior changes for total sample
Domain # Items Range M SD
Health surveillance 4 0-3 0.85 0.86
Diet 16 0-6 0.69 1.15
Physical activity 4 0-3 0.21 0.57
Illness prevention 8 0-5 0.18 0.69
Medications 5 0-3 0.04 0.27
Stress 4 0-2 0.04 0.24
Total 41 0-15 2.00 2.53
# of domains 6 1-6 1.22 1.20








Table 4-6. Mean values or frequencies for perceived control scales
Item


I can do something
Strongly disagree
Somewhat disagree
Neutral
Somewhat agree
Strongly agree

Doctors can do somethinga
Strongly disagree
Somewhat disagree
Neutral
Somewhat agree
Strongly agree

It is up to chance, b
Strongly disagree
Somewhat disagree
Neutral
Somewhat agree
Strongly agree

Composite score (z-score)


2.95 + 1.18
28 (15%)
39 (20%)
53 (28%)
58 (30%)
14 (7%)

2.63 + 1.11
28 (15%)
74 (39%)
38 (20%)
47 (23%)
8 (4%)

3.26+ 1.11
15 (8%)
37 (19%)
39 (20%)
83 (24%)
17 (9%)

2.79 + 0.99


Note: Data are n (%) and means + SD. a Scored as follows: Strongly Disagree = '1',
Disagree = '2', Neutral = '3', Agree = '4', Strongly Agree = '5'. b Variable not used
in composite score.








Table 4-7. Mean values or frequencies for perceived risk scales
Item
Relative risk 3.46 + 1.05
Much less 13 (7%)
Somewhat less 15 (8%)
About the same 61 (32%)
Somewhat higher 74 (39%)
Much higher 27 (14%)

Belief about when child may develop diabetes 1.88 + 0.68
Never 53 (28%)
Unsure 113 (59%)
Yes, but not for a long time from now 22 (12%)
Yes, in the near future 4 (2%)

Risk estimation
Overestimate 12 (6%)
Accurate 76 (40%)
Underestimate 80 (42%)
Don't know/don't remember 24 (13%)

Risk composite (z-score) -.004 + 0.74

Note: Data are n (%) and means + SD. a Scored as follows: Much less = '1', Somewhat
less = '2', About the same= '3', Somewhat higher = '4', Much higher = '5'.
Scored as follows: Never = '0', Unsure = '2', Yes, but not for a long time from now
= '3', Yes, in the near future = '4'. c Overestimate = '3', Accurate = '2',
Underestimate = '1', Don't know/don't remember- '0' but value not included in
analyses.









Table 4-8. Mean values or frequencies for anxiety/worry scale


Item
Worry
Never
Rarely
Sometimes
Often
Always


1.00 + 1.02
73 (38%)
66 (34%)
39 (20%)
8 (4%)
6 (3%)


Anxiety (10-item STAI)b

Anxiety composite (z-score)


30.79 + 9.66


0 + 0.91


Note: Data are n (%) and means + SD.a Scored as follows: Never = '0', Rarely = '1',
Sometimes = '2', Often = '3', Always = '4'. b Predicted full scale score based on 10
item measure.









Table 4-9. Intercorrelations and coefficient alphas for coping variables
Domain 1 2 3 4 5 6 7


1. Problem focused

2. Seeks social support

3. Avoidance

4. Wishful thinking
5. Self-blame
6. Total score


0.81
0.64*

0.15

0.53*
0.22*

0.84*


0.73
-0.02

0.31'"
0.12

0.78**


0.36
0.37*

0.33**
0.37


0.70
0.28** 0.53
0.74*** 0.44*** 0.86


Note: Coefficient alphas are presented in boldface along the diagonal. p< .01. p<
0.001.






73


Table 4-10. Mean scores for Ways of Coping-Revised (WCC-R) scales
Constructa # Items

Problem focused 15 0.48 + 0.24
Seeks social support 6 0.54 + 0.31
Avoidance 6 0.13 + 0.11
Wishful thinking 10 0.25 + 0.23
Self-blame 3 0.03 + 0.13

Total score 42 0.29 + 0.14

Note: Data are means + SD. a Scored as the mean of the items in each subscale. Each
item in subscale is scored as '0' = No, '1' = Yes.





74


Table 4-11. Mean values or frequencies for information seeking scale
Item
Any information source 115 (60%)

Literature 63 (33)
Doctor 46 (24)
Family/friend 33 (17)
Television 27 (14)
Internet 22 (12)

# of information sources 0.99 + 1.04

Note: Data are n (%) and means + SD. aScored as Yes= '1', No = '0'.













CHAPTER 5
RESULTS

Sample Characteristics

Compared to mothers who were eligible but did not complete the current interview

(n = 176), participants in this current study (n = 192) were significantly more likely to be

married (p < 0.001) and older at time of notification (p < 0.01) and the current interview

(p < 0.001) (for those who were not contacted, age was estimated based on end date of

data collection 4/1/03) (Table 5-1). Additionally, they had higher levels of education (p <

0.01) and annual family income (p < 0.001). There were no differences between the two

groups in terms of ethnicity. Overall, these results suggest the current sample was a

highly select sample of mothers who were more economically stable and possessed more

personal resources than mothers in the original larger sample. It is important to consider

the sample bias in interpreting results of this study, as the behaviors of these mothers may

not be reflective of the general population.

For mothers who participated in the current interview, 85% completed all three

previous PANDA Part II interviews versus 57% of eligible non-participating mothers (x

(1, N = 368) = 35.12, < 0.001). Participating mothers had a higher number of

completed interviews (p < 0.001). Participation rates differed significantly for the

longitudinal component of the PANDA study (Part II), which involves periodic

autoantibody screening. Sixty one percent of current study participants and 49% of non-

participants completed at least one autoantibody screening (X2 (1, N = 368) = 5.02 p <

0.05). Participating mothers had a higher number of blood draws (p < 0.001). These








participation rates are less than rates reported in Finland, where approximately 80% of

infants who were genetically screened joined their antibody surveillance study (Kupila et

al., 2001).

As assessed in the initial interview, there was no significant difference in anxiety

scores, as measured by the state STAI, between participating and non-participating

mothers (Table 5-1). There was no significant difference between the two groups of

mothers in their perceived likelihood that their child would develop diabetes in the future.

However, at the time of the initial interview, mothers who participated in the current

interview reported greater accuracy in estimating their child's risk status than mothers

who did not participate (p < 0.01) and fewer mothers underestimated their child's risk (p

< 0.05) (Table 5-1).

There were no differences between children of participants versus non-participants

in regards to age, genetic risk status, sex, only child status, or family history (Table 5-2).

Objective 1

Hypothesis 1.1

At the outset of the study, it was hypothesized that reported behavior changes

endorsed would most likely correspond to recommendations for the treatment of diabetes

(ADA, 2001) and the prevention of type 2 diabetes (ADA, 2002a, 2002b; Pierce et al.,

1995), including changes in diet and physical activity patterns.

The questionnaire's design permitted the use of both open and closed ended

questions within each behavioral domain. Descriptive analyses of each behavioral

construct, both as dichotomous and continuous variables, were conducted, including

frequencies, means, standard deviations, and correlations. Qualitative data from other








open-ended questions addressing advice received and perceived control were also coded

as descriptive data.

Open-Ended Questions

The initial open-ended item assessing behavioral change simply asked whether

mothers did anything special to reduce their child's risk of developing type 1 diabetes

(yes or no). Fifty-five mothers (29%) responded that they had done something

preventative. Open-ended questions were also asked at the beginning of each of the six

domains and again, at the end of the interview to assess maternal recall of behavior

changes. At least one spontaneous behavior was reported in response to domain specific

open ended questions by sixty nine mothers (39%), somewhat more than were identified

through the initial broad question, yielding a total of 118 spontaneous responses (Table 5-

3). Of these, 51 mothers indicated making a change in their child's diet and/or exercise

(74%), corresponding with recommendations to prevent and/or treat type 2 diabetes

(Table 5-4). One domain, medications, yielded no spontaneous responses and stress only

yielded one response. Sixty three percent of the responses to the open-ended questions

were later addressed in forced choice items asked subsequently in each domain.

To further examine the hypothesis that mothers were following recommendations

for the prevention of type 2 diabetes, responses to open-ended questions regarding advice

received and maternal beliefs were analyzed to determine if mothers' reported actions

were based on a premise that a healthy lifestyle is an effective prevention method for type

1 diabetes. An open-ended question assessing maternal beliefs about what they could do

to prevent their child from developing type 1 diabetes was asked of mothers who agreed

or strongly agreed that they could do something to prevent their child from developing

type 1 diabetes (Table 5-5). Seventy two mothers (38%) reported believing they could do








something preventative, with 108 responses generated. Of these, 61 mothers reported

dietary and/or exercise changes (92%).

Additionally, 46 (24%) mothers reported receiving advice from a medical

professional, generating 63 pieces of advice, and 33 (20%) mothers reported receiving

advice from family or friends, generating 39 pieces of advice (Table 5-6). Ninety percent

of mothers reported following advice from a medical professional and 95% reported

following advice from family members/friends. Of the advice received from medical

professionals, 43% suggested making healthier dietary and physical activity changes. Of

the advice received from family members or friends, 28% of advice from family

suggested healthy lifestyle changes in diet and exercise.

Forced Choice Questions

Forced choice items were asked with yes or no responses to assess maternal

recognition of reported behavior changes. These items were used to assess specific

behaviors and were expected to yield more positive responses than the use of open-ended

questions.

Results based on the forced choice items within each domain, indicated that out of

192 mothers, 129 (67%) reported changing at least one behavior in an attempt to prevent

diabetes from developing in their at risk child (M =2.00, SD = 2.53). Domain scores

were calculated for each of the six possible categories of behavior determined a priori.

Of those who reported at least one behavior change, 30% reported two to three changes,

24% reported four to six changes, and 8% reported changing more than six behaviors (M

= 2.98, SD = 2.57) (Table 5-7). Changes in health surveillance behaviors were most

frequently endorsed (59%), including blood glucose monitoring and watching for signs of

diabetes development. Changes in child's diet (34%) were the next most commonly








reported, followed by changes in physical activity (14%), illness prevention (9%),

medications (3%), and stress (3%) (Table 5-8). The item most frequently endorsed (>

10%) was checking for specific signs of type 1 diabetes (50.5%). An open-ended follow-

up question asked mothers to specify the nature of the symptoms they look for in their

children. Ninety seven mothers reported they look for signs of diabetes in their at risk

child, each responding with approximately two signs each (M = 2.16). Seventy nine

percent of mothers reported at least one correct diagnostic criterion type 1 diabetes (i.e.,

polyuria, polydipsia, weight loss, and increased appetite), 32% identified behaviors that

were not indicative of diagnosis, but were related to diabetes symptomatology (i.e., signs

of hyperglycemia or diabetic ketoacidosis), and 45% identified signs that were not related

to diabetes (Table 5-9). Only 5% of mothers did not identify one correct or related

symptom of type 1 diabetes. Of those who reported an accurate symptom, 38% also

listed inaccurate symptoms. Additionally, testing the child's blood glucose level either at

home or at a physician's office, feeding the child less soda, juice and other sweet foods,

and encouraging the child to exercise more often were the next most commonly endorsed

behavior changes. Items that might indicate maternal overprotectiveness or items

suggesting unwarranted use of medications were rarely endorsed. Reported behavior

changes ranged across domains for those endorsing more than one behavior change with

only 19% reporting changes within only one domain (M = 2.33, SD = 1.02). This

suggests that mothers engaged in a wide variety of behavior changes.

In comparing forced choice item responses with responses to open-ended questions,

results indicated that significantly more mothers endorsed forced choice items rather than

made spontaneous responses. This suggests that mothers may either have had difficulty








recalling behaviors that were not as salient, with forced choice items serving as a

recognition task to help refresh their memory. Or, perhaps there may have been a

demand characteristics associated with presenting individual specific behaviors in yes/no

format. According to forced choice items, 67% of mothers reported making at least one

behavior change versus 36% of mothers responding to open ended questions (Table 5-

10). All mothers who spontaneously reported behavior changes also responded similarly

to forced choice items, so there were no mothers who spontaneously reported behavior

change who did not also report changes according to forced choice items. In comparing

forced choice versus open-ended questions, the primary difference was that mothers were

less likely to spontaneously report changes in health surveillance that were later identified

through forced choice items. Mothers may not consider increased health surveillance as a

way of actively preventing diabetes.

Reported behaviors specific to healthy lifestyle changes, within diet and exercise

domains, consistent with recommendations for prevention and treatment of type 2 were

coded and compared to address the hypothesis that behavior change would likely

correspond with recommendations for prevention of type 2 diabetes. Overall, based on

responses to open-ended questions, 51 mothers (27%) reported making at least one such

behavior change (Table 5-4) and 59 mothers (31%) indicated a similar behavior change

via responses to forced choice items (Table 5-8). Behaviors related to recommendations

for prevention of type 2 diabetes were more prevalent among open-ended responses than

among forced choice responses, in which health surveillance changes and overall dietary

changes in general were more frequently reported.








Objective 2

Exploratory model testing was conducted through the use of logistic regression

analyses predicting whether mothers reported behavior change. As stated previously, due

to the non-normal distribution of reported behavior scores, the outcome measure of

behavior was examined dichotomously, comparing mothers who reported at least one

behavior change (1 = > 1 behavior change) versus mothers who reported none (0 = no

behavior change). Regressions were also conducted to predict whether a behavior change

was spontaneously reported in response to open ended questions; however, results were

nearly identical to using the forced choice items and therefore, quantitative analyses

based on responses to open ended questions were reported.

In each regression model, predictor variables were entered in blocks according to

hypothesized relationships from prior literature. Each block of variables was added

successively. When each block was added to the model, only variables that were

significant at p< 0.10 were retained. For these analyses, several variables were recorded

for ease of interpretation. Due to the sample's unbalanced distribution by maternal race,

minority ethnic groups were collapsed into one group and maternal ethnicity was

categorized as "Caucasian" (1) and "not Caucasian" (0). Maternal marital status was

coded as "1" for married and "0" for single, separated, widowed or divorced. Child's sex

was coded as 1 "male" and 2 "female". Only child status coded as 1 "yes" and 0 "no".

The first block of variables entered into the regression model contained one

variable, time elapsed between notification and current interview, to control for effects of

time. The second block of variables contained maternal demographic variables,

including maternal education level, ethnicity, marital status, number of children, and age

at the time of the interview. The third block entered contained child demographic








variables, including child's sex, whether an only child, and age at the time of interview.

Family history of diabetes was also included in this block, using two dichotomous

variables: (1) the presence of a first-degree relative with diabetes (yes/no) or (2) the

presence of a second or higher degree relative (yes/no).

The fourth block of variables contained the hypothesized predictor variable.

Predictor variables consisted of standardized composite scores on measures of perceived

control, risk perception, and anxiety, as well as total scores on measures of coping and

information seeking. Participation in PANDA Part II study was also used as

dichotomous predictor variable (yes/no). Reliability analyses were conducted on

composite scores suggesting that internal consistency was fair for these variables (Table

4-6). Each of the following hypotheses was examined separately. Within each model,

main effects were examined as well as interactions between perceived control and other

predictors, where noted. Only significant predictors were retained. Ultimately, a final

model was produced from these separate models to account for the highest classification

rate in the behavioral outcome variable. To account for type 1 error, a more conservative

level of significance was chosen at p < 0.01 and this is noted where appropriate.

Hypothesis 2.1

It was hypothesized that mothers who perceived they have control over their child

developing diabetes would be more likely to report engaging in behavior changes. Based

on statements regarding perceived control, which required an agree/disagree response,

38% percent of mothers reported believing they could do something and 27% believed

doctors could do something to prevent their child from developing type 1 diabetes.

Meanwhile, 52% reported believing it was up to chance or fate whether their child

develops type 1 diabetes (Table 4-6).








Hierarchical logistic regression analyses were conducted, using the composite score

for perceived control (belief that mother could do something or medical professional

could do something) to predict whether any behavior change was reported (yes/no) when

controlling for demographic factors (Table 5-11). Results indicated that mothers whose

children had a first degree relative with diabetes were significantly more likely to engage

in behavior change (odds ratio = 24.22, p < 0.001) and maternal perceived control was

not a significant predictor of behavior change, resulting in an overall model that

accounted for 67.5% overall correct classification.

Hypothesis 2.2

It was hypothesized that mothers who perceived their children to be increased risk

for type 1 diabetes would be more likely to report engaging in behavior change.

Hierarchical logistic regression was conducted as described previously; however, both

actual risk and the risk composite score were entered as the final block in the logistic

regression model. Results indicated that again, the presence of a first degree relative was

a significant predictor of behavior change (odds ratio = 18.98, p < 0.01). The child's

actual risk was found not to be significant. When controlling for actual risk, perceived

risk was a significant predictor (2.32, p< 0.01) (Table 5-12). Mothers who perceived

their children to be at greater risk were more likely to engage in behavior change. This

model resulted in an overall classification rate of 68.4%. An interaction between

perceived control and perceived risk was also tested and was not significant.

Hypothesis 2.3

It was hypothesized that mothers who were more anxious would be more likely to

report engaging in behavior change. This hypothesis was tested by entering the anxiety

composite score as the final block in the logistic regression model. Results indicated that








anxiety, as measured at the initial interview following notification, was not a significant

predictor of subsequent behavior change and was not retained in the final model.

However, mothers who were more anxious at the time of the current interview were more

likely to report behavior change (odds ratio = 2.98, p< 0.001). This model resulted in a

correct classification rate of 72.9% (Table 5-13). Follow-up analyses were conducted to

determine if there was an interaction between anxiety and perceived control; however,

none was found. Results demonstrated that while anxiety remained a significant

predictor, there was no main effect of perceived control, nor was the interaction term

significant, suggesting that mothers who were more anxious were more likely to report

behavior change to prevent diabetes in their child regardless of their level of perceived

control over the situation.

Hypothesis 2.4

It was hypothesized that mothers who used more coping strategies, particularly

active coping (i.e., problem-focused, seeking social support), would be more likely to

report engaging in behavioral change. Data were available for 176 mothers who

completed the Ways of Coping Checklist- Revised (WCC-R) at the 4-month interview.

In separate logistic regression models, each coping scale score was entered as the final

block of variables. Results indicated that after controlling for the significant effect of the

presence of a first degree relative, problem-focused coping (odds ratio = 10.72, P< 0.01),

seeking social support (odds ratio = 4.99, p< 0.01), and wishful thinking (odds ratio =

14.48, p< 0.01) were significant predictors of behavior change (Tables 5-14, 5-15, and 5-

16). While the two active coping factors were significantly related to behavior change, a

more passive coping style, wishful thinking, was also significant and to a relatively








higher degree. Item analysis of the wishful thinking scale indicated that this scale

included items related to optimistic thinking but also a desire for the problem to "go

away" or "be over with." It may be that wishful thinking reflects a sense of optimism and

urgency that might be associated with engaging in preventative actions believed by

mothers to be efficacious. Additionally, total coping as measured by the mean of all

reported coping behaviors was a significant predictor of reported behavior change (odds

ratio = 160.06, p< 0.001) (Table 5-17).

Hypothesis 2.5

It was hypothesized that mothers who engaged in information seeking and/or were

given recommendations by medical professionals or other family members related to

behavior change, would be more likely to report engaging in behavior change. Overall,

60% reported receiving information from at least one source, and the mean number of

sources was 0.99 (SD = 1.04). Overall, 33% reported receiving diabetes-specific

information from a book or other literature, 14% reported watching diabetes-related

television programming, and 12% reported seeking information using the internet (Table

4-11). As stated previously,

Fifty nine percent of mothers reported talking to their physician about their child's

genetic risk screening results. Of those, 41% reported receiving advice from their

physician, with over 89% reportedly taking their physician's advice. When specifically

asked in an open-ended question about the nature of the guidance given, mothers

specified a wide range of advice (Table 5-6). The most frequent advice given was to

monitor their child and promote a healthy lifestyle. Additionally, six mothers were told

to continue with PANDA study and five mothers were told by their physicians not to

worry about their child's risk.








Advice from family friends was similarly assessed. Eighty-six percent reported

talking to a family member or friend. Seventy percent reported talking with their spouse

about their child's genetic screening results, 63% reported talking with the child's

grandparent, 13% reported taking to a family member of friend who has diabetes and

32% reported talking with a family member or friend who does not have diabetes.

Seventeen percent of mothers reported receiving advice from at least one family

member/friend. Typically, advice was given by a child's grandparent (62%), followed by

spouse (15%) and friend or family member who does not have diabetes (15%), then

friend or family member who does have diabetes (5%). When specifically asked in open-

ended questions about the advice that was given, most frequent advice was to help child

maintain a healthy diet and five mothers were told not to worry. (Table 5-7)

Logistic regression was used to determine if the number of information sources

predicted the likelihood of engaging in behavior change. The number of information

sources was entered as the last block of predictor variables in a logistic regression model

(Table 5-18). When controlling for the presence of a first degree relative (odds ratio =

26.31, p < 0.01), those with more sources of diabetes-specific information were

significantly more likely to report engaging in behavior change (odds ratio = 2.27, p<

0.001). The presence of a first degree relative combined with the degree of information

sources together resulted in an overall classification rate of 74.3%.

Follow-up logistic regression analyses were conducted to determine if an

interaction was present between perceived control and the number of information

sources; none was found.








Hypothesis 2.6

It was hypothesized that mothers who continued their participation in the

Prospective Assessment of Newborn Diabetes Autoimmunity (PANDA) study by

participating in periodic blood testing for antibodies, would be more likely to report

behavior changes. Overall, 61% of mothers participated in at least one subsequent blood

draw. Twenty-six percent participated in two or more. However, when asked in the

interview, 174 mothers reported accurately whether they participated in part II of the

study (91%), and 3 reported participating when they actually have not (1%), and 15

reporting they had not participated when they actually had (8%).

Surprisingly, in the logistic regression model, registry participation using either

number of blood draws or continued participation (yes/no) did not predict to reported

behavior change (Table 5-19a, b). Mothers who continued with the PANDA study were

no more likely to report engaging in preventative efforts, despite their already active

participation in health surveillance.

Summary Model

Logistic regression was conducted to determine which of the previously listed

variables were most predictive of behavior change. As in previous analyses, family

history, characterized by presence of a first-degree relative, was entered as the first block

of variables as it had been found to be consistently significant in all previous models.

Actual risk was entered next in the model followed by all six variables also found to be

significant at the 0.01 level in previous models (i.e., perceived risk, anxiety, number of

information sources, problem focused coping, seeking social support, and wishful

thinking) were subsequently entered simultaneously. Problem focused coping and








seeking social support were dropped from the resulting model, as they were not

significant.

The final logistic regression model showed the presence of a first degree relative

was once again a significant predictor of behavior change (odds ratio = 19.34, P = 0.01).

Number of information sources, anxiety, perceived risk, and wishful thinking were also

significant predictors (Table 5-20). Overall, the model's classification rate was 77.7%.

Objective 3

It was hypothesized that mothers who reported modifying behaviors would show a

greater reduction in anxiety over time than mothers who did not report engaging in

behavior change. To examine this hypothesis, hierarchical linear regression was

conducted similarly to logistic regression procedures described for Objective 2, except

the dependent variable was the anxiety composite score, a continuous variable. Anxiety

at the initial interview was entered as the first block of variables followed by same

ordering of blocks of variables of demographic variables described previously. Reported

behavior change and the composite score for perceived control were entered as the final

(fourth) block of predictor variables to determine if behavior contributed significantly to

anxiety at the final follow-up interview, above and beyond the effect of initial anxiety

and demographic predictors. In a follow-up model, the interaction term between

behavior change and perceived control was added as the fifth block of predictors.

Results indicated that initial anxiety was a significant predictor of anxiety at the

current interview (J = 0.42, p< 0.001) accounting for 22% of the variance (Table 5-21).

Current age of the child (B = -0.18, p< 0.01) along with the presence of a first degree

relative (j = 0.29, p< 0.001) and the presence of a second or higher degree relative (3 =

0.18, p< 0.01) together accounted for an additional 13% of the variance (p < 0.001.








Reported behavior change (j = 0.24, p< 0.001) and perceived control (j = 0.20, p< 0.01)

were entered as the final block and both were found to be significant predictors,

accounting for an additional 9% of variance (p < 0.001). Overall, the model accounted

for 43% of the total variance. In the follow-up model, the interaction term was added and

not found to be significant. Results indicated that initial anxiety was the primary

predictor of anxiety at the time of the current interview. However, above and beyond

initial anxiety, mothers whose children were younger and had a relative with diabetes

were more anxious at the time of the current interview. Mothers who reported at least

one behavior change were significantly more anxious at both post-notification and

current interviews (as measured by the STAI only) than mothers who reported no

behavior changes (initial interview: M = 42.75, SD = 14.54 versus M = 36.80, SD =

12.54, t(1, 189) = 2.76, p < 0.01) (current interview M = 32.57, SD = 10.28 versus M =

27.14, SD = 7.02, t(1, 190) = 4.29, p < 0.001). Contrary to the original hypothesis, when

controlling for demographics and initial anxiety, mothers who reported at least one

behavior change and who perceived greater control over the onset of diabetes in their

children were more anxious at the time of current interview than mothers who did not.

This suggests behavior change maintains, rather than reduces anxiety over time for these

mothers.

Objective 4

Questions regarding behavior change used in the current interview were

developed from the DPT-1 survey (Johnson, 2002) and therefore, dichotomous scoring

for some of the questions in the current interview were comparable. The database from

the current study was merged with the maternal report data from the DPT-I study. Only

data collected from mothers who were not aware of the study results were included (n =








116). Of these mothers, 63 (53%) had children who participated in the control group of

the study and 53 (47%) had children enrolled in the experimental arm. Children whose

mothers completed the DPT-1 survey were significantly older than children in the

PANDA sample, ranging in age from 5 to 19 years old (M = 12.18, SD = 3.24) (t(2, 306)

= 32.02, p < 0.001). Reported maternal behavioral data from these two populations were

compared on 17 overlapping variables. Analyses were conducted across the

corresponding individual questions and similar domain scores.

We hypothesized mothers of genetically at risk children would be less likely to

report behavior change than mothers of ICA+ children enrolled in Diabetes Prevention

Trial-1 (DPT-1) ( = 116). On questionnaire items that were shared by both studies,

43.2% of mothers whose children were in the DPT-1 study and 33.3% of mothers in the

current sample reported at least one behavior change (Table 5-22). However, this

difference was not significant (p = 0.08). Mothers in the two samples reported similar

proportions of behavior change in the domains of diet and exercise; however,

medications differed by groups with mother from the DPT-1 sample reporting greater use

of medications/supplements.

There were few significant differences between the two samples of mothers on

specific items. Mothers in the DPT-1 sample were nearly four times more likely to report

feeding their children more diet and sugar free drinks (p< 0.001), and more often reported

feeding their children less regular soda (p< 0.05), whereas, mothers in the current sample

more often reported feeding their children less juice (p< 0.05). Administering vitamins

(p< 0.05) and administering insulin at home (p< 0.05) were practices that were

significantly more common in the DPT-1 sample. This is not surprising given that 53 of





91

the mothers had children who were in the experimental arm of the study involving home

insulin injections (46%) and the question involved giving "extra" insulin above and

beyond study protocol. Out of the 5 mothers who reported giving their child insulin, 4

(80%) were mothers of children enrolled in the experimental group.








Table 5-1. Comparisons of maternal demographic variables between participants in current
sample versus those eligible who were unable to be contacted or declined
participation (N= 368)
Unable to
Completers contact/declined Total
(n = 192) (n = 176) (n = 368) F (1, 434)
or_


Maternal age at notification
Current maternal agea


Race
Caucasian
African American
Hispanic
Asian/Other


Mothers level of education
High school or less
Some college/trade school
College degree or beyond

Marital Status (married)


Annual income
(in $10,000 intervals)

Number of Children


33.49 + 5.36
33.67 + 5.38


162 (84.9)
6(3.1)
13 (6.8)
10 (5.2)


45 (23.7)
62(32.1)
85 (44.2)

164 (85.4)

4.97+2.50


2.09+ 1.11


27.71+5.35
31.21+ 5.33


135 (76.7)
17 (9.7)
17(9.7)
7 (4.0)


50 (28.2)
74(41.8)
53 (29.9)

112(64.4)

3.82 + 5.30


2.03 + 1.26


29.28+5.63
32.41+5.44


297 (80.9)
23 (6.3)
30 (8.2)
17 (4.6)


95 (25.9)
137 (36.8)
138 (37.3)

276 (75.5)

4.43 + 2.40


2.06 + 1.18


Number of previous 39.83*
interviews


Number of blood draws (II)
0
1
2
3


Anxietyb


Belief about when child may
develop diabetes
Never


9 (4.7)
23 (12.0)
160 (83.3)


74(39.1)
47 (24.5)
29(15.1)
41(21.4)

40.74+14.13


36 (20.3)
41(23.2)
100 (56.5)


90 (50.8)
59 (33.3)
20(11.3)
8 (4.5)


39.83 + 14.29


45 (12.0)
64 (17.2)
260 (70.8)


164 (44.7)
106 (28.6)
49 (13.4)
49 (13.4)


40.30 + 14.19


43 (22.4) 34 (19.3) 77 (20.8)


14.42***
12.18**


7.32


11.72**


18.71**

20.22***


44.03**


0.70


1.90


43 (22.4) 34 (19.3)


77 (20.8)