:

25 £235.: ..
6 um w ;ﬁﬁumm

.w ..
4:2. ‘
W. 1.35....

. ‘ a H
AW-MW-Mwﬁyu.
ta... :,

ém .

p“.-

_ . . 5
5a.. 9,

- ‘ . .

”32.1

.0; . .

Fug?“

A in

“can“ A.
p. x
.5.\ n
.2.

L... 3.

.. 1....lt}?
Ju (~41! TL

a; lull

futiknuwﬁq.
.83.: e:

.

1.1, . z .
Pb?

19.
sunnn
.53:
.. s.

4.x a.”

 

 

2| Ir 1'

Tn! ‘ 1. 14.1..Tﬂdkf 1.4.3.1.

 

n. I J .a . .I ~\ ~ ‘ ~ ». . ,1
33:31.31, 3‘ ‘ 11¢--.ﬂ4rasu...,,...1l... .31 - . MUNWJ . \L‘a .3 J1}, . as w .2 a.» .3
l .9. . .1: . .uf, . ‘

‘ 1.-

3 .
g52"2:}m LIBRARY
Michigan State
University

This is to certify that the
thesis entitled

EXPLORING FACTORS ASSOCIATED WITH
LOW-INCOME MOTHERS' INFANT FEEDING CHOICES

presented by

ALLISON CHURILLA

has been accepted towards fulfillment
of the requirements for the

 

MS. degree in Park, Recreation & Tourism Resources

[ﬁg/«A W

M'éjor Profes§oi=~e5+gh ature

De. camber I 2003

 

Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

PLACE IN REIURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/ClRC/DateDue.p65~p.15

 

EXPLORING FACTORS ASSOCIATED WITH
LOW-INCOME MOTHERS’ INFANT FEEDING CHOICES

By

Allison Churilla

A THESIS
Submitted to
Michigan State University

in partial fulﬁllment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Park, Recreation and Tourism Resources

2003

 

ABSTRACT

EXPLORING FACTORS ASSOCIATED WITH
LOW-INCOME MOTHERS’ INFANT FEEDING CHOICES

By

Allison Churilla

The purpose of this study was to explore infant feeding choices in the context of
external inﬂuence from individuals and groups in mothers’ social networks, adjustment to
and perceptions of their own maternal and sexual femininity during and after pregnancy,
orientation toward speciﬁc feeding methods, and socio-demographic and cultural factors.
The sample for this study was composed of 200 women that participated in the Special
Supplemental Nutrition Program for Women, Infants, and Children (WIC) in two urban
areas of Michigan during the months of June and July 2003. Primary data was collected
from questionnaires administered to WIC clients and secondary data on WIC clients was
made available from the WIC Division of the Michigan Department of Community
Health. Two regression models were proposed that predicted mothers’ infant feeding
behavior on (1) discourses of maternal and sexual femininity and (2) authority of other
individuals. 0 n b oth m odels, m others’ b elief in the beneﬁts of breastfeeding was the
strongest statistically signiﬁcant predictor of breastfeeding percentage. Two other beliefs
were statistically signiﬁcant negative predictors on both models: belief that breastfeeding
limits lifestyle habits and belief that breastfeeding is painful. Implications for practice
suggest heightened attention to mothers’ needs and desires; implications for future
research focus on further reﬁning existing scales and employing qualitative methods to

further explore these results.

ACKNOWLEDGEMENTS

It is with deep gratitude that I thank my academic advisor, Dr. Dwayne Baker, for
allowing me to spread my wings on this project. This project required him to embark on a
journey outside his typical area of interest. With some certainty, I would also venture to
guess that, on some occasions, this journey carried him outside his typical comfort zone
as well. Under these circumstances, there are few individuals that would have embraced
this project with such vigor and invested such time, resources and care into my academic
career. Moreover, it is a rare faculty member that takes such effort to cultivate a personal
relationship with their student based on respect and genuine interest in life outside the
classroom. It was a stroke of luck that I crossed paths with such a faculty member and it
is this relationship that I will value long after this project has been completed.

From the beginning of this project, Dr. Baker recognized the strengths that several
colleagues 0 ould c ontribute to the p roject a nd m ade g reat e fforts to help me develop
relationships with other experts in the ﬁelds of sociology, medicine, and nutrition. On this
note, I also extend great thanks to the three faculty members that agreed to serve on my
master’s thesis committee. Dr. Linda Blum, associate professor in the Department of
Sociology at the University of New Hampshire, was asked to serve on the committee
based on her extensive work on the topic of infant feeding from a sociological
perspective. Dr. Blum’s thoughtful attention to the development of the proposal and ﬁnal
product was invaluable, as she challenged me to critically examine my own assumptions

about sociological research, infant feeding, race and ethnicity, and class.

iii

 

I also extend warm thanks to Dr. Margot Kurtz, professor in the Department of
Family and Community Medicine at Michigan State University, who agreed to serve on
my thesis committee despite her great geographical distance from the United States for a
large portion of the project. Dr. Kurtz offered great strength to my project in several
ways, not least 0 f w hich w as he r intense kno wledge o f quantitative methods and her
empirical understanding of women’s breasted experiences in relation to breast cancer.
Moreover, Dr. Kurtz was a source of endless support and encouragement throughout the
project, which contributed greatly toward my ability to face project challenges and
ultimately complete this project.

My ﬁnal committee member, Dr. Beth Olson, assistant professor in the
Department of Food Science and Human Nutrition at Michigan State University, was a
considerable resource in this project. It was Dr. Olson’s interest in infant feeding that ﬁrst
ignited my interest in this topic. Dr. Olson’s passion for the topic was obvious and she
possessed great knowledge of and experience working with administrators, staff and
participants in the Michigan WIC program. Dr. Olson was also a rich resource for
identifying the latest research on this topic and she was constantly on the watch for
literature that would add dimension to my own research. I am incredibly grateful for her
periodic correspondence throughout the project and for her willingness to meet with me,
at short notice no less, when I was conﬁonted with speciﬁc challenges or opportunities.

The ﬁnancial support provided for this project was an incredibly generous
contribution to the process and completion of this study. For this, particular thanks are

directed to Karen Shirer of Michigan State University Extension, Family and Consumer

iv

Sciences. This support allowed greater access to resources for the study that enhanced the
breadth of the project.

In addition to the support I received from members of my thesis committee and
the ﬁnancial support I received, I was privileged to work with several individuals “in the
ﬁeld” that were committed to the success of the project. These individuals included Diane
Ayers, Pat Benton, Lynette Biery, Kathy Daly-Koziel, Dr. Kobra Eghtedary, Nancy
Erickson, Kathy Koss, Diane Revitte, LeeAnne Roman, David Sachau, and ofﬁce staff
and nutritionists at participating WIC clinics in the Kent County Health Department and
the Detroit City Health Department. I would like to express particular gratitude to Monica
Smith, WIC coordinator at the Kent County Health Department, for her dedication to this
project. It was my pleasure to work with an individual that demonstrated such passion for
her work and for the people she works with.

In my days spent in Room 17, I was lucky to work with individuals that also
demonstrated this passion and commitment. I am so grateful for the unending support
offered by Ariel Rodriguez, whose care and good humor added great warmth to Room
17. The ﬁiendship we developed over two years of sharing an ofﬁce space is an
invaluable outcome of my time at Michigan State. I am also thankful for the camaraderie
of other students in the department that, over my two years in the department, challenged
me to grow as a student and as a person.

Lastly, there are several people that deserve particular thanks for their support not
only in this project, but in all aspects of my life. I was blessed with strong support from
friends and family, w hich helped in w ays they will ne ver fully kno w. M y v ery g 00d

friend, Rob Palmer, was the person that let me talk, grumble, cry and laugh—often in the

wee hours of the morning—and never let me lose sight of the important things in this
process and in life. He was a true friend for this entire journey, struggled alongside me,
and shares this accomplishment with me. My brother, Jesse Churilla, has taught me a
great deal about having fun with life and was the person that helped me enjoy myself in
this process with our visit in April and our occasional conversations over the telephone.
And very special thanks go to my parents, Doug and Jean Churilla, who have always
made me feel like I am one of the most important and special people in the world. My
mom listened through countless telephone conversations to the challenges and successes
of the project and, regardless of how I felt before talking to my mom, I always felt better
after our conversations. My dad gave me the best advice of anyone throughout the
duration of this project, which made all the difference in my ability to embrace
challenges of the project. My dad never ended a telephone conversation with me without
asking, “Is there anything you need from us?” Their patience through my struggles and

pride in my accomplishments is my ultimate reward.

vi

TABLE OF CONTENTS

Page
LIST OF TABLES ...................................................................... viii
LIST OF FIGURES ..................................................................... xi
Chapter 1: INTRODUCTION ......................................................... 1
The Discursive Field of Infant Feeding .................................. 2
Sources of External Inﬂuence ............................................. 3
Maternal and Sexual Discourses of Infant Feeding .................... 8
Study Purpose and Objectives ............................................. l 1
Chapter 2: LITERATURE REVIEW ................................................. l6
Maternal F emininity ........................................................ 17
Sexual Femininity .......................................................... 20
Sources of External Inﬂuence ............................................. 24
Chapter 3: METHODS .................................................................. 37
Study Purpose ............................................................... 37
Study Participants ........................................................... 38
Study Design and Procedure .............................................. 40
Study Instrument ............................................................ 45
Chapter 4: RESULTS ................................................................... 68
Kent County Study Sample ................................................ 73
Detroit Study Sample ...................................................... 87
Study Objective 1: Comparison by Urban Location ................... 98
Study Objective 2: Assessment of Measurement Reliability ......... 123
Study Objective 3: Predicting Infant Feeding Behavior by Systems
of External Inﬂuence ....................................................... 134
Study Objective 4: Predicting Infant Feeding Behavior by Body
Image and Attitudes toward Motherhood ................................ 145
Chapter 5: DISCUSSION ............................................................. 153
Managerial Recommendations ............................................ 161
Research Implications ...................................................... 164
Study Limitations ........................................................... 167
Conclusion .................................................................. 1 73
Appendix A: MEASUREMENT OF FACTORS RELATED TO
MOTHERS’ INFANT FEEDING BEHAVIOR ..................... 175

vii

Appendix B: SURVEY INSTRUMENT ............................................. 184

Appendix C: APPROVAL LETTER FROM THE MICHIGAN STATE
UNIVERSITY COMMITTEE ON RESEARCH INVOLVING
HUMAN SUBJECTS ................................................... 192

REFERENCES ........................................................................... 194

viii

LIST OF TABLES

Number Page
1) Sampling Frame by WIC Agency and Clinic ........................... 39
2) WIC Secondary Data Extraction Measures and Original
Instruments ................................................................... 58
3) Expected and Obtained Study Sample by WIC Agency ............... 70

4) Comparison of Kent County Study Participants and Eligible and
Ineligible WIC Client Groups on Socio-Demographic
Characteristics ............................................................... 74

5) Comparison of Kent County Study Participants and Eligible and
Ineligible WIC Client Groups on Participation in Other Support
Programs ..................................................................... 80

6) Comparison of Kent County Study Participants and Eligible and
Ineligible WIC Client Groups on Characteristics of Prenatal Care... 82

7) Comparison of Kent County Study Participants and Eligible and
Ineligible WIC Client Groups on Infant Feeding Characteristics. . 84

8) Comparison of Detroit Study Participants and Eligible and
Ineligible WIC Client Groups on Socio-Demographic
Characteristics ............................................................... 88

9) Comparison of Detroit Study Participants and Eligible and
Ineligible WIC Client Groups on Participation in Other Support
Programs ..................................................................... 92

10) Comparison of Detroit Study Participants and Eligible and
Ineligible WIC Client Groups on Characteristics of Prenatal Care... 94

11) Comparison of Detroit Study Participants and Eligible and
Ineligible WIC Client Groups on Infant Feeding Characteristics. . 96

12) Comparison of Study Participants by WIC Agency on Socio-
Demographic Characteristics .............................................. 100-101

13) Comparison of Study Participants by WIC Agency on Participation
in Other Support Programs ................................................ 107

ix

14) Comparison of Study Participants by WIC Agency on
Characteristics of Prenatal Care ........................................... 109

15) Comparison of Study Participants by WIC Agency on Infant
Feeding Characteristics ..................................................... 1 l 1

16) Overall Mean Item Ratings and Rankings on Scales of the

Breastfeeding Attrition Prediction Tool (J anke, 1994) ................. l 16
17) Statistically Signiﬁcant Differences on Positive and Negative

Beliefs about Breastfeeding by WIC Agency ........................... 118
18) Overall Mean Item Ratings on Sources of External Inﬂuence ........ 119

19) Overall Mean Item Ratings and Rankings on Subscales of the
Maternal Adjustment and Maternal Attitudes Questionnaire
(Kumar, Robson and Smith, 1984) ....................................... 121

20) Statistically Signiﬁcant Differences on Body Image Items and
Attitudes toward Motherhood and the Baby by WIC Agency. . . . . 122

21) Items and Item Statistics for Factored Subscales of the Positive
Breastfeeding Sentiment Scale (Janke, 1994) ........................... 125

22) Items and Item Statistics for Factored Subscales of the Negative
Breastfeeding Sentiment Scale (J anke, 1994) ........................... 127

23) Items and Item Statistics for F actored Subscales of the Body Image
Subscale (Kumar, Robson and Smith, 1984) ............................ 130

24) Items and Item Statistics for F actored Subscales of the Attitudes
toward Motherhood and the Baby Subscale (Kumar, Robson and
Smith, 1984) .................................................................. 132

25) Zero Order Correlation Matrix of Dependent and Independent
Variables in Regression Model A ......................................... 138

26) Results of Stepwise Regression Analysis Predicting Breastfeeding
Percentage on Systems of External Inﬂuence ........................... 141

27) Results of Stepwise Regression Analysis Predicting Breastfeeding
Percentage on Systems Of External Inﬂuence (Alternative Model). 143

28) Zero Order Correlation Matrix of Dependent and Independent
Variables in Regression Model B ......................................... 147-148

 

 

29)

30)

A1)

Results Of Stepwise Regression Analysis Predicting Breastfeeding
Percentage on Body Image and Attitudes toward Motherhood and
the Baby ...................................................................... 149

Results of Stepwise Regression Analysis Predicting Breastfeeding
Percentage on Body Image and Attitudes toward Motherhood and

the Baby (Alternative Model) ............................................. 151

Measurement of Factors Related to Motherhood and Infant

Feeding ........................................................................ 175
xi

LIST OF FIGURES

Number Page

1) Proposed Regression Model ‘A’ Predicting Current Infant Feeding
Behavior on Systems of External Inﬂuence ............................. 63

2) Proposed Regression Model ‘B’ Predicting Current Infant Feeding
Behavior on Maternal Body Image and Attitudes Toward
Motherhood and the Baby ................................................. 66

3) Revised Regression Model A ............................................. 140

xii

Chapter 1
INTRODUCTION

Infant feeding is a socially situated phenomenon, wherein mothers’ choices may
be complex decisions that draw on medical knowledge, standards of responsible
motherhood, and social expectations of feminine sexuality. In the context of these social
expectations of women and mothers, infant feeding choices may be complex decisions
that are associated with expectations and support from family members, friends, health
professionals, and other social actors. For mothers making choices about infant feeding,
social expectations may also inﬂuence their perceived attractiveness, their ability to
adjust to motherhood and the new baby and, as a consequence, their orientation toward
methods of infant feeding. As a result, the amount of “choice” mothers exercise in their
infant feeding decisions may be inseparable from the social contexts in which mothers
make their decisions (Carter, 1995).

Racial, ethnic, and class-based factors may also shape and add dimension to the
social contexts in which mothers make choices about infant feeding. Infant feeding
choices and experiences may assume vastly different meanings across these categories
and, as a result, choices about infant feeding may be closely bound to issues of class
respectability, cultural norms, medical expectations, and historical precedent (Beal,
Kuhlthau, & Perrin, 2003; Blum, 1999; Weller & Dungy, 1986).

In this sense, mothers’ infant feeding “choices” may be largely guided by factors
that are external to the mother-infant dyad. At best, these choices are shaped by these
factors; at the extreme, these choices are constrained by social and cultural expectations.

Th erefore, in exploring these constructs, “the goal ought not to be the separation of this

particular decision from its surrounding social variables in order to make it an
unconstrained one; infant-feeding decisions are necessarily and productively connected to
a host of other choices and constraints” (Law, 2000, p. 445). As indicated, these choices
and constraints may include perceived expectations of others, adjustment to motherhood
in the context of maternal and sexual expectations of femininity, personal beliefs about
feeding methods, and racial, ethnic, and class-based factors.

Accordingly, the purpose of this study was to explore infant feeding choices in the
context of external inﬂuence from individuals and groups in mothers’ social networks,
adjustment to and perceptions of their own maternal and sexual femininity during and
after pregnancy, orientation toward speciﬁc feeding methods, and socio-demographic and
cultural factors.

The Discursive Field of Infant Feeding

This study drew on Foucault’s theories of discourse, power and the construction
of knowledge to explore how medical, maternal and sexual expectations are translated
into ways of knowing and understanding practices of infant feeding. Foucault commences
his discussion of authority with the concept of discourse, or the ways that individuals and
institutions speak about a particular topic (Foucault, 1978/1990, 1984). Based on the
variety of perspectives from which individuals a nd institutions m ay view a p articular
topic, it is possible for multiple discourses to exist on the topic that overlap and reinforce
one another, or diverge and compete with one another. These multiple discourses
constitute a discursive ﬁeld.

Drawing on Foucault, Carter (1995, p. 29) argues that infant feeding can be

interpreted as a discursive ﬁeld that encompasses “discourses concerned with femininity,

sexuality, mothering, class, race and gender relations.” This breadth of discourses
presents mothers with multiple ways of interpreting methods of infant feeding. These
ways of understanding infant feeding may be translated into bodies of knowledge
surrounding the topic of infant feeding, wherein individuals and the institutions in which
they are invested may construct bodies of knowledge that follow from the ways the topic
is discussed. Accordingly, Foucault deﬁnes discourse in the context of power, wherein
beliefs and meanings may be constructed to convey authority (Foucault, 1977/1995,
1978/ 1990, 1984).

Following from this theoretical framework, the current study aimed to investigate
infant feeding behavior in the Foucaultian context of two discourses of femininity—
matemal and sexual—and their intersections with race, education, and orientations
toward and away from breastfeeding and formula feeding. Further, based on the potential
authority conveyed by individuals and institutions in mothers’ social networks, this
investigation examined the individuals and systems of external inﬂuence that were
associated with mothers’ feeding behaviors.

Sources of External Inﬂuence

The inﬂuence of actors external to the mother-infant dyad on mothers’ infant
feeding choices may vary by source, perceived expectations of each source, and the
degree to which mothers value these perceived opinions. Research has identiﬁed a range
of inﬂuential actors and supportive contacts in mothers’ infant feeding choices
(Giugliani, Caiaffa, Vogelhut, Witter, & Perman, I994; Humphreys, Thompson, &
Miner, 1998; Janke, 1994; Jones, 1986, 1987; Matich & Sims, 1992). In order to

distinguish sources of social support by their relationship with the mother, sources of

support and inﬂuence may be grouped into family and friends, professional sources, and
community. These categories correspond with contexts of the ecological model of
breastfeeding developed by Tiedje et a1. (2002), which was constructed to explore the
inﬂuence of external environments on mothers’ infant feeding choices.

Expectations of mothers’ infant feeding choices may be communicated directly,
perhaps most notably by the healthcare delivery system, or indirectly through community
norms and societal expectations. Between these extremes, expectations may also be
expressed by social actors in the form of encouraging suggestions, biting remarks, or
indifferent comments about mothers’ infant feeding practices. In this context of direct
and indirect expectations, mothers’ perceptions of others’ expectations may be accurate
or inaccurate representations of how others feel. Nonetheless, perceptions may be
powerful inﬂuences on mothers’ orientations toward methods of infant feeding and
subsequent infant feeding choices (Janke, 1994).

As one of the most immediate contexts of mothers’ external environments, family
and ﬁiends may exert inﬂuence on mothers’ infant feeding choices in both supportive
and constraining capacities (Grossman, F itzsimmons, Larsen-Alexander, Sachs, &
Harter, 1990; Guttman & Zimmerman, 2000; Matich & Sims, 1992). Family inﬂuences
on mothers’ infant feeding choices may be particularly important because these
individuals may be most likely to provide emotional support in mothers’ decision-making
processes and tangible support in the household (Grossman et a1., 1990). In particular, the
baby’s father may be particularly inﬂuential in shaping mothers’ orientations toward
feeding methods and, as a result, health professionals have been overwhelmingly

encouraged to include the father in breastfeeding promotional efforts.

For mothers whose social network relies on extended family and ﬁiends,
emotional and tangible support for mothers’ infant feeding choices may be provided in
different, and yet equally supportive, capacities. These social networks may be
particularly characteristic of African-American and Latin-American working-class
mothers, constructed on values of “community-based independence,” reciprocity, and
shared child-rearing (Blum & Deussen, 1996; Dominguez & Watkins, 2003). In this
function, family and friends may offer mothers the opportunity to make infant feeding
choices in the context of encouraging and supportive environments.

While these social networks may be supportive, they may also—at the same
time—discourage or constrain mothers’ choices about feeding methods. Dominguez and
Watkins (2003, p. 119) describe the potential tensions that can arise among mothers’
family and friends network as “a constant negotiation—yielding resources, but not
without signiﬁcant conﬂict.” In addition to providing support, therefore, friends have also
been cited as one of the most potentially disapproving sources of support for
breastfeeding, particularly by expressing concern about mothers’ ability to maintain their
social lifestyle (Guttman & Zimmerman, 2000). As well, family members have also been
cited as a source of breastfeeding discouragement when they are at least partially
responsible for care of the infant (Guttman & Zimmerman, 2000). As a result, in
attempting to negotiate support and constraints from family and friends, mothers may
limit their own infant feeding choices out of consideration of factors related to
convenience, freedom, and shared childcare with family and friends.

Perhaps one of the most direct social actors in inﬂuencing mothers’ infant feeding

choices has been the healthcare delivery system in the United States. This institution has

been explicit about advising mothers on the advantages of breastfeeding. At an individual
level, health professionals may counsel mothers on the superior infant health and
nutrition beneﬁts of breastfeeding. On a larger scale, the medical community and health
professionals have partnered with federal departments to promote breastfeeding as a
national health objective. In 1979, the United States Department of Health and Human
Services (HHS) instigated a national planning process labeled the Healthy People
initiative to improve public health and medicine. At ten—year intervals, the initiative
identiﬁes health objectives and reports outcomes (US. Department of Health and Human
Services, 2000; US. Department of Health and Human Services (HHS), 1979, 1990,
2000a). Objectives of the most recent report, Healthy People 2010, aim for 75 percent
breastfeeding rate among mothers in the early postpartum period, 50 percent continuing
to breastfeed at six months, and 25 percent continuing to breastfeed at one year (U .S.
Department of Health and Human Services (HHS), 2000a, 2000b). Each of these goals
targets a sizeable increase in breastfeeding from 1998 baseline rates.

In order to achieve this goals and other goals related to infant health and nutrition,
the US. Department of Health and Human Services (HHS) and the Food and Nutrition
Service (FNS) began co-hosting biannual meetings of the Breastfeeding Promotion
Consortium (BPC) in 1990. These meetings provide leaders of over 25 health
professional associations, breastfeeding advocacy groups and Federal agencies an
Opportunity to exchange ideas and work collaboratively to increase rates Of breastfeeding
in the United States to those recommended by Healthy People 2010 (Food and Nutrition

Service (FNS), 2003).

Professional advice on the advantages of breastfeeding has often been conﬁned to
discussions about infant health and development at the neglect of other considerations in
mothers’ infant feeding choices (Blum, 1999). Notably, this professional advice
advocating breastfeeding may also be conﬁned to mothers that engage in behaviors and
possess characteristics that indicate responsible motherhood. Indeed, some women may
be advised to not to breastfeed when breastmilk may threaten infant health, as in the case
of “women who use illicit drugs, who have active, untreated tuberculosis, or who test
positive for HIV, as well as those who use certain prescribed drugs” (U .S. Department of
Health and Human Services (HHS), 2000a, p. 16-47). In these circumstances, Blum
(1999) is explicit in highlighting medical contraindications that conﬂate characteristics
associated with unsafe breastmilk with racial, ethnic and class-based assumptions about
reliable mothers. Blum (1999, p. 142) cautions that, in light of recent emphasis on diverse
approaches to breastfeeding promotion in the healthcare delivery system, an approach
that treats groups of mothers as suspect may function “punitively by ﬂattening ambiguous
facts” such as race, ethnicity and class. As a result, in deﬁning public health objectives,
some mothers may be categorized according to class-based and culturally-situated
expectations of responsible, and irresponsible, motherhood.

Along these lines, a second objective of Healthy People 2010 emphasizes
breastfeeding promotion among groups with exceptionally low rates of breastfeeding
initiation and dur ation. Ac cording to the r eport, “increasing the r ate 0 f b reastfeeding,
particularly among low-income and certain racial and ethnic populations less likely to
begin breastfeeding in the hospital or to sustain it throughout the infant’s ﬁrst year, is an

important public health goal” (US. Department of Health and Human Services (HHS),

 

2000a, p. 16-47). In its intention, this 0 bjective aims to inc rease infant a nd m atemal
health among populations determined to be at particular risk. However, such an objective
has the potential to be “both controlling and helpful” by developing greater resources for
support that are sensitive to cultural and class-based differences and, at the same time,
carefully monitoring and guiding infant feeding practices among these populations
(Blum, 1999, p. 141).

Researchers have noted that expectations of family and ﬁiends, professional
sources, and community may interact in varying, overlapping, and competing capacities
to provide support or exercise inﬂuence on mothers’ decisions (Carter, 1995; Tiedje et
al., 2002). Perhaps most notably, professional recommendations regarding infant feeding
have often been aligned with societal beliefs about responsible motherhood. Research
suggests that many mothers are aware of and believe in the superior infant health beneﬁts
of breastfeeding (Brownell, Hutton, Hartman, & Dabrow, 2002; Gabriel, Gabriel, &
Lawrence, 1986; Schmied & Lupton, 2001). This may suggest that social constructions of
responsible motherhood are necessarily tied to medical recommendations and expert
advice. Wall (2001, p. 593) concurs, arguing that motherhood encompasses “not only
notions of the naturalness and purity of breastfeeding and breast milk but also the
scientiﬁc and medical colonization of reproduction and child rearing.”

Maternal and Sexual Discourses of Femininity

However, this medical model of motherhood, with its emphasis on infant health
and nutrition, m ay 0 verlook o r o versimplify t he importance of other factors that may
inﬂuence mothers’ infant feeding choices. For some women, the infant feeding decision

may induce conﬂict between expectations of the medically conscientious mother and the

sexually attractive woman (Stearns, 1999; Young, 1990). Each set of social expectations
calls for the emergence of a different set of feminine characteristics. Accordingly, social
expectations of the medical mother portray a woman that is nurturing and responsive to
the needs of her child. Further, this woman is a resource for the healthy development of
her children and therefore subscribes to the medical model of “breast is best.”

Even among mothers that subscribe to the superior health beneﬁts of
breastfeeding, a large portion of mothers nonetheless decide not to initiate breastfeeding
or stop breastfeeding prior to professional recommendations. One reason for this apparent
discrepancy may be that breastfeeding, particularly in the presence of others or in public,
may be a context in which expectations of the maternal body conﬂict with expectations of
the sexual body (Stearns, 1999). Social expectations of the feminine sexual body portray
a woman that is sexually alluring and physically attractive to her partner and to others. In
a culture that sexualizes breasts, breastfeeding may introduce an alternative approach
toward breasts that is not sexual, but rather is nurturing and life-giving. Young (1990, p.
197) describes mothers’ resulting infant feeding choices as “divided,” wherein women
are forced “to ide ntify w ith o ne 0 r a nother im age 0 f w omanly p ower—the nur turing,
competent, selﬂess mother, always sacriﬁcing, the soul of goodness; or the ﬁery,
voluptuous vamp with the power of attraction, leading victims down the road of pleasure,
sin, and danger.”

In the presence of potentially conﬂicting social expectations of mothers and
women, mothers’ attitudes toward their body and motherhood may be shaped. Maclean
(1990, p. 3) explains, “Women’s beliefs, attitudes, and experiences do not develop in

isolation. They reﬂect not only a woman’s unique personality and preferences but also

the social forces and Viewpoints that are a part of the world in which she lives.” Further,
socially guided attitudes may orient mothers toward or away from speciﬁc methods of
infant feeding. Diverse, and potentially competing, social expectations of the maternally
nurturing woman, the medically conscientious woman, and the sexually attractive woman
may shape mothers’ preferences for speciﬁc infant feeding methods. As a result, mothers’
beliefs and their infant feeding choices may reference a medical model of motherhood,
while at the same time taking account of sexual expectations of femininity. Therefore,
although women’s beliefs are Often assumed to be highly personal, they are also situated
within a social context of potentially conﬂicting norms and expectations about mothers
and women.

The interaction of social expectations of women and mothers may inﬂuence
mothers’ infant feeding choices in constraining and liberating capacities. As mothers
make choices about methods of infant feeding, mothers’ decisions necessarily take place
in environmental contexts of social support and inﬂuence, expectations of maternal
responsibility, and accepted norms of sexual femininity. Examining how these contexts
interact with mothers’ orientations toward infant feeding and how they subsequently
inﬂuence mothers’ infant feeding choices may provide insight into the discursive ﬁeld of
infant feeding.

As Blum (1999) and others have indicated, discussions of infant feeding must
necessarily account for class—based and culturally-situated expectations of motherhood
(Grossman, Larsen-Alexander, Fitzsimmons, & Cordero, 1989; Jacobson, Jacobson, &
Frye, 1991; Kurinij, Shiono, & Rhoads, 1988; Li, Fridinger, & Grummer—Strawn, 2002;

Weller & Dungy, 1986). Low income women and speciﬁc groups of racial and ethnic

10

minority women may be placed with a different set of expectations from external social
actors and the wider community. As a result, these populations may orient themselves
toward and away from feeding methods in culturally and historically speciﬁc ways.
Understanding the processes that guide and shape mothers’ choices, with speciﬁc
sensitivity to demographic and cultural characteristics, might add to understandings of the
discursive ﬁeld of infant feeding as well.

Study Purpose and Objectives

Because infant feeding choices are made in these social contexts, the purpose of
this study was to explore factors associated with infant feeding behavior among low-
income women. Four objectives guided this study.

Objective 1. The ﬁrst objective of this study described overall characteristics of a
sample of postpartum, low-income mothers from two urban areas in Michigan and
compared these characteristics by urban location. One major provider of services for
these populations is the Special Supplemental Nutrition Program for Women, Infants, and
Children (WIC). This program provides food assistance and educational services to low-
income pregnant women, breastfeeding and non-breastfeeding mothers, infants, and
children that are determined to be at nutrition risk. Because low income and speciﬁc
racial and ethnic groups have been targeted as a national priority for improving public
health, mothers that participate in the WIC program may be particularly prone to racially-
and class-situated inﬂuences in their infant feeding choices. This study therefore drew its
sample from clients of this program in two major metropolitan areas of Michigan.

At the federal level, WIC is administered by the Food and Nutrition Service

(FNS) of the United States Department of Agriculture. As a federal program, Congress

11

appropriated $4.687 billion to the WIC program for the 2003 ﬁscal year, which the ENS
dispenses to states to provide WIC program services (Food and Nutrition Service (FNS),
2003). Services include nutrition counseling and referrals to health care providers and
other social services. WIC participants may also receive food beneﬁts, which are
distributed through monthly vouchers that allow participants to purchase speciﬁed foods
that are high in protein, calcium, iron, and vitamins A and C. Program eligibility is based
on income guidelines, state residency, and acknowledged “nutritional risk,” including
medically-related risks or diet-related risks as determined by a qualiﬁed health
professional (Food and Nutrition Service (FNS), 2003). However, WIC is not an
entitlement program and WIC beneﬁts may not be available to all eligible applicants. As
a result, guidelines have been established to prioritize applicants based on their level of
nutrition risk.

Although WIC provides formula to participants that choose not to breastfeed,
WIC has historically supported breastfeeding through federal legislation that requires
breastfeeding promotion and authorizes enhanced beneﬁts for breastfeeding mothers
(Food and Nutrition Service (FNS), 2003). When the program ofﬁcially became
permanent in 1975, federal legislation allowed breastfeeding mothers to participate in
WIC through one year postpartum, whereas non-breastfeeding mothers were eligible
through six months postpartum. Congress placed increased emphasis on WIC’s
breastfeeding promotion efforts in 1989 through legislation that called for standards of
breastfeeding promotion at the federal and state level. These standards required states to
designate administrative staff for breastfeeding support and allocate a speciﬁed amount

of their budget toward breastfeeding promotion. In recent years, Congress has also

12

required the USDA to develop a national breastfeeding promotion plan and has
authorized states to use food funds for the purchase or rental of breastpumps.

Of the $4.687 billion appropriated to the WIC program for the 2003 ﬁscal year,
Michigan was allocated over $128 million in federal grants for WIC program
administration, nutrition services, and food beneﬁts (Food and Nutrition Service (FNS),
2003). In 2002, Michigan’s WIC Division provided services and beneﬁts to over 200,000
women, infants, and children (FNS, 2003; MDCH, 2003). There are 50 WIC agencies
throughout the state, each of which manages several WIC clinics throughout the area.
Study participants for the current investigation were recruited from clinics managed by
WIC agencies in two major metropolitan areas of Michigan: Kent County and Detroit.

Objective 2. The second objective of this study was to test the measurement
reliability of scales and subscales of two instruments developed to measure aspects of
motherhood and infant feeding among this sample of WIC participants. Sampling
methods used to test the reliability of these scales may have limited the applicability of
the scales to more diverse sample populations. As a result, this objective provided an
initial step toward exploring reliability of the scales among a diverse sample of low
income women.

The ﬁrst scale, the Maternal Adjustment and Attitudes Questionnaire (Kumar,
Robson, & Smith, 1984), contains ﬁve subscales that measure maternal conﬁdence and
adjustment to aspects of the maternal role, including Body Image, Somatic Symptoms,
Marital Relationship, Attitudes to Sex, and Attitudes to Pregnancy/Motherhood and the
Baby. The original instrument was tested among two samples of women clients at

antenatal clinics in London; the majority of women in these samples were manied or

13

 

stably cohabiting with the baby’s father (Kumar et al., 1984). For the current
investigation, marital status among the sample may be more diverse and may add to the
reliability of other scales that do not rely on a marital or cohabiting status. Consistent
with the purpose of this study, two subscales were selected from the instrument to be
tested among this sample in order to explore mothers’ perceptions of their own sexual
attractiveness and maternal femininity: body image and attitudes toward motherhood and
the baby.

The second instrument, the Breastfeeding Attrition Prediction Tool (Janke, 1994)
contains four scales that measure predictors of breastfeeding duration: Positive Sentiment
toward Breastfeeding, Negative Sentiment toward Breastfeeding, Social and Professional
Support, and Breastfeeding Control. The instrument was originally tested among a
convenience sample of 201 breastfeeding women, the majority of whom were white,
married, had obtained schooling beyond high school, and were middle- to upper-income
(Janke, 1994, p. 101). Similar to this sample, the sample for the current investigation was
relatively homogenous with respect to income, as one criterion of WIC eligibility is low
income. However, the sample that was drawn for the current study was c onsiderably
more diverse with respect to all other characteristics, particularly with reference to race,
marital status, and education. Thus, the current investigation tested measurement
reliability of the Positive and Negative Breastfeeding Sentiment Scales in order to extract
categories of mothers’ beliefs about infant feeding methods.

Objective 3. The third objective of this study examined the extent to which
mothers’ infant feeding behavior was inﬂuenced by sources of inﬂuence external to the

mother-infant dyad. Because research suggests that infant feeding behavior may be

14

moderated by other factors such as race/ethnicity (Baranowski et al., 1983; Blum, 1999;
Park, Meier, & Song, 2003), and mothers’ personal orientations toward feeding methods
(Li et al., 2002; Losch, Dungy, Russell, & Dusdieker, 1995), these characteristics were
included as control variables in this model of mothers’ infant feeding choices.

For this model, mothers’ infant feeding behaviors was examined in relation to
three systems of external inﬂuence: family and friends, the healthcare delivery system,
and community sources. Measurement of these characteristics drew on the format of
Janke’s (1994) Social and Professional Support Scale to explore external systems of
inﬂuence on mothers’ choices.

Objective 4. The fourth objective examined the extent to which infant feeding
behavior is predicted by mothers’ perceptions of sexual and maternal femininity. Two
subscales Of the Maternal Adjustment and Maternal Attitudes Questionnaire (Kumar et
al., 1984) were used to measure mothers’ perceptions Of their body and mothers’
adjustment to motherhood and the mothering role. These subscales were labeled Body
Image and Maternal Attitudes toward Motherhood and the Baby, respectively.

Chapter 2 explores the literature related to mothers’ infant feeding behavior,
including sources of external inﬂuence, discourses of femininity, and socio-demographic

and cultural characteristics that may interact with these factors.

15

Chapter 2
LITERATURE REVIEW

Foucault’s theories of the construction of knowledge and power provide an
appropriate approach to exploring how social expectations of mothers and women are
translated into ways of knowing and understanding practices of infant feeding. For
Foucault (1978/1990), discourses are central to the construction of knowledge and are
expressed in the ways individuals speak about a particular topic. Based on the diverse
perspectives from which individuals and institutions approach speciﬁc topics, multiple
discourses may exist surrounding one topic. These discourses need not be complementary
and, in fact, Foucault explains the evolution of discourses on a topic as divergent.
Accordingly, “it is not simply in terms of a continual extension that we must speak of this
discursive growth; it should be seen rather as a dispersion of centers from which
discourses emanated, a diversiﬁcation of their forms, and the complex deployment of the
network connecting them” (Foucault, 1978/1990, p. 34). As will be discussed, research
on discourses particular to the topic of infant feeding has indicated that there may be
multiple and divergent discourses surrounding the topic that add complexity to mothers’
feeding behaviors.

The existence of several discourses on a unique topic constitutes a discursive
ﬁeld, which Carter (1995) discusses with particular reference to infant feeding.
According to Carter (1995), infant feeding can be interpreted as a discursive ﬁeld that
references multiple, and potentially competing, discourses on the topic. Theorists have
aimed to identify a range of discourses surrounding the topic of infant feeding, including

discourses of risk, religion, ethics, law, sexuality, and maternity (Law, 2000; Maher,

l6

1992; Nadesan & Sotirin, 1998; Traina, 2000; Young, 1990), two discourses particular to
expectations of femininity have been empirically examined in their relationship with
infant feeding beliefs and behavior.

Maternal Femininity

Expectations of women as mothers are one context in which women may make
decisions about how to feed their infant. Murphy (2000) reported that mothers may feel
considerable guilt when, after acknowledging the superior infant health beneﬁts of
breastfeeding, they nonetheless decide to introduce formula to the infant. In interviews
with thirty-six ﬁrst-time mothers, Murphy (2000) cited the potential conﬂation of
breastfeeding with constructions of responsible motherhood, concluding that among her
study sample the “subsequent decision to introduce formula milk raised questions about
their identities as ‘good mothers’” (Murphy, 2000, p. 303).

Further, Murphy (2000) points to this ﬁnding as an indication of the strength Of
the medical ﬁeld in guiding discourse on maternal femininity. Wall (2001) added to this
ﬁnding in her content analysis of informational materials on infant feeding made
available to expectant and postpartum mothers in Canada. Results of this study indicated
that informational material p resented he alth a nd nutrition information int andem with
messages about responsible, caring and generous motherhood. However, as the research
of Schmied and Lupton (Schmied & Lupton, 2001) suggests, not all mothers experience
motherhood and infant feeding in this connected and harmonious manner. For these
mothers, the researchers concluded that the practice of breastfeeding was experienced

instead as a constraint to mothers’ feelings of independence and autonomy.

l7

Indeed, women’s transition into motherhood may introduce new obligations that
require heightened attention to the needs of the infant, which Arafat et al. (1981) have
suggested may occur at the expense of mothers’ relationships with other members of their
social networks. According to these researchers, the apparent incongruence between the
maternal role and mothers’ relationships with other individuals may have implications for
mothers’ infant feeding behavior. In particular, the researchers suggested that mother-
partner attachment may be stronger among mothers that choose to formula feed, while the
mother-infant attachment may be stronger among mothers that choose to breastfeed
(Arafat et al., 1981). However, other research has contradicted the assumption that
motherhood necessarily extracts from mothers’ ability to engage in extended social
networks. Rather, this research has concluded that it is this social network that deﬁnes
motherhood in speciﬁc contexts.

Most notably, Stack (1975/1997) and others (Dominguez & Watkins, 2003;
Kaplan, 1997) have added dimension to these constructions of motherhood by layering
race and ethnicity on interpretations of the maternal role. In Stack’s ( 1975/1997)
ethnographic study of African-American families in an urban community, Stack
recounted constructions of motherhood as largely reliant on extended social networks of
family and friends throughout the community. In this sense, the maternal role was
encompassed in a range of social actors in the mothers’ social network and mothering
responsibilities were shared throughout this community of kin and friends.

Kaplan’s (1997) ethnographic research focused speciﬁcally on African-American
teenage mothers and their constructions of motherhood. For many of the African-

American teenage mothers in he r s tudy, K aplan ( 1997) no ted that m otherhood w as a

18

means of gaining some degree of control over their lives. She explains, “All of the teen
mothers thought motherhood would bring them love and happiness. Some believed
motherhood gave them middle-class stability. Others though it gave them a focus in their
lives, providing them with a sense of belonging and, perhaps, the means to ﬁnd what they
so urgently needed—security and protection” (Kaplan, 1997, p. 181).

Contrary to Stack’s (1975/1997) conclusions, however, Kaplan (1997) reported
that members of these mothers’ extended social networks typically exhibited disapproval
of teenage motherhood. Dominguez and Watkins (2003) reported similar potential for
kin-based conﬂict based on their interviews with ﬁve African-American and ﬁve Latin-
American mothers. In particular, although the researchers cited family and ﬁiends as
potential resources for social support and mobility among African-American and Latin-
American mothers, Dominguez and Watkins (2003, p. 121) also noted that these
relationships held the potential for unpredictability, irregularity, interference, and
exploitation.

Following from this line of ethnographic and qualitative research on constructions
of motherhood, Blum (1999) and Blum and Deussen (1996) focused on infant feeding
behavior in the context of culturally-based constructions of motherhood. In interviews
conducted with 27 white and 26 black middle-class and working-class mothers, these
researchers concluded that mothers’ conceptions of responsible and respectable
motherhood may vary by class and racial background (Blum, 1999; Blum & Deussen,
1996). In particular, Blum (1999) noted that constructions of motherhood among African-
American mothers relied on values of independence, shared parenting with other social

supports, and reciprocity. Conversely, Blum (1999) described greater emphasis on the

19

presence of a male partner to establish a sense of credibility among white, working class
mothers.

Based on the work of these several researchers, it may be deduced that the
complexity of motherhood is necessarily bound to racial and ethnic interpretations and
experiences of motherhood. Based on the work of Blum (1999) and Blum and Deussen
(1996), which is speciﬁc to the topic of infant feeding, it may also be theorized that
constructions Of motherhood, as they intersect with race, may be associated with infant
feeding behavior and experiences.

Sexual F emininity

Yalom (1997) and Traina (2000), among others (Schmied & Lupton, 2001; Shaw,
2003; Young, 1990), have theorized that women’s roles as mothers, wives and partners,
patients, co-workers and academic peers may situate women at the intersection of
numerous expectations about women’s breasts. Transmitted through individuals and
institutions in mothers’ social networks, mothers’ beliefs about the function of breasts
may play an important role in mothers’ infant feeding behavior (Jones, 1987; Stearns,
1999). In particular, the literature suggests that mothers’ beliefs about the function and
symbolic character of women’s breasts may inﬂuence mothers’ orientation toward or
away from breastfeeding, particularly in the context of sexual and maternal constructions
of breasts (Jones, 1987; Stearns, 1999). According to Stearns (1999) and others (Jones,
1987; Young, 1990), within this dichotomy of beliefs about the functions of women’s
breasts, one category of beliefs centers on the sexual breast as a symbol of women’s
attractiveness. The other category of beliefs focuses on the maternal breast as a symbol of

nurturance and natural motherhood. The apparent contradiction between these two

20

categories of beliefs has led several researchers to explore the relationship between these
beliefs and mothers’ infant feeding behavior. The research ﬁndings of these studies have
consistently indicated that mothers’ preference for formula feeding may be related to
sexualized constructions of women’s breasts (Jones, 1987; Maehr, Lizarraga, Wingard, &
Felice, 1993; Steams, 1999).

Stearns’ (1999) conclusions suggest that mothers’ awareness of the sexual
symbolism of women’s breasts may be particularly heightened when breastfeeding in
public or in the immediate presence of others. In interviews with 51 breastfeeding
mothers, Stearns noted that mothers attempted to exercise particular discretion when
breastfeeding in public and negotiated these feedings by ensuring that their bodies, and
particularly their breasts, were carefully concealed throughout the feeding. Discretion
was also important in situations where male family members, acquaintances, and
strangers were present. Based on her results, Stearns (1999, p. 321) argued that “women
actively create the good maternal body before an audience that is more familiar and
comfortable with the sexualized breast than the nurturing breast. The construction of the
good maternal body involves constant vigilance to how the breastfeeding is viewed by
others.” Similarly, in their exploration of women’s experiences breastfeeding in public,
Sheeshka et a1. (2001) noted mothers’ feelings of vulnerability while breastfeeding
publicly at shopping malls and in restaurants. Sheeshka et a1. (2001) reported that
mothers in this study attempted to negotiate these feelings using similar means as the
mothers in Stearns’ (1999) study.

Both of these studies further explored the presence of other supportive individuals

as potential mediators of mothers’ feelings of vulnerability. In light of potentially

21

constraining beliefs about the female body, Stearns concluded that many women enjoyed
experiences of breastfeeding in contexts where they felt greater control over their
surroundings and in situations where they perceived greater support and familiarity with
other individuals present. Sheeshka et a1. (2001) reported similar conclusions but
hastened to add that mothers’ support was derived more particularly from the presence of
other supportive women.

However, Jones (1987) research suggested that mothers’ perceptions of male
partners’ beliefs about the function of breasts may also be inﬂuential in mothers’ feeding
behavior. However, Jones (1987) added complexity to this relationship by suggesting that
women’s beliefs about breasts, and perceptions of their partner’s beliefs, may be class-
based. Based on interviews with 1525 women in a postpartum unit, Jones (1987, p. 83)
reported that working-class women were more likely to believe that breasts were markers
of feminine sexuality and attractiveness. Across categories of class, women in this study
also perceived that their male partners possessed similar beliefs to their own (Jones,
1987). F ollowing t his s tudy, ho wever, there ha 8 b een a r elative s carcity o f additional
research exploring the relationship of class with sexual beliefs.

Age has been studied more extensively as a factor associated with mothers’
beliefs about their breasts and reasons for selecting particular feeding methods. In
particular, adolescent mothers’ infant feeding behavior ha 5 b een c ompared with a dult
mothers’ infant feeding behavior, with particular emphasis on describing differences in
reasons for infant feeding choices between these two age groups (Baisch, Fox, Whitten,
&Pajewski, 1989; B rownell et al., 2002; Grossman et al., 1989; Maehr et al., 1993;

Robinson, Hunt, Pope, & Garner, 1993). Overall, researchers’ conclusions have

22

suggested that the developmental stage of adolescence may cause young mothers to be
particularly inﬂuenced by their own beliefs about sexualized breasts and body image.

Following from these conclusions, researchers have conjectured that adolescent
mothers’ greater propensity to formula feed may be related to their attention to social
norms regarding feminine body image and the s exual s ymbolism o f w omen’s b reasts
(Brownell et al., 2002; Grossman et al., 1989; Maehr et al., 1993; Robinson et al., 1993).
For example, Robinson et al. (1993) surveyed 84 adolescent WIC participants to describe
attitudinal characteristics of breastfeeding and formula feeding adolescent mothers.
Although the study was largely descriptive, results indicated that approximately one-third
(n = 29) of the eighty-four study participants agreed with the statement, “Breastfeeding
makes your breasts sag,” and thirty-six respondents agreed with the statement, “I would
feel embarrassed if someone saw me breastfeed” (Robinson et al., 1993, p. 1312). As a
result, the researchers reasoned that adolescents may be particularly susceptible to
sexualized social beliefs about women’s bodies. Adolescents’ tendency to View breasts as
sexual was also cited in a study conducted by Maehr et a1. (1993), who reported that more
adult expectant mothers cited convenience and naturalness as reasons for their decision to
initiate breastfeeding. The researchers suggested that, due to their stage in development,
“the teen mother may indeed not yet view her breasts as a natural means for infant
feeding” (Maehr et al., 1993, p. 456). Rather, adolescent mothers may be particularly
prone to view breasts as sexual rather than maternal.

The sum of research presented above surrounding maternal and sexual femininity
suggests multiple ways of understanding breastfeeding in the context of knowledge about

functions of w omen’s b reasts a nd r esponsible m otherhood. Ac ross b oth dis courses 0 f

23

femininity, researchers cited mothers’ social networks as key factors in constructing
expectations of maternal and sexual femininity. From this line of research, several
researchers have suggested that members of these social networks and the institutions in
which they are invested may add further dimension to mothers’ feeding decisions
(Giugliani et al., 1994; Guttman & Zimmerman, 2000; Humphreys et al., 1998; Matich &
Sims, 1992). In particular, it has been conjectured that mothers’ infant feeding behavior
may be shaped by expectations of external actors that support and enforce the authority of
various discourses of femininity and various bodies of knowledge surrounding infant
feeding. Indeed, Foucault deﬁnes these bodies of knowledge in the context of power,
wherein facts, beliefs and meanings are constructed to convey authority and inﬂuence
(1977/ 1995; 1978/1990; 1984). In line with this research, additional studies have focused
speciﬁcally on members of these social networks and their inﬂuence on mothers’ feeding
behavior.
Sources of External Inﬂuence

Research has identiﬁed a range of potential actors that may exert inﬂuence upon
mothers’ infant feeding choices. These sources may be categorized by their relationship
with the mother, as actors’ Viewpoints and degrees of inﬂuence may depend on the type
of social relationship each maintains with the mother.

The ecological model of breastfeeding developed by Tiedje et al. (2002) provides
a model for categorizing social actors according to their relationship with the mother. The
model relies on a human ecology perspective, which locates the individual in the context
of external systems of inﬂuence that operate on the individual and on one another (Tiedje

et al., 2002). These systems are often illustrated as concentric circles, wherein the

24

individual is p laced at the c enter 0 f s everal c ompounding external systems. Applying
such an approach to the ﬁeld of breastfeeding, the researchers situate the mother-infant
dyad at the center of the model surrounded by four external contexts that exert inﬂuence
on mothers’ infant feeding choices and experiences. These four external contexts are
family, the healthcare delivery system, community, and the societal/cultural context. For
the current study, three contexts of the ecological model of breastfeeding were used to
categorize potential actors in mothers’ infant feeding decisions: family and friends, the
healthcare delivery system, and community sources.

Identifying Social Actors. Early studies on external social actors focused
primarily on identifying sources of external inﬂuence in mothers’ infant feeding
behaviors. In several of these initial studies, results indicated that mothers may be
reluctant to attribute their infant feeding behaviors to any external actors. In particular,
Arafat, Allen and Fox (1981) asked mothers to identify individuals that decided how her
infant would be fed and whose advice was important in their initial infant feeding
decision. Respondents indicated that, for the most part, the initial infant feeding decision
was In ade b y t he m other a lone w ith minimal e xtemal inﬂuence (Arafat et al., 1981).
Similarly, Jones (1987) asked mothers to provide reasons for initiating their chosen
method of infant feeding. Less than 6% of respondents attributed any part of their
decision to the inﬂuence of their husband, friends and relatives, and medical and nursing
staff (Jones, 1987). However, the results of a study conducted by Bevan, Mosley, Lobach
and Solimano (1984) indicate that this apparent reluctance to attribute infant feeding

behavior to external actors may be moderated by mothers’ chosen method of feeding,

25

wherein mothers that chose to formula feed were more likely to cite “no one” was
inﬂuential in their decision.

The Baby’s Father. Despite indications that mothers’ infant feeding choices are
made independently, several studies discuss the role of the baby’s father in inﬂuencing
mothers’ infant feeding behaviors. The inﬂuence of the baby’s father was even suggested
in the above studies that attributed little inﬂuence to external actors. In comparing
mothers’ emotional relationship with their partner to mothers’ infant feeding behavior,
Arafat, Allen and Fox (1981) suggested that formula feeding mothers may report greater
emotional attachment with their partner than breastfeeding mothers. As well, in exploring
the mothers’ perceptions of partners’ preferences, Jones (1987) noted that mothers’ infant
feeding preferences were typically aligned with the baby’s father if he was reported to
have a deﬁnite preference. In particular, women that chose not to initiate breastfeeding
indicated more partners with perceived preferences for formula feeding, no preference, or
unknown preference (Jones, 1987). In an earlier study, Bevan et al. (1984) cited similar
results among mothers in their sample.

To examine actual preferences of the baby’s father in relation to intended feeding
method, Freed, Fraley and Schanler (1992) surveyed 258 middle- to upper-income
expectant fathers. Their instrument asked fathers to respond to attitudinal items about
perceived advantages and disadvantages of breastfeeding, including nutrition, bonding,
breastfeeding in public, and factors related to respectable motherhood. Further, the
instrument w as unique in a ddressing fathers’ a ttitudes toward p otential e ffects o n the
sexual relationship and partners’ concerns about the mother’s ﬁgure. Fathers that

indicated their partners’ intention to formula feed were reported to have greater concerns

26

about breastfeeding not being natural, being bad for breasts, causing their partner to be
less attractive, and interfering with sex (Freed et al., 1992). Taken together, the results of
this study and those above suggest that there may be an association between partners’
preferences and mothers’ intended feeding behavior.

Additional research has attested to the inﬂuence of partners’ preferences on
mothers’ actual infant feeding behavior. In their study of 200 breastfeeding and formula
feeding mothers, Giugliani et a1. (1994) reported that, regardless of demographic
variables, partners’ preferences were important in mothers’ initiation of breastfeeding.
Similarly, in interviews conducted with 133 mothers and their signiﬁcant others, Kessler,
Gielen, Diener-West and Paige (1995, p. 103) reported that 79% of mothers interviewed
selected the baby’s father as their signiﬁcant other, or the person whose opinions
“mattered the most to her” in her infant feeding choices. The preferences of mothers’
selected signiﬁcant other were signiﬁcantly related to respondents’ intention to
breastfeed, and intention mediated mothers’ initiation and continuation of breastfeeding
(Kessler et al., 1995). The results of a study conducted by Whelan and Lupton (1998)
among a sample of low income, breastfeeding women were similar, wherein continuation
of breastfeeding was associated with partners’ positive or neutral breastfeeding
preferences. Based on these results, partners’ preference for breastfeeding may be a
signiﬁcant factor in mothers’ decision to initiate and continue breastfeeding.

Further research has shifted focus from partner’s preferences to partner’s support,
particularly in relation to mothers’ intended feeding behaviors. In attempting to predict
mothers’ intention to breastfeed, Matich and Sims (1992) explored three categories of

support provided by the baby’s father in relation to mothers’ intention to breastfeed:

27

tangible, emotional and informational. Across all three categories of support, mothers that
intended to breastfeed indicated greater support from the baby’s father than mothers that
intended to formula feed (Matich & Sims, 1992). Two of these categories of support were
signiﬁcant in a longitudinal study of mothers’ interpersonal relationships and feeding
choices conducted by Isabella and Isabella (1994). These researchers reported that across
all sources of social support, mothers indicated receiving the greatest emotional and
instrumental support in their feeding choices from their partner. Conversely,
informational support was strongest from doctors and nurses (Isabella & Isabella, 1994).

Nonetheless, a study conducted by Humphreys, Thompson and Miner (1998)
suggests that the baby’s father may indeed be an inﬂuential source of informational
support. In their study of 1,001 low-income, expectant mothers, Humphreys et a1. (1998)
reported that hearing about breastfeeding beneﬁts from a range of sources (the baby’s
father, family members, and lactation consultants) was positively associated with
intention to breastfeed. After controlling for previous breastfeeding experience, however,
only the relationship between informational support from the baby’s father and intention
to breastfeed remained constant across groups (Humphreys et al., 1998). Mothers’
intention to breastfeed may therefore necessarily rely on all three types of support from
the baby’s father: emotional, tangible and informational.

In addition to inﬂuencing mothers’ intention to breastfeed, support ﬁ'om the
baby’s father has been explored in relation to breastfeeding duration with mixed results.
Isabella a nd Isabella ( 1994) indic ated that m others w ho e xpressed g reater s atisfaction
with support they received from the baby’ father also continued to breastfeed their infant

for longer periods of time. However, in the development of the Breastfeeding Attrition

28

Prediction Tool (Janke, 1994), support from the baby’s father failed to load at the 0.30
level on Social and Professional Support Scale. Results from these two analyses suggest
an elusive relationship between support from the baby’s father and breastfeeding
duration.

Other Family and Friends. The inﬂuence of other family and ﬁiends on mothers’
infant feeding behaviors is supported in the literature. Dix (1991) indicated that support
from family and friends was one of the most inﬂuential factors in mothers’ selection of
either feeding method. Giugliani et a1. (1994, p. 161) developed similar conclusions in
their study of social support among breastfeeding and formula feeding mothers. The
researchers added that the inﬂuence of family and friends on mothers’ feeding decisions
held constant across demographic and cultural factors such that “decisions about infant
feeding are a function of familial and group attitudes, regardless of maternal age, ethnic
group, educational level, and marital status.”

Mothers’ degree of perceived support from family and friends may vary by
mothers’ infant feeding behavior. In particular, several studies indicate that breastfeeding
mothers report greater support from family and friends than formula feeding mothers.
Grossman et al. (1990) noted the breastfeeding mothers received greater support from the
grand(mother) and friends than formula feeding mothers. Examining the inﬂuence of
these two sources on breastfeeding duration, Whelan and Lupton (1998) reported that
continuation of breastfeeding was positively associated with support from these sources.

However, while the above research suggests that the inﬂuence of family and
friends may be positive, Guttman and Zimmerman’s (2000) study of 154 predominantly

low income mothers indicated positive support was more characteristic only among

29

mothers that chose to breastfeed. Their results suggested that friends may be one of the
most disapproving sources of social support for breastfeeding, particularly by expressing
concern about mothers’ ability to maintain their social lifestyle. Family members were
also cited as a source of breastfeeding discouragement when the family member was at
least partially responsible for care of the infant (Guttman & Zimmerman, 2000).

Healthcare Delivery System. As an authority on infant and maternal health, the
healthcare delivery system and medical discourses on femininity may exercise particular
inﬂuence on mothers’ infant feeding behavior. On a national level, breastfeeding has
been promoted by the medical community, health professionals, and federal departments
to promote breastfeeding as a national health objective. The Healthy People initiative,
which commenced in 1979 as a national planning scheme to improve public health and
medicine, espouses maternal and infant health as a major national health objective (U .S.
Department of Health and Human Services (HHS), 1979, 1990, 2000a). Each report has
underscored the importance of breastfeeding for infant health outcomes and objectives of
the most recent report, Healthy People 2010, aim for 75% breastfeeding rate among
mothers in the early postpartum period (an 11% increase from 1998), 50% continuing to
breastfeed at six months (a 21% increase from 1998), and 25% continuing to breastfeed
at one year (a 9% increase from 1998) (US. Department of Health and Human Services
(HHS), 2000a, p. 16-46).

These goals have been adopted by the healthcare delivery system, which may
maintain contact with mothers during prenatal care and education, the delivery, and
postpartum care and counseling. Nonetheless, despite potentially numerous contacts

during pregnancy and in the postpartum period, several studies indicate that the inﬂuence

30

of health professionals on mothers’ decision to initiate breastfeeding may be minimal. In
interviews with 81 recent postpartum women at an urban hospital, Dix (1991) concluded
that the inﬂuence of health professionals on respondents’ decision to initiate
breastfeeding or bottle feeding was not as strong as the respondent’s family and other
social inﬂuences. This result has been supported by several other studies examining
social support and inﬂuence of actors in the healthcare delivery system on mothers’
initial feeding decision (Giugliani et al., 1994; Sullivan & Jones, 1986).

In order to test this relationship, further research on the healthcare delivery system
has moved in the direction of categorizing support provided by health professionals in
relation to mothers’ infant feeding behavior. In particular, studies have focused
speciﬁcally on informational support provided by health professionals and mothers’
subsequent decisions. Using this approach, Isabella and Isabella’s (1994) longitudinal
study of mothers’ interpersonal relationships and feeding choices concluded that, indeed,
informational support was strongest from mothers’ doctors and nurses. Mothers’
attendance in classes also provided women with greater emotional and informational
support and was a predictor Of mothers’ intention to breastfeed (Matich & Sims, 1992).

To examine whether the relationship between informational support and intention
to breastfeed translates into actual behavior, McLeod, Pullon and Cookson (McLeod,
Pullon, & Cookson, 2002) administered surveys to 490 expectant mothers just prior to
delivery, which asked them two questions related to informational support and guidance
they received or desired. Almost one-third of these women (31%) believed that they

needed more information about breastfeeding. Results of data analysis indicated that

31

women that expressed a need for greater informational support were less likely to be
breastfeeding at six to ten weeks postpartum (McLeod et al., 2002).

In response to the apparent inﬂuence of informational support on infant feeding
choices, some health care providers and clinics have implemented peer support programs
to enhance mothers’ social support network. Several researchers have employed
experimental design to evaluate the effectiveness of these programs on aspects of infant
feeding choices, particularly with reference to initiation and duration of breastfeeding. I
Most of these program evaluations indicated that participation in peer support programs
increased rates of breastfeeding initiation and duration, and may have enhanced women’s
ability to reach their infant feeding goals (Dennis, Hodnett, Gallop, & Chalmer, 2002;
Pugh, Milligan, & Brown, 2001; Quarles, Williams, Hoyle, Brimeyer, & Williams, 1994;
Schafer, Vogel, Viegas, & Hausafus, 1998; Shaw & Kaczorowski, 1999). Thus, although
the general inﬂuence of the healthcare delivery system on mothers’ infant feeding
behavior may be minimal in the context of other social actors (Dix, 1991; Giugliani et al.,
1994; Sullivan & Jones, 1986), the particular provision of informational support by health
providers may be inﬂuential.

Community Sources. Community expectations of mothers’ feeding behavior may
be expressed in a number of public outlets, including general public places, work and
school, and through the media. In general public places, Guttman and Zimmerman (2000)
suggested that mothers’ perceptions of community norms may affect initiation and
duration of breastfeeding. Study participants were asked about to recall experiences when
they witnessed other women breastfeed in public; of the women that had witnessed such

an event, half of all breastfeeding mothers and 40% of formula feeding mothers perceived

32

negative public reactions to the breastfeeding mother. Among the breastfeeding mothers,
the researchers speculated that the apparent discrepancy between personal feelings
toward the experience and perceived social norms may cause breastfeeding to be a
constrained experience for these mothers, particularly in public (Guttman & Zimmerman,
2000).

To examine actual public attitudes toward breastfeeding, Li, Fridinger and
Grummer-Strawn (2002) analyzed 12 items related to breastfeeding that appeared on the
Healthstyles 2000 national mail survey. Surveys were administered to men and women
and asked study participants to indicate their perceptions of mothers’ ability to maintain
their lifestyle while breastfeeding, appropriateness of breastfeeding in public, proper age
at which to wean a baby, and potential physical complications for the mother. Negative
public perceptions were greatest on issues of mothers’ ability to maintain her lifestyle
(45% agreed that breastfeeding mothers must give up too many lifestyle habits),
breastfeeding in public (27% felt that breastfeeding in public was embarrassing for the
mother), and appropriate age to wean the infant (31% believed babies should be fed
cereal or baby food by 3 months; 31% felt that one-year-old children should not be
breastfeed by their mother) (Li et al., 2002). Results indicate that mothers’ infant feeding
choices may be made in the c ontext o f p otentially c onstraining p ublic p erceptions o f
breastfeeding.

On a different level of community inﬂuence, several researchers have examined
the speciﬁc context of the workplace in relation to mothers’ feeding behavior. Similar to
the approach above, Seijts (2002) explored public perceptions of organizational

attractiveness based on policies and accommodations made to breastfeeding mothers.

33

Organizations that made a ccommodations to b reastfeeding m others w ere p erceived a s
more attractive and more fair; study participants also indicated that they would be more
likely to apply to and accept jobs from organizations perceived as more accommodating
(Seijts, 2002). Nonetheless, employers’ perception of value added to their organization
through such measures may be less favorable. In their sample of 85 employers and their
attitudes toward breastfeeding, Libbus and Bullock (2002) reported that although
employers tended to be willing to facilitate an environment open to breastfeeding, many
also indicated that they saw little value added to their organization through providing
such accommodations. Further, a large percentage of the employers indicated they had
not had experience with breastfeeding and, hence, had not personally made
accommodations to breastfeeding mothers in the workplace.

The results of this study suggest that workplace supervisors may be willing to
accommodate breastfeeding mothers, but that actual workplace accommodations may still
be lacking. As a result of this workplace environment, Lindberg (1996) theorized that
social expectations of the good mother and qualities of the productive worker may induce
role conﬂict for breastfeeding mothers. Lindberg’s analysis of secondary data indicated
that mothers were more likely to wean their infant during the month they commenced
working. Among mothers that entered employment, women employed in part-time
positions were more likely to breastfeed and to continue to breastfeed through
employment than women employed in full-time positions (Lindberg, 1996). Thus, the
literature suggests that there may be a constraining relationship between maternal

employment and breastfeeding, potentially inﬂuence mothers’ initial and subsequent

34

feeding behaviors. This relationship may or may not hold for mothers’ school
environment, which was not discussed in the above studies.

Interacting Factors. Several studies have explored the effects Of mothers’
demographic and cultural characteristics on sources of external inﬂuence. Grossman,
Fitzsimmons, Larsen-Alexander, Sachs and Harter (1990) interviewed 220 new mothers
to describe and compare characteristics of low and upper income mothers that chose to
breastfeed or formula feed. Overall, breastfeeding mothers reported greater support in
their infant feeding decision from all sources, including the baby’s father, their mother,
ﬁiends, health professionals, books and television, and advocacy/education groups. By
income, the social support patterns of low income breastfeeding mothers more closely
resembled upper income breastfeeding mothers—who reported greater social support
from all sources—than low income formula feeding mothers (Grossman et al., 1990).

Similarly, ethnicity has been explored in relation to mothers’ social support
networks. Baranowski et al. (1983) explored the relationship between breastfeeding
initiation and sources of social support across three ethnic groups: African-American,
Mexican-American, and Anglo-American. Results indicated that the most important
source predicting African-American mothers’ decision to initiate breastfeeding was a
close friend. Among Mexican-American mothers, the (grand)mother was most important
and among Anglo-American mothers, the male partner predicted most of the variance
(Baranowski et al., 1983). These results suggest the potential inﬂuence of speciﬁc
demographic and cultural values in identifying sources of external inﬂuence and

predicting mothers’ infant feeding behavior.

35

 

In sum, the above literature suggests a range of discourses on femininity that may
be transmitted through various social actors, and which may inﬂuence mothers’ infant
feeding behavior. The apparent overlap between and contention among discourses of
femininity—both maternal and sexual—suggests that mothers’ infant feeding choices
may exist at the intersection of personal beliefs, discourses of femininity, and socio-
demographic factors (Blum, 1999; Blum & Deussen, 1996; Carter, 1995; Kaplan, 1997;
Maher, 1992; Murphy, 2000; Steams, 1999). Adding to this picture, the inf luenceof
external actors and the institutions in which they are invested may also inﬂuence
mothers’ feeding behavior (Baranowski et al., 1983; Giugliani et al., 1994; Grossman et
al., 1990; Grossman et al., 1989; Guttman & Zimmerman, 2000; Humphreys et al., 1998;
Janke, 1994; Matich & Sims, 1992). Based on this line of research, the ways in which
these discourses and actors interact with one another to predict mothers’ feeding behavior
formed the focus of the current investigation. To this end, Chapter 3 outlines study
participants, instrumentation, and procedure that aided in the accomplishment of the four

study objectives outlined in Chapter 1.

36

Chapter 3
METHODS
Study Purpose

To review, the purpose of this study was to explore factors associated with infant
feeding choices among low-income women. In particular, this study aimed to accomplish
four research objectives. First, this study intended to describe characteristics of a sample
of women that participated in the Special Supplemental Nutrition Program for Women,
Infants, and Children (WIC) in two urban areas in Michigan and aimed to explore how
these characteristics compared across geographic location.

Second, this study was designed to test measurement reliability of two subscales
of the Maternal Adjustment and Attitudes Questionnaire (Kumar et al., 1984) and two
scales of the Breastfeeding Attrition Prediction Tool (Janke, 1994) among this sample of
WIC participants. Two subscales of the Maternal Adjustment and Maternal Attitudes
Questionnaire were selected for this objective: Attitudes toward Motherhood and the
Baby and Body Image (Kumar et al., 1984). The Negative Breastfeeding Sentiment and
Positive Breastfeeding Sentiment Scales from Janke’s (1994) Breastfeeding Attrition
Prediction Tool were also assessed for this study objective.

Lastly, this study was designed to examine the extent to which mothers’ infant
feeding behavior was inﬂuenced by social contexts in which mothers made their feeding
choices. Two models were developed to predict infant feeding behavior on the basis of
two 5 ocial c ontexts. T he first m odel, w hich formed the third 0 bjective o f t he current
investigation, aimed to explore mothers’ infant feeding behavior in relation to

individuals/groups in the mother’s social sphere. For this objective, three systems of

37

external inﬂuence were examined: family and friends, the healthcare delivery system, and
community sources.

The fourth research Objective aimed to explore mothers’ infant feeding choices in
relation to maternal attitudes toward motherhood and the baby and maternal body image.
These constructs were used to operationalize mothers’ perceptions of maternal and sexual
femininity. Both models of mothers’ infant feeding behaviors necessarily included factors
of race/ethnicity, education, and mothers’ beliefs about infant feeding methods as control
variables.

Study Participants

The sample for this study was drawn from women that participated in the Special
Supplemental Nutrition Program for Women, Infants, and Children (WIC) in two urban
areas of Michigan during the months of June and July 2003. Questionnaires were
administered to WIC participants that met the eligibility criterion and indicated a
willingness to participate in the study. Eligibility to participate was determined by the
amount of time since the woman’s last live birth. In particular, women were considered
eligible if they had delivered a live birth within the previous three months. This time
period was selected following conversations with WIC administrators from the Michigan
WIC Division, who indicated that although they advised exclusive breastfeeding for a
period longer than three months, they acknowledged introduction of solids typically
occurred among their clients by six months postpartum.

The study aimed for a sample of 100—200 study participants from each location. In
order to establish a sampling frame for distributing surveys to WIC clinics in each urban

location, WIC client ﬁgures were obtained from the Kent County Health Department and

38

Detroit Health Department that reported the total number of active WIC clients in
April/May 2003 by clinic. These ﬁgures accounted for all WIC participants with active
enrollment in the speciﬁed month, including pregnant women, breastfeeding and non-
breastfeeding postpartum mothers, infants, and children. Based on these ﬁgures, a
sampling frame was established to distribute surveys proportionally by the number of
active WIC clients at each clinic and within each WIC agency (urban location). The

sampling frame for both urban locations is presented in Table 1.

Table 1: Sampling Frame by WIC Agency and Clinic

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

April/May 2003 Proportion of Total . .
Count” WIC Client Group Surveys Distributed
Kent County Health Dept. 16,378 39.6% 200
Clinic K1 2,879 7.0% 35
Clinic K2 5,152 12.5% 63
Clinic K3 670 1.6% 8
Clinic K4 2,343 5.7% 29
Clinic K5 2,375 5.7% 29
Clinic K6 591 1.4% 7
Clinic K7 2,368 5.7% 29
Detroit City Health Dept. 24,964 60.4% 315
Clinic D1 3,092 7.5% 38
Clinic D2 3,655 8.8% 44
Clinic D3 3,716 9.0% 56
Clinic D4 2,185 5.3% 27
Clinic D5 3.095 7.5% 38
Clinic D6 3.379 8.2% 41
Clinic D7 2,291 5.5% 28
Clinic D8 1,673 4.0% 20
Clinic D9 1,878 4.5% 23
Both Locations 41,342 100.0% 515

 

 

 

 

A All women, infants and children with active enrollment in WIC in the speciﬁed month

Although the study aimed for a sample of 100-200 study participants from each
location, larger quantities of questionnaires were distributed to WIC clinics to account for

incomplete questionnaires or refusal to participate. As indicated in Table 1, 200

39

 

questionnaires were distributed to Kent County WIC clinics and 315 questionnaires were
distributed to Detroit clinics. Initial survey administration procedures dictated that these
questionnaires be administered to clients at the sixteen participating WIC clinics for a
period of three weeks commencing in mid-June 2003. This decision was made following
discussions with staff at each location, who indicated that three weeks would be an
appropriate length of time to recruit the desired number of study participants from each
location. However, after three weeks of survey administration, it was determined through
conversations with WIC administrators at each urban location that an additional four
weeks of questionnaire administration was needed to obtain the desired number of
completed questionnaires.
Study Design and Procedure

Data analysis for the current investigation involved the merging of two sets of
data on study participants’ feeding behavior. The primary data set contained data
collected from questionnaires administered to WIC clients at sixteen clinics throughout
Kent County and Detroit. The secondary data set was extracted from WIC participant
records maintained by the WIC Division 0 f t he Michigan D epartment o f C ommunity
Health (MDCH). This data set contained all active WIC participants enrolled in the
sixteen participating clinics as of June 24, 2003. Primary and secondary data sets were
merged using WIC identiﬁcation numbers, which are unique family numbers assigned to
women and their children in order to maintain participant records at the local and state
level. This required that study participants write their WIC identiﬁcation number, and no
other identiﬁers, on the cover of their questionnaire in order to match the corresponding

secondary data record. A detailed description of data collection procedures follows,

40

including primary data collection procedures by urban location and details on the
secondary data extraction.

Primary Data Collection. Questionnaires were administered to clients at seven
clinics managed by the Kent County Health Department and nine clinics that operated
under the Detroit City Health Department for a period of seven weeks in June and July
2003. The target sample size for the study was 100-200 study participants from each
location to facilitate comparison between locations and draw conclusions about the
sample. Following discussions with staff at each location, three weeks was determined to
be an appropriate length of time to recruit the desired number of study participants.
However, based on the number of completed surveys collected following three weeks of
survey administration, a cooperative decision was made with WIC administrators at each
site to extend the data collection timeline by four additional weeks.

Discussions with staff at each location determined procedure for administering
questionnaires to WIC clients. Separate administration procedures were constructed for
each location to maximize the researcher’s use of available resources and to minimize
strain on WIC staff. As a r esult, p rocedures for a dministering que stionnaires differed
between the Kent County Health Department WIC Agency and the Detroit City Health
Department WIC Agency. Questionnaire administration procedures for each location are
described in detail below.

Surveys were administered to clients at seven WIC clinics managed by the Kent
County Health Department. Following discussions with staff at this location, it was
decided that questionnaires would be administered to WIC clients by members of the

Breastfeeding Subcommittee of the Kent County Health Department WIC Agency. This

41

subcommittee was composed of twelve members that counseled women at clinics
throughout the Kent County WIC Agency. The group was selected to administer the
survey based on two considerations. First, the subcommittee indicated particular interest
in information that can be gleaned from the study and was therefore willing to commit
resources to the project. This also minimized strain on other WIC clinic staff. Second,
following several grammatical revisions of the instrument, WIC staff remained uncertain
about study participants’ ability to read and understand portions of the questionnaire. As
a result, the Breastfeeding Subcommittee volunteered time to assist in distributing
questionnaires to WIC clients that met study criteria. Members of the committee
approached potential study participants while they were waiting for their scheduled
appointment to offer eligible women the Opportunity to participate in the study.
Questionnaires were also administered to WIC clients at nine clinics managed by
the Detroit City Health Department. The procedure for administering surveys in this
second study location differed slightly from Kent County distribution procedures.
Following discussions with WIC coordinators and nutritionists in this department, it was
decided that eligible study participants would be offered the opportunity to participate in
the study by WIC clinic clerks. This group of WIC staff was been selected to distribute
questionnaires because W1C clients were required to make contact with this individual
upon checking in and before leaving the clinic. As a result, staff felt these individuals
could efﬁciently recruit study participants during the clinic check-in process. Upon
checking in the with clinic clerk, WIC clients were asked to complete necessary internal

paperwork to maintain their WIC certiﬁcation. After completing this paperwork and

42

returning forms to the clinic clerk, eligible WIC participants were offered the opportunity
to participate in the study.

WIC staff at both locations were instructed to recruit study participants from all
WIC clients that met eligibility criteria for the study. Eligibility to participate was
determined by the amount of time since the woman’s last live birth. In particular, women
were considered eligible if they had delivered a live birth within the previous three
months. This time period was been selected following conversations with WIC
administrators from the Michigan WIC Division, who indicated that they advise
exclusive breastfeeding through a longer period of time but acknowledge introduction of
solids typically occurs by six months. Further, although previous research has extended
this period of time to a maximum of twelve months postpartum (Bevan et al., 1984), a
shorter time period was chosen for this study to enhance women’s ability to recall
information from their third trimester of pregnancy, since their baby was born, and at the
time of survey administration. In an attempt to obtain a sample of clients that resembled
the diversity of WIC populations at each clinic, no other selection criteria was set (i.e.,
age, race/ethnicity, etc.)

Informed consent was obtained from study participants at all sites. During their
initial explanation of the project, WIC staff at each location informed clients that their
participation in the study was completely voluntary and that their decision to participate
or not to participate would not affect the services and/or beneﬁts they received from the
WIC program. Women that agreed to participate were be asked to read, sign and return a
consent form that outlined the purpose of the study, rights as a study participant,

incentives, and potential risks and beneﬁts of participating. An additional copy of the

43

consent form was made available for study participants’ records. After returning a signed
consent form, study participants were asked to complete their questionnaire in the clinic
waiting area while they were waiting for their scheduled appointment. Questionnaires
were expected to take approximately twenty minutes to complete. WIC staff that
administered the questionnaires at each location indicated that the questionnaire typically
took at least this amount of time to complete.

As an incentive for participating in the study, participants were offered the
opportunity to be entered into a drawing for one Of eight $25 gift certiﬁcates to a local
grocery store. Participants that were interested in being included in this drawing were
asked to provide their contact information on a separate card included with the
questionnaire. The card was clipped to the front of the questionnaire in a manner that
allowed the study participant to detach their contact information from their completed
questionnaire. Contact information was not recorded on any data set and was only
maintained in hard copy. Following the incentive drawing, all prize entry forms were
destroyed.

In order to protect the conﬁdentiality of respondents’ information, study
participants were asked to place their completed questionnaires and prize drawing entry
slips in a clearly marked, locked box in the clinic waiting room. WIC staff was instructed
to place signed consent forms in the same collection box. Questionnaires, prize entry
drawing slips, and consent forms were collected from WIC clinics by the researcher at
intervals throughout the seven weeks of survey administration and upon completion of

the study.

44

 

Secondary Data Collection. Respondents were asked to write their WIC
identiﬁcation number, a unique number assigned to each WIC client and known only to
WIC administrative staff and the client, on the cover of their survey. This number was
used to match an extraction of secondary data from records maintained by the Michigan
WIC Division to respondents’ answers. Once the primary and secondary data sets were
linked, the WIC identiﬁcation number was removed from the data set and a unique
identiﬁcation number was assigned for each respondent. The code sheet of WIC
identiﬁcation numbers and the new identiﬁcation numbers remained locked in a secure
location and were destroyed upon the completion of the study.

Study Instrument

This study collected primary data from questionnaires administered to WIC
clients and drew on secondary data from the WIC Division of the MDCH to accomplish
the four study objectives. In particular, primary and secondary data were used to explore
characteristics that described the study sample and to assess the degree to which the
sample was representative of other WIC clients at the same urban location, to compare
across geographic location, to test measurement reliability of several existing scales and
subscales included in the questionnaire, and to predict mothers’ infant feeding behavior
based on her social contexts.

To accomplish these objectives, the questionnaire included measures of
respondents’ current infant feeding behavior, systems of external inﬂuence, body image
during pregnancy, attitudes toward motherhood and the baby since the baby was born,

beliefs about breastfeeding, and socio-demographic characteristics. Measurement of each

45

set of characteristics is described below, followed by a discussion of the secondary data
that was obtained from the WIC Division of the MDCH.

Current Infant Feeding Behavior. Mothers’ current infant feeding behavior was
calculated as a breastfeeding percentage, wherein the number of times respondents
currently fed their infant breastmilk in a typical day was divided by the total number of
times they fed their infant in a typical day. Based on this equation, breastfeeding
percentage may range from zero (0% breastfeeding) to one (100% or exclusive
breastfeeding). In models predicting infant feeding behavior for objectives 3 and 4 of the
investigation, breastfeeding percentage ﬁrnctioned as the dependent variable in data
analysis.

In response to research that suggests that the experience of pumping breastmilk
may inﬂuence mothers’ beliefs and infant feeding behavior (Blum, 1999), respondents
were asked to distinguish between how many times they feed their infant pumped
breastmilk in a typical day and how many times they feed their infant breastmilk from the
breast in a typical day. These measures were collapsed into one in the computation of
breastfeeding percentage. However, where applicable, results on mothers’ use of a breast
pump and frequency of pumping breastmilk are reported in Chapter 4.

Also related to computation of the breastfeeding percentage, it was decided that
the current investigation would be limited to an examination of two feeding methods:
breastfeeding and formula feeding. This decision was made based on conversations with
WIC staff and administrators, wherein it was concluded that the sample of mothers for
the current study was intentionally limited to three months postpartum to involve women

that had not yet introduced solid foods. However, for the purpose of comprehensively

46

exploring mothers’ feeding behaviors, study respondents were asked whether they had
fed their baby anything other than breastmilk or formula. This characteristic was explored
as a descriptive characteristic only, and was not included as an independent variable
predicting breastfeeding percentage.

Several other characteristics of mothers’ infant feeding behavior were also
measured and reported for descriptive purposes. Timing of mothers’ initial infant feeding
decision was explored by asking respondents to indicate when they ﬁrst made their
decision to or not to breastfeed. Five response categories were offered to respondents:

99 ‘6'

In the second 3

9, 66'

“before I became pregnant, In the ﬁrst 3 months of my pregnancy,
months of my pregnancy,” “in the third 3 months of my pregnancy,” and “after my baby
was born.” These ﬁve categories were later collapsed into three in o rderto facilitate
comparisons using Chi-Square analysis. These three categories included before
pregnancy, during pregnancy, and after birth.

Initiation of breastfeeding was measured by respondents’ answer to the question:
“Have you ever—at any time—fed your baby breastmilk?” Respondents that answered in
the affirmative were considered to have initiated breastfeeding, whereas respondents that
indicated “no” were considered not to have initiated breastfeeding. Additionally, WIC
secondary data was available on mothers’ initiation of formula feeding for a portion of
the study sample. This characteristic will be described in greater detail below.

Lastly, duration of breastfeeding was measured as the number of weeks between
the infant’s date of birth and the time at which the infant was weaned from breastmilk. If

the infant was still breastfeeding at the time of survey administration, duration of

breastfeeding was calculated as the amount of time between the infant’s date of birth and

47

the completion of the current investigation (dated as August 22, 2003). There were three
collapsed categories of breastfeeding duration that encompassed 0 through 10 days, 11
through 31 days, and longer than one month. Most often, mothers that were still
breastfeeding at the time they completed the survey were computed as having breastfed
longer than one month.

Systems of External Inﬂuence. The format of the Social and Professional
Support Scale of the Breastfeeding Attrition Prediction Tool (Janke, 1994) was used to
measure the inﬂuence of external actors in respondents’ infant feeding choices.
Development and testing of this tool, in addition to a comparison with other tools in the
literature, is described in detail in Appendix A. The scale measured respondents’
perceptions of others’ expectations and the value respondents place on others’ opinions
on three systems of external inﬂuence: family and friends, the healthcare delivery system,
and community sources.

The inﬂuence measure was computed as a product of two separate measures on
the same source of inﬂuence: mother’s perception of a breastfeeding expectation from
each source and mother’s motivation to comply with their opinions. Perception of a
breastfeeding expectation was measured by asking respondents to rate their level of
agreement with the following statement: “When I ﬁrst decided how to feed my baby,
these people though that I should deﬁnitely breastfeed by baby...” Motivation was
measured by asking respondents to indicate their level of agreement with the statement:
“I value these people’s opinions on how I should feed my baby...” On both measures,
respondents were presented with a ﬁve-point Likert scale that ranged from strongly

disagree to strongly agree, with an Option to select “don’t know or does not apply.”

48

J anke (1994) suggested computing the inﬂuence score for each source of external
inﬂuence as a product of the expectation score and the value score. According to J anke’s
method of computation, respondents were presented with response categories that ranged
from strongly disagree (coded as a 1) to strongly agree (coded as a 5). The inﬂuence of
each source could therefore range from 1 to 25, indicating degrees of support f oror
discouragement of breastfeeding.

Alternatively, for the current investigation, coding on the expectation and value
scores for each source Of external inﬂuence differed in order to compute an inﬂuence
score that suggested strength and direction of infant feeding inﬂuence toward
breastfeeding or toward formula feeding. T herefore, e xpectation r anged from s trongly
disagree (coded as a -2) to strongly agree (coded as a 2) to indicate the source’s
expectation toward formula feeding or toward breastfeeding. The expectation score was
then multiplied by the value score (coded from 1 to 5) to indicate the degree to which the
respondent valued the source’s perceived orientation toward breastfeeding or toward
formula feeding. Based on this computation, inﬂuence scores for each source could range
ﬁ'om -10 (strong inﬂuence in the direction of formula feeding) to +10 (strong inﬂuence in
the direction of breastfeeding).

The original Social and Professional Support Scale contained twelve potential
sources of inﬂuence. Five sources from the original instrument were directly translated
onto the questionnaire for the current study. Four additional sources from the original
instrument were translated onto the questionnaire, but were collapsed into categories on
the current instrument. In particular, Janke’s (1994) original sources of “other relatives”

and “my sister” were embedded in “my family” for the current study. Additionally, “the

49

baby’s doctor” and “my doctor” were collapsed on the current instrument to read “my
doctor(s).” Following discussions with WIC administrators and staff, the last three
sources from the original instrument were determined to be considerably broad for the
purposes of this study. In particular, it was determined that “my childbirth educator” and
“La Leche League” should instead be made speciﬁc to this population of WIC clients by
including two items speciﬁc to the WIC program: “WIC staff” and “Breastfeeding Peer
Counselors.” The last item, “people important to me,” was removed owing to its vague
detail about the relationship of these individuals to the mother.

AS a result, other sources that were included on the current instrument that were
developed following discussions with WIC administrators and staff and review of social
support and inﬂuence literature (Dennis & Faux, 1999; Dusdieker, Booth, Seals, &
Ekwo, 1985; Humphreys et al., 1998; Janke, 1994; Matich & Sims, 1992). In total,
fourteen sources were included in the questionnaire. Six sources were originally placed in
the family and friends system of external inﬂuence, including the baby’s father’, my
mother’, my mother-in-law’, my family, my best friend*, and my other friends.
Following discussions with a n e xpert p anel, o ne 3 dditional s ource w as p laced in t his
group that was originally placed in the healthcare delivery system of inﬂuence. This
source, Breastfeeding Peer Counselors, was relocated to the family and friends system of
external inﬂuence owing to the personal, peer-type relationship developed between
breastfeeding peer counselors and mothers. This relationship was also maintained in the
literature (Arlotti, Cottrell, Lee, & Curtin, 1998; Dennis et al., 2002; Pugh et al., 2001),

thereby supporting the decision to move the source. The healthcare delivery system was

 

° Item translated directly from the Social and Professional Support Scale of the Breastfeeding Attrition
Prediction Tool (Janke, 1994).

50

measured in three items: my doctor(s), my midwife’, and WIC staff. Lastly, the
community sources system of external inﬂuence included four sources: other people at
my school, other people at my work, my community, and people in public places.

Mean scores of each system of external inﬂuence were computed by taking the
mean of inﬂuence scores for all sources in the system. Mean scores for each system of
external inﬂuence functioned as a separate independent variable predicting mothers’
infant feeding behavior for study objectives 3 and 4.

Maternal Body Image. Maternal body image was measured using the Body
Image subscale of Kumar et al.’s Maternal Adjustment and Maternal Attitudes Scale
(1984). An in-depth review of the development of this instrument and a comparison to
other scales in the literature is located in Appendix A. To summarize, the subscale
contains twelve items that ask respondents to consider aspects of their body image during
their third trimester of pregnancy. Sample items include: “I have liked the shape of my
body,” “I have felt attractive,” and “I have felt proud of my appearance.” Response
categories were b ased o n a f ive-point L ikert s cale t hat r anged from s trongly dis agree
(coded as a 1) to strongly agree (coded as a 5). In order to extract factors of body image,
this subscale was factor analyzed as part of study objective 2. Mean scores for each factor
of body image were computed by taking the mean item ranking of all items included in
the extracted factor. Subsequently, factors of body image were analyzed as independent
variables in data analysis for study objectives 3 and 4.

Maternal Attitudes toward Motherhood and the Baby. The Attitudes toward
Motherhood and the Baby subscale of the Maternal Adjustment and Maternal Attitudes

Scale (Kumar et al., 1984) was used to measure respondents’ orientation toward

51

motherhood and the maternal role. Once again, a detailed description of the development
of this instrument and a comparison with other scales in the literature is located in
Appendix A. As an overview, this subscale contains twelve items that ask respondents
about their attitudes toward pregnancy, motherhood, and their baby since their baby was
born. Example items include: “I have been wondering whether my baby will be healthy
and normal,” “I have been worrying that I might not be a good mother,” and “The
thought of having more children has appealed to me.” Response categories were based on
a ﬁve-point Likert scale that ranged from strongly disagree (coded as a l) to strongly
agree (coded as a 5). As with the Body Image subscale, items on the current subscale
were factor analyzed to extract factors of attitudes toward motherhood and the b aby.
Mean scores were then computed by factor by taking the mean of item ratings of all items
that loaded on the factor. Mean factor scores were used as independent variables in
predicting mothers’ infant feeding behavior for study objectives 3 and 4.

Beliefs about Breastfeeding. Two additional scales of the Breastfeeding Attrition
Prediction Tool (Janke, 1994) measured respondents’ beliefs about breastfeeding: the
Positive Breastfeeding Sentiment Scale and the Negative Breastfeeding Sentiment Scale.
These subscales measured respondents’ beliefs in the advantages and the disadvantages
of breastfeeding. A detailed de scription o f instrument de velopment a nd a c omparison
with other instruments in the literature can be found in Appendix A.

On the original instrument, respondents’ beliefs were one of two components
measuring respondents’ attitudes toward breastfeeding. In particular, attitude was a
product of two measures: belief and outcome evaluation. Belief was measured by asking

respondents to indicate their level of agreement with each of 29 statements about

52

advantages and disadvantages of breastfeeding. Janke (1994) measured outcome
evaluation by asking respondents to rate the importance of each advantage or
disadvantage in choosing a feeding method. For example, a matched belief and outcome
evaluation pairing from the instrument included “Breastfeeding is more convenient than
formula feeding” (belief) and “Using a feeding method that is convenient is important to
me” (outcome evaluation). Both items were rated on Likert-type scales that ranged from
strongly disagree to strongly agree. The product of two complimentary measures formed
the respondent’s attitude on the item; the sum of all attitudinal items on each subscale
indicated the respondent’s positive or negative breastfeeding sentiment score. A higher
score on either scale indicated greater positive or negative breastfeeding sentiment.

The Positive and Negative Breastfeeding Sentiment Scales contained a total of 29
belief items related to the advantages and disadvantages of breastfeeding and 29
corresponding outcome evaluation items. One additional item was added to the
instrument in response to literature that suggests women encounter signiﬁcant constraints
to breastfeeding in places outside of their homes, including work, school and public
places (Blum, 1999; Chezem, Montgomery, & Fortman, 1997; Steams, 1999). This
additional item on the questionnaire read, “It is difﬁcult to ﬁnd places to breastfeed
outside your home.”

Although Janke (1994) suggested the use of both sets of questions for the
attitudinal measure, a decision was made to include only belief items on the instrument
for the current study. This decision was based on the researcher’s primary focus on

maternal beliefs about breastfeeding as they relate to one outcome: current breastfeeding

53

behavior. Other intended outcomes, although potentially inﬂuential in mothers’ feeding
decisions, were outside the focus of the current study.

As a result, fourteen items were included on the instrument for the current
investigation related to positiVe breastfeeding sentiment. An additional sixteen items
were included on the instrument related to negative breastfeeding sentiment. Response
categories on the breastfeeding sentiment items were consistent with the original
instrument developed by Janke (1994). Respondents could respond to each statement
based on a scale that ranged from strongly disagree (coded with a l) to strongly agree
(coded with a 5). T he fourteen it ems o n p ositive b reastfeeding sentiment and sixteen
items on negative breastfeeding sentiment were factor analyzed separately to extract
factors related speciﬁcally to advantages and disadvantages of breastfeeding. Mean
scores for each factor of positive breastfeeding sentiment and negative breastfeeding
sentiment were calculated by taking the mean of item ratings of all items that loaded on
the factor. Mean factor scores on positive and negative breastfeeding beliefs were then
entered as predictors of infant feeding behavior in subsequent data analysis.

Socio-Demographic Characteristics. Several items on the questionnaire
addressed respondents’ socio-demographic characteristics: age, race/ethnicity, education,
marital status, and current employment/student status. Although several of these variables
were available from the secondary data set, these variables were included on the
questionnaire to supplement secondary socio-demographic and household information
maintained by the WIC Division of the MDCH.

Secondary Data. The Michigan WIC Division of the MDCH agreed to make

secondary data available that could be linked with respondents’ questionnaire answers

54

using their WIC identiﬁcation number. The process of secondary data extraction required
that desired ﬁelds be extracted on a speciﬁed date to encompass all active WIC clients
enrolled at a speciﬁc point in time. Based on the original timeline that anticipated three
weeks of survey administration, data was extracted on the complete active WIC client
group enrolled at the sixteen clinics as of June 24, 2003. The extraction was completed
by administrators at the Michigan WIC Division of the MDCH, who then made the data
set available for the current investigation. The only identiﬁer linking the primary and
secondary data sets was the WIC identiﬁcation number.

Data maintained by the WIC Division of the MDCH was collected from
enrollment paperwork and self-reported health history forms that WIC participants are
required to complete to enroll in the program and maintain WIC certiﬁcation. There are
three points at which women are required to complete paperwork to maintain
certiﬁcation. These data collection points correspond to the three categories of women
eligible to participate in the WIC program: pregnant, breastfeeding postpartum, and non-
breastfeeding postpartum. Accordingly, health history forms are required of each woman
WIC participant at each of the following time intervals throughout their WIC certiﬁcation
period: during pregnancy, immediately postpartum, and six months postpartum.

During pregnancy, the Pregnant Woman’s Health History asks WIC participants
to provide information on socio-demographic characteristics of the expectant mother,
characteristics associated with the expectant mother’s prenatal care, environmental
conditions, family resources, and alcohol and tobacco use. If the woman continues her
certiﬁcation after birth, she is required to complete the Postpartum Woman’s Health

History during her ﬁrst visit after giving birth. This initial visit is most often scheduled

55

during the ﬁrst two weeks postpartum. The health history for this six-month period of
postpartum certiﬁcation asks women for information similar to the Pregnant Woman’s
Health History, but also asks women to provide information about their baby’s health and
their infant feeding practices. Non-breastfeeding mothers can maintain certiﬁcation for as
long as six months. WIC participants that continue to breastfeed through six months are
eligible to extend their WIC certiﬁcation to one year and are required to complete a Six-
Month Recertiﬁcation Health History.

As a result, the secondary data set maintained by the WIC Division contains
temporal records of clients’ health, infant and child health, environmental factors, and
demographic characteristics. Data extracted for the current investigation contained only
the most recent characteristics of WIC participants and not temporal characteristics
associated with changing behaviors.

Four secondary data sets were extracted for the current investigation that
contained data on mothers, infants and children active at each WIC participating clinic in
each urban location. The ﬁrst two sets of secondary data contained information on
expectant mothers and mothers with active enrollment in WIC by urban location.
Information contained in these data sets represented the most recent information on the
client and was based on the most recent health history form completed to maintain
certiﬁcation.

The second two sets of secondary data made available for the current study
contained information on active infant and child participants at each clinic in each urban
location. Health histories for these two groups of eligible \VIC participants are completed

by the guardian to establish the infant’s or child’s WIC certiﬁcation. The WIC Infant

56

Health History asks the guardian to provide information on their infant’s race/ethnicity,
method of infant feeding and eating history, infant’s overall health, parents’ physical and
mental health, and environment conditions. Infants and children may be certiﬁed for up
to one year; children are eligible to participate through age ﬁve. For the purposes of this
study, information on children was excluded from data analysis. Accordingly, secondary
data on infants was merged with corresponding secondary data on mothers.

Table 2 outlines the measures extracted from WIC Division data sets and

indicates which instruments collected these measures.

57

Table 2: WI C Secondary Data Extraction: Measures and Original Instruments

 

Pregnant
Woman’s Health

History

History

Postpartum
Woman’s Health

Six-Month
Recertiﬁeation
Health History

WIC Infant
Health History

Other

 

WIC Characteristics

 

WIC identiﬁcation number

 

WIC agency

 

WIC clinic

 

Type of WIC certiﬁcation

 

Original certiﬁcation date

XXXXX

 

Soda-Demographic Measures

 

Race/Ethnicity (mother)

 

Race/Ethnicity (infant)

 

Date of birth (mother)

 

Date of birth (most recent infant)

 

Gender (infant)

 

Marital status

 

Education

 

Household size

 

Household income

 

Participation in Other Support Programs

 

Medicaid

 

Family Independence Program

 

Food Stamps

XXX

 

Characteristics of Prenatal Care

 

Initiation of prenatal care

 

Location of prenatal care

 

Payment for prenatal care

XXX

XXX

 

Infant Feeding Behaviors

 

Initiation of breastfeeding

 

Initiation of formula feeding

 

Current breastfeeding status

><><><><

 

 

Duration of breastfeeding

 

 

 

 

 

 

 

In order to accomplish the objectives of this study, primary data collected from
questionnaires administered to WIC clients and secondary data maintained by the

Michigan WIC Division of the MDCH were merged to gain greater information on the

58

study sample. Primary and secondary data was merged based on WIC identiﬁcation
numbers, which participants were asked to record on the cover of their questionnaire.
Once the data sets were merged, characteristics of the mother were matched on several
ﬁelds to ensure consistency between information contained in the primary data set and
that contained in the secondary data set. In the event of speciﬁc inconsistencies between
primary and secondary characteristics, data was defaulted to one characteristic or the
other based on the nature of the characteristic. Speciﬁcally, if the characteristic was a
static characteristic, or one that does not change over time, the characteristic defaulted to
the secondary data set. This decision was based on the assumption that the WIC Division
maintains accurate records of p articipants’ static information, including r ace/ethnicity,
date of birth, and infant gender.

Alternatively, if the characteristic was a dynamic characteristic, or one that may
change over time, data was defaulted to the primary data set. This decision was based on
the timing of data collection for primary and secondary data. In particular, dynamic
variables contained in the primary data set gave a more recent representation of study
participants’ characteristics on those characteristics with the potential to change. As
previously described, secondary data on WIC participants is collected at three time
intervals and may reﬂect information that was dated from up to six months prior to the
date of survey completion.

After merging the primary and secondary data sets and checking for
inconsistencies, the study sample was explored in relation to the larger group of WIC
clients at the same location to determine the degree to which the study sample was

representative of other WIC clients. For this exploration, the larger WIC client group was

59

divided into two based on potential eligibility to have participated in the study. Using
these three groups for comparison (the study sample, the eligible WIC client group, and
the ineligible WIC client group), the current study explored the extent to which study
results could potentially explain other WIC clients at each urban location. Following
these comparisons, data analysis was conducted to explore the four study objectives that
guided the study.

Study Objective 1: Sample Description and Comparison. The ﬁrst Objective of
this study aimed to describe characteristics of the sample and to compare across
geographic location. The following characteristics were examined for this objective:
socio-demographic characteristics, participation in other matemal-child support
programs, factors a ssociated with p renatal c are, infant feeding behavior, beliefs about
breastfeeding, systems of external inﬂuence, maternal body image, and attitudes toward
motherhood and the baby. For these comparisons, characteristics were defaulted to the
primary or s econdary da ta 5 et b ased o n type 0 f c haracteristic: static (defaulted to the
secondary data set) or dynamic (defaulted to the primary data set). This procedure was
described previously. Additionally, where applicable, comparisons were made between
the larger eligible WIC client groups by urban location to determine whether differences
between the study samples would be expected based on differences between the larger
WIC client groups.

Study Objective 2: Measurement Reliability of Existing Instrumentation. The
second objective of this investigation was directed at testing measurement reliability of
existing subscales measuring maternal body image and attitudes toward motherhood and

the baby, and an existing scale measuring maternal beliefs about breastfeeding among

60

this sample of WIC participants. Two subscales of the Maternal Adjustment and Maternal
Attitudes (Kumar et al., 1984) and two scales of the Breastfeeding Attrition Prediction
Tool (J anke, 1994) were examined using exploratory factor analysis. On these scales and
subscales, exploratory factor analysis was used to determine whether the measurement
model could be improved by developing subscales that potentially reﬂect sub-factors of
breastfeeding beliefs, maternal body image, and attitudes toward motherhood and the
baby.

For this analysis, criteria were established that dictated circumstances under
which speciﬁc items of each subscale or scale were retained or removed from the factor
structure. These criteria were based on magnitude of factor loading and cross-loading on
two factors. In particular, items that failed to load at a level of 0.30 or higher were
removed from the factor structure. As well, items that loaded on more than one factor
simultaneously were examined for two characteristics. If the difference between the
factor loadings was greater than 0.05, the item was clustered on the factor on which its
factor loading was greatest. However, prior to placing the item on this factor, the item
was examined for its theoretical ﬁt with other items on the factor.

Based on these criteria, factors of each construct were extracted. Items that loaded
on each factor were analyzed for common themes to explore conceptual ﬁt with other
items that loaded on the factor. In order to be consistent with previous research in the
areas of body image, maternal adjustment, and beliefs about breastfeeding, themes were
gleaned from the items based on a review of relevant literature on these constructs.
Relying on previous literature, factors were labeled according to the common themes that

ran throughout the items that loaded on the factor. This analysis extracted several factors

61

for each construct, which were used in subsequent regression analyses as outlined in
Objectives 3 and 4 below.

Study Objective 3: Predicting Mothers ’ Infant Feeding Behavior on Systems of
External Inﬂuence. The third objective of this study aimed to predict respondents’
current infant feeding behavior on systems of external inﬂuence. The original proposed
regression model predicted current infant feeding behavior, measured as a breastfeeding
percentage, on three systems of e xtemal influence (family a nd friends, the he althcare
delivery system, and community sources) and their interactions. In addition, three other
independent variables were included in the regression model to control for respondents’
race/ethnicity, education, and beliefs about breastfeeding. In order to control for these
variables, it was decided that stepwise regression would be used and these ﬁelds were
entered into the regression model at the outset. The proposed model is presented as

Figure l.

62

Figure 1: Proposed Regression Model ‘A ’ Predicting Current Infant Feeding Behavior
on Systems of External Inﬂuence

 

 

IFB = a + D} (R) ‘I' b2 (E) + b3 (BBF) ‘1’ b4 (FF) + b5 (HDS) + D6 (CS) + b7 (FF'HDS) +
b3 (FF-CS) + b9 (HDS°CS) + b10(FF'HDS°CS) + e

Where:

IFB = Current Infant Feeding Behavior

R = Race

E = Education

BBF = Beliefs about Breastfeeding
FF = Family and Friends (System of External Inﬂuence)
HDS = Healthcare Delivery System (System of External Inﬂuence)

CS = Community Sources (System of External Inﬂuence)
= Intercept
bx = Average change in infant feeding behavior associated with one unit change in

the independent variable, holding all other independent variables constant

e = Error

 

Following data collection, the proposed model was revised based on limited
response to items measuring mothers’ perceived expectations of and valued place on
community sources. In particular, a large portion of the study sample indicated “don’t
know or does not apply” on these items, thereby impeding computation of an inﬂuence
score for each of these four sources. In the revised model, the inﬂuence of the community
sources system of inﬂuence and its interactions were removed from the proposed model.
The revised regression model for this study objective appears in Chapter 4.

Measurement of each variable has been described in some detail previously.

Brieﬂy, the dependent variable was measured as a breastfeeding percentage and

63

 

computed as the number of times the infant was fed breastmilk in a typical day divided
by the number of total feedings in a typical day. Socio-demographic control variables
were recoded into three dummy variables that measured race/ethnicity on the basis of
Black/Aﬁican-American or not Black/African-American, education on the basis of
obtaining a high school degree (or the equivalent) or obtaining less than a 12th grade
education, and another education variable that measured obtaining some college
education or no college-level education. Two positive breastfeeding sentiment factors and
three negative breastfeeding sentiment factors, in addition to one solitary item related to
negative breastfeeding sentiment, were measured as separate mean scores of all items that
loaded on the factor. Lastly, mean scores were computed for each system of external
inﬂuence based on inﬂuence scores for each individual/group in the speciﬁed system.
Variables were entered for the regression model in the order presented in Figure 1 using
stepwise procedures.

After results had been obtained for the predicted regression model, independent
variables in regression equation were tested for the presence of multicollinearity
according to the procedure outlined by Lewis-Beck (1980). High multicollinearity exists
when an independent variable is highly correlated with one or more independent
variables in the predicted regression equation. As discussed by Lewis-Beck (1980), the
presence of high multicollinearity threatens to reduce the reliability of parameter
estimates in the obtained regression equation. Although examination of the bivariate
correlations between each set of independent variables has the potential to suggest the
presence of multicollinearity, Lewis-Beck (1980) recommends regressing each

independent variable on all other independent variables to investigate the amount of

64

variance explained. Using this procedure, high multicollinearity can be detected if the
adjusted R2 obtained for any of the models approached 1.00. This result would indicate
the presence of a strong correlation between the (in)dependent variable and one or more
independent variables.

For the current investigation, in the event that high multicollinearity was detected,
the “offending” variable was removed from the original regression model as
recommended by Lewis-Beck (1980). However, as this procedure risks speciﬁcation
error in the revised model, Lewis-Beck (1980) recommends running a separate regression
model on the Offending variable(s). This process for investigating multicollinearity and
resolving issues with high multicollinearity was applied for the current study objective
and for the predicted regression model outlined in Study Objective 4.

Study Objective 4: Predicting Mothers’ Infant Feeding Behavior on Maternal
Body Image and Attitudes toward Motherhood The fourth objective of this study aimed
to predict respondents’ current infant feeding behavior based on maternal body image and
attitudes toward motherhood. Current infant feeding behavior, measured as a
breastfeeding percentage, was regressed on these two constructs. In addition, three other
independent variables were included in the regression model to control for respondents’
race, education, and beliefs about breastfeeding. Using stepwise regression, the control
variables were entered into the regression model ﬁrst, followed by body image factors
and factors related to attitudes toward motherhood and the baby. Measurement of the
dependent variable and control variables was described previously in Study Objective 3.
For the introduction of body image and attitudes toward motherhood and the baby into

the regression model, mean scores were computed for each extracted factor of these

65

subscales. In total, two factors related to body image and four factors related to attitudes
toward motherhood, in addition to one solitary item on the attitudes toward motherhood
construct, were entered into the regression model. Stepwise regression was used and the
regression model for this objective is presented in Figure 2.

Figure 2: Proposed Regression Model ‘B ’ Predicting Current Infant Feeding Behavior
on Body Image and Attitudes toward Motherhood and the Baby

 

 

IFB = a + b] (R) + b2 (E) + b3 (BBF) 'I" D4031) + b5 (AM) + b6 (BI'AM) + e

Where:
IFB = Current Infant Feeding Behavior
R = Race
E = Education
BBF = Beliefs about Breastfeeding
B1 = Maternal Body Image
AM = Attitudes toward Motherhood and the Baby
= Intercept
bx = Average change in infant feeding behavior associated with one unit change in

the independent variable, holding all other independent variables constant

e = Error

 

Based on the number of factors extracted from each of the Body lrnage subscale
and the Attitudes toward Motherhood and the Baby subscale, the decision was made to
exclude their interactions from the regression model. In particular, the sample size for the
current study was not sufﬁciently large to allow the number of interactions possible
between seven factors. As a result, the regression model presented in Figure 2 was
revised to remove interactions between body image and attitudes toward motherhood and

the baby. After running the proposed regression model, independent variables were

66

 

inspected for high multicollinearity using the procedure outlined for Study Objective 3.

Results of this and other data analysis procedures are presented in the subsequent chapter.

67

Chapter 4
RESULTS

The purpose of this study was to explore factors associated with infant feeding
choices among low-income women. In accordance with the study purpose, questionnaires
were distributed to seven clinics in Kent County, Michigan, and nine clinics in Detroit,
Michigan, that a dministered the S pecial S upplemental Nutrition Program for Women,
Infants and Children. In order to distribute and administer questionnaires proportionally
by clinic size at each of the sixteen WIC clinics, the sampling frame for the study was
based 0 n the total num ber o f W IC p articipants s erved at p articipating clinics in each
location during the month of April/May 2003. These ﬁgures included all active women,
infants and children enrolled in the WIC program during the speciﬁed month.

Based on April/May 2003 ﬁgures, 60.4% of all WIC participants were clients of
Detroit clinics and 39.6% were participants in Kent County clinics. Relying on this
distribution of participants by urban location, 315 questionnaires were delivered to the
Detroit City Health Department in packets proportional to clinic size and 200
questionnaires were delivered to the Kent County Health Department to be distributed
proportionally by clinic size. An additional 25 questionnaires were included in each
delivery to account for unusable surveys due to incompletion or refusal.

Dates of administration at each location spanned from June 16 through August 1,
2003, which constituted seven weeks of questionnaire administration. The study aimed to
recruit approximately 100-200 participants from each study location and, based on
April/May 2003 client counts, a smaller proportion of Kent County study participants

was expected. However, proportions of surveys obtained from each urban location

68

differed from expected proportions. In total, 229 completed questionnaires were collected
across both locations, wherein 119 completed questionnaires were collected from WIC
participants in Kent County clinics (52.0%) and 110 questionnaires were collected ﬁ'om
WIC participants in clinics in Detroit (48.0%). Contrary to expected proportions, study
participants in Kent County constituted a larger segment of the study sample than study
participants from Detroit. This result may be partially explained by questionnaire
distribution, administration and external management procedures, which differed between
the Kent County Health Department and the Detroit City Health Department based on
WIC staff resources. Expected and obtained study samples by WIC agency and clinic are
displayed in Table 3. Also presented in Table 3 are June 2003 ﬁgures on women with
active enrollment in WIC, categorized by potential eligibility to participate and

ineligibility to participate in the study.

69

225:. m 3255 05 E Eat: o>= a 8 FEB ﬁlm we: :2: E059: “actuate b..=£m=m o
5:9: coEuoam 05 E 0:5 5 EoE=oEu 3:8 5:5 5:83 3:0 m
55:. 3.50% of E OS» 5 EoEzoEo 038m 5:: 56:5 use 8.55 .8an =< <

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

>346 woﬁm wood om com :\:o.oo~ m _ m $062 New; 2333.5 53—
2 ......................................... 535:: 2:5
8m x: an o o :3 8 :2. ME: 8 2:5
ea 8. N2 c m :2 8 so: 5.. 8 2:5
:2 so. as. _ 2 so.” am :2 33 5 2:5
8m :8 E 2 A $3 3 2.2 $2 8 2:5
mm mm. as m m. 2.3 mm :2 33 8 2:5
.8 8. 8m _ : :2 R 2.2 22H 3 2:5
:3 EN :3 m 2 an: 9.. so... 23 8 2:5
:8 N. m :3 o m. s2 3. ea.” 33 8 2:5
3. m: o; c E :3 mm 2.? Noon 5 2:5
83 o _ 3 Sam 2 so :3: m: 5.8 SSA 4%: 5.8: Eu 2.55
m .......... m 8 can ................ 585:: 2:5
as m2 .5 m 8 and am e: A £3 5 2:5
Q :m 5 o m :3 A as... am 3. 2:5
.3 3.. mm o a :3 am so; 23 Q 2:5
mom a: w _ A o A 2.2 on 2.3 2.2 3. 2:5
:2 mm a: o a $3. 2 $2. :3 0. 2:5
:3 gm 8: _ t :3 m... an”. 22 mo. 2:5
pom E E 4 3 sea 2 :3 EH 3. 2:5
949m www 80% 2 2: $.on com £66m wwﬁc— anon 5.8: @550 «:3—
ﬂggtem 3592:5— Esoo 38260 9:60:00 @2539 35::me €203.25 $1300
vim beam mmcm 25.. 93:5 mzotzm :oEOQEm @855 5509:; moom
22235 :2225 22235 p22w=m . . age?
9.9.0 2.2m 0;? 295% mam oEEn— wEEEam

£55 33 553‘ USA :8 maﬁa. 3.35% 553.5 33 3883 “m 035—.

70

Eligibility to participate in the study was determined by the time elapsed since the
mother’s last live birth. Mothers that were up to three months postpartum were eligible to
participate. As indicated in Table 3, 29 respondents in the study sample were excluded
from subsequent data analysis due to ineligibility to participate in the study. Among this
group of ineligible study participants, 28 respondents were more than three months
postpartum (13 from Kent County and 15 from Detroit) and an additional respondent
from Detroit did not provide information on the infant’s date of birth. The secondary data
set was consulted on this characteristic, but information on this infant’s date of birth was
not available on the secondary data set either. As a result, data analysis was conducted on
primary data obtained from the questionnaires completed by of 200 postpartum WIC
participants. This data set was composed of 106 respondents from Kent County clinics
and 94 respondents from Detroit clinics.

Also outlined in Table 3 are ﬁgures on the larger WIC client group at each urban
location, which were based on an extraction of June 2003 active WIC participants at the
seven clinics in Kent County and nine clinics in Detroit. The ﬁgures outline active
women WIC participants at each clinic, above and beyond the study sample, that were
either eligible to participate or ineligible to participate in the study. These ﬁgures do not
include infants and children enrolled at the sixteen participating clinics. As Table 4
indicates, there were 888 active WIC participants in Kent County and 1,510 active WIC
participants in Detroit that were potentially eligible to participate in the study. According
to these ﬁgures, the eligible WIC client group at Kent County composed 37.0% of

eligible study participants while Detroit composed 63.0%. This reiterates the

71

discrepancies between surveys expected from each urban location and eligible surveys
obtained from Kent County (52.0%) and Detroit (48.0%).

Initial analysis on the eligible study sample compared characteristics of the study
sample with those of the larger group of WIC clients in each urban location, labeled the
WIC client group. Comparisons were performed to explore the degree to which the study
sample was representative of other eligible study participants at each location andto
investigate whether these c haracteristics w ere s irnilar to the larger group of ineligible
study participants as well. For this analysis, women WIC clients were categorized into
two groups based on eligibility to participate in the study, as speciﬁed in Table 3.
Resulting comparisons examined characteristics of women at each urban location across
three groups: the eligible study sample, WIC client group eligible study participants, and
WIC client group ineligible study participants. Similarities and differences across the
three groups were explored on four categories of characteristics: socio-demographic
characteristics, participation in other support programs, factors associated with prenatal
care, and infant feeding behaviors.

Comparisons were performed by urban location because of (1) differences in
proportionality between the obtained study sample (52.0% from Kent County; 48.0%
from Detroit) and the expected study sample (39.6% from Kent County; 60.4% from
Detroit) and (2) differences in socio-demographic composition of WIC participants in
each urban location. Chi-square analyses were performed to compare the eligible study
sample, the eligible WIC client group, and the ineligible WIC clients group on

categorical characteristics. Additionally, interval-level data was compared across these

72

three groups using analysis of variance. Results of all three categories of Kent County
comparisons are reported ﬁrst, followed by comparisons on Detroit groups.
Kent County Study Sample

As discussed, the Kent County study sample included 106 women. This group
was compared with two Kent County WIC client groups that did not participate in the
study. These groups were categorized by eligibility to participate in the study, wherein
888 Kent County WIC clients were eligible and 2,640 were ineligible. Comparisons were
ﬁrst made across these three groups on socio-demographic characteristics, wherein
differences between the three groups were statistically signiﬁcant on four of the
characteristics. For the purpose of reporting results, the Kent County study sample may
be abbreviated as “SSK,” the eligible WIC client group as “EK,” and the ineligible WIC
client group as “1K.” Results of comparisons on socio-demographic characteristics are

outlined in Table 4.

73

 

Table 4: Comparison of Kent County Study Participants and Eligible and Ineligible
WI C Client Groups on Socio-Demographic Characteristics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Kent County Eligible Ineligible
Eligible WIC Client WIC Client
Study Sample Group Group
Count Percent Count Percent Count Percent

Race/Ethnicity * 105 ----- 888 ----- 2,640 -----
Black/African-American 23 21.9% 218 24.5% 535 20.3%
White/Non-Hispanic 66 62.9% 458 51.6% 1,427 54.1%
Hispanic/Latina 14 13.3% 174 19.6% 583 22.1%
Other 2 1.9% 38 4.3% 95 3.6%

. A 24.9 5.6 25.6 5.7 25.1 5.6

Age (m years) (Mean) (SD) (Mean) (SD) (Mean) (SD)

Marital Status 105 ----- 887 ----- 2,614 -----
Married 40 38.1% 350 39.5% 915 35.0%
Not married 65 61.9% 537 60.5% 1,699 65.0%

Education Level *** 105 ----- 879 ----- 2,611 -----
Less than 12’h grade, no diploma 25 23.8% 307 34.9% 1,019 39.0%
High school degree or GED 29 27.6% 341 38.8% 899 34.4%
Some college, no degree 36 34.3% 194 22.1% 558 21.4%
College degree or higher 15 14.3% 37 4.2% 135 5.2%

—_

Infant Gender 106 ----- 888 ----- 997 -----
Male 56 52.8% 443 49.9% 511 51.3%
Female 50 47.2% 445 50.1% 486 48.7%

Infant Race/Ethnicity *** 28 ----- 888 ----- 996 -----
BIack/African-American 6 2 1.4% 271 30.5% 248 24.9%
White/Non-Hispanic 17 60.7% 372 41.9% 459 46.1%
Hispanic/Latina 2 7.1% 208 23.4% 250 25.1%
Other 3 10.7% 37 4.2% 39 3.9%

. B h”, 2.3 1.0 4.1 1.0 8.1 2.2

Infant Age 0” months) (Mean) (SD) (Mean) (SD) (Mean) (SD)

 

 

A Study Sample, n=105; Eligible WIC Client Group, n=888; Ineligible WIC Client Group, n=2,640
B Study Sample, n=106; Eligible WIC Client Group, n=888; Ineligible WIC Client Group, n=997

* p < 0.05
*** p < 0.001

74

 

The racial/ethnic distribution of each group was compared and across all three
groups, the largest proportion of each was White/Non-Hispanic (SSK=62.9%;
EK=51.6%; IK=54.1%), with rather sizeable proportions of Black/Aﬁican-American
participants (SSK=21.9%; EK=24.5%; IK=20.3%) and Hispanic/Latina participants
(SSK=13.3%; EK=19.6%; H<=22.1%). Differences between the racial/ethnic
distributions of the three groups were statistically signiﬁcant (xz=15.065, df=6, p=0.020)
and analysis of the adjusted standardized residuals suggested that there were two notable
differences across the three groups. The ﬁrst difference existed between the eligible WIC
client group and the ineligible WIC client group on proportion of Black/African-
American WIC participants, wherein the eligible WIC client group was composed of a
larger percentage of Black/African-American WIC participants (24.5%) than the
ineligible WIC client group (20.3%). The second difference existed between the study
sample and the ineligible WIC client group on proportion of Hispanic/Latina women,
wherein the sample did not contain as large a portion of Hispanic/Latina study
participants (13.3%) as the ineligible WIC client group (22.1%). For the primary
objective of these comparisons, which focused on the degree to which the study sample
was representative of the eligible WIC client group, neither of these differences was of
considerable concern. However, these results suggest that the study sample and the
eligible WIC client group were not representative of the ineligible WIC client group on
this characteristic, limiting the ability to generalize subsequent analysis of the study
sample to the overall WIC client group.

A second characteristic that was statistically different across the three Kent

County groups was education level obtained (x2=39.773, df=6, p=0.000). Adjusted

75

standardized residuals were computed and compared to investigate where the differences
between the three groups existed. Results of this analysis suggested that the study sample
was composed of a considerably greater proportion of women with at least some college
education (48.6%) than the two WIC client groups, while the eligible WIC client group
contained a greater proportion of women with a high school degree or the equivalent
(38.8%) than the study sample or the ineligible WIC client group. Lastly, the ineligible
WIC client group possessed a greater proportion of women with less than a 12th grade
education (39.0%) than the other two groups. These results indicate that the study sample
was, to a large extent, more educated than the larger eligible WIC client group or other
ineligible WIC participants at the same location.

Differences in education level obtained between the Kent County study sample
and the WIC client groups may have resulted from the study methodology, which relied
on a self-administered questionnaire written only in English to collect primary data on
study participants. Additionally, WIC staff was available in a limited capacity for
assisting study participants in completing the questionnaire. This required that potential
study participants were able to read written English and that they were able to
independently read and understand questionnaire items. As a result, potential study
participants’ ability to participate in the study may have been limited by their level of
education obtained.

Differences in marital status across the three Kent County groups were not
statistically signiﬁcant, but did approach statistical signiﬁcance (x2=5.862, dﬁ—Z,
p=0.053). Across all three groups, the majority of WIC participants in all three groups

were not married (SSK=61.9%; EK=60.5%; IK=65.0%), which included women that

76

were never married, divorced or separated, and widowed. Analysis of the adjusted
standardized residuals indicated that these differences approached signiﬁcance between
the two WIC client groups. In particular, more women in the eligible WIC client group
were married (39.5%) than in the ineligible WIC client group, where more women were
not married (65.0%).

Similarly, differences in mean age of women in each group approached statistical
signiﬁcance, where mean age among the Kent County study sample was 24.9 years of
age (SD=5.6), 25.6 years of age among the eligible WIC client group (SD=5.7), and 25.1
years of age among the ineligible WIC client group (SD=5.6). The minimum age across
all three groups was 14.1 years of age, which was characteristic of a woman in the
ineligible WIC client group. Maximum age was 49.0 years old and belonged to a woman
in the eligible WIC client group. Analysis of variance indicated that differences in mean
age were not statistically signiﬁcant across the three groups (F=2.422, dﬁ3,632,
p=0.089), but that differences approached statistical signiﬁcance. As with marital status,
this difference that approached statistical signiﬁcance existed between the eligible and
ineligible WIC client groups.

Three socio-demographic characteristics of mothers’ infants were also compared
and differences between the Kent County study sample and the two Kent County WIC
client groups were statistically signiﬁcant on two. For these comparisons, pregnant
women were necessarily excluded from analysis and the ineligible WIC client group was
reduced by approximately 1,600 women due to inability to describe characteristics of an

unborn infant.

77

Infant gender was distributed similarly across all three groups and, as would be
expected, proportions in each group approached approximately 50.0% of each gender.
The study sample was composed of slightly more women with male infants (52.8%), as
was the ineligible WIC client group (51.3%). The eligible WIC client group was almost
evenly distributed, with 49.9% of women having male infants. Differences between the
three groups were not statistically signiﬁcant (x2=0.549, df=2, p=0.760).

On the two other infant characteristics analyzed (race/ethnicity and age), both
were signiﬁcantly different between the groups even at a=0.05. However, both of these
results should be interpreted cautiously for distinct reasons, as described following each
result.

Differences between groups on racial/ethnic distribution of infants were
statistically signiﬁcant (x2=16.329, df=6, p=0.760) and, as indicated, this result should be
interpreted cautiously. Infants’ race/ethnicity was only available from the secondary data
set, as it was not collected on the questionnaire administered to WIC clients. Because
primary data was linked with secondary data on a smaller portion of the sample,
secondary information on infants’ race/ethnicity was available on a considerably smaller
group. As a result, information on infants’ race/ethnicity was severely limited among the
Kent County study sample. Analysis of the adjusted standardized residuals indicated that
the eligible WIC client group contained a smaller proportion of White/Non-Hispanic
infants (EK=41.9%) than the other two groups (SSK=60.7%; IK=46.1%) and a larger
proportion of Black/Afn'can-American infants (30.5%) than the ineligible WIC client
group (24.9%). However, to reiterate, these results should be interpreted cautiously due to

the small portion of the study sample accounted for on this characteristic (n=28).

78

Lastly, analysis of variance on mean infant age indicated that differences were
statistically signiﬁcant (F=1,626.222, df=l,990, p=0.000). Multiple comparisons were
run on this characteristic using Tukey HSD tests, which showed that differences were
signiﬁcant between each of the three. However, this result may be at least partially
related to the time at which data was collected on women in each group. The study
sample and the eligible WIC client group were composed of women that had given birth
only after March 1, 2003, whereas the mothers in the ineligible WIC client group may
have given birth anytime between July 1, 2002 and February 28, 2002. As a result, the
mean infant age among the ineligible WIC client group (8.1 months old) was
considerably larger than the mean age of the other two groups. Additionally, mothers in
the study sample were typically administered questionnaires on their ﬁrst visit to their
WIC clinic following childbirth. This appointment was generally scheduled within the
ﬁrst two weeks of the infant’s date of birth, thereby lowering the mean age of infants in
the study sample. As a result, although there were signiﬁcant differences on mean infant
age between the three groups, these differences were largely the result of timing of
primary and secondary data collection.

Next, a group of comparisons was performed on WIC clients’ participation in
three other types of support programs administered by the Family Independence Agency
for low-income mothers and their children. These programs included Medicaid, the
Family Independence Program (FTP), and the F ood Stamps P rogram. T hese p rograms
provide assistance for pregnant and postpartum women with dependent children in the
three forms: medical coverage, cash assistance, and food vouchers, respectively. Table 5

outlines participation in each of these programs by comparison group.

79

Table 5: Comparison of Kent County Study Participants and Eligible and Ineligible
WI C Client Groups on Participation in Other Support Programs

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Kent County Eligible Ineligible
Eligible WIC Client WIC Client
Study Sample Group Group
Count Percent Count Percent Count Percent
Medicaid *** 83 ----- 888 ----- 2,635 -----
Medicaid participant 53 63.9% 650 73.2% 1,520 57.7%
Not a Medicaid participant 30 36.1% 238 26.8% 1,115 42.3%
m m 83 ----- 888 ----- 2,635 -----
FIP participant 11 13.3% 189 21.3% 268 10.2%
Not an FIP participant 72 86.7% 699 78.7% 2,367 89.8%
Food Stamps *** 83 ----- 888 ----- 2,635 -----
Food Stamps participant 24 28.9% 339 38.2% 651 24.7%
Not a Food Stamps participant 59 71.1% 549 61.8% 1,984 75.3%
Overall Participation *** 83 ----- 888 ----- 2,635 ----
Participant in 0 programs 30 36.1% 227 25.6% 1,068 40.5%
Participant in 1 program 28 33.7% 318 35.8% 946 35.9%
Participant in 2 programs 15 18.1% 169 19.0% 370 14.0%
Participant in 3 programs 10 12.0% 174 19.6% 251 9.5%

 

 

 

 

*** p < 0.001

As indicated in Table 5, the program with the greatest proportion of participation

across all three groups was Medicaid (SSK=63.9%; EK=73.2%; IK=57.7%) with smaller

proportions participating in the Family Independence Program (SSK=13.3%; EK=21.3%;

IK=10.2%) and the Food Stamps Program (SSK=28.9%; EK=38.2%; IK=24.7%). There

were statistically signiﬁcant differences between the groups on participation in each of

these programs (Medicaid: x2=67.779, df=2, p=0.000; FIP: x2=72.632, df=2, p=0.000;

Food Stamps: x2=59.645, df=2, p=0.000). Once again, results should be interpreted with

caution, as 26 women in the study sample were excluded from analysis due to lack of

secondary information. However, analysis of the standardized adjusted residuals on each

80

 

of these signiﬁcant differences indicated that the differences existed between the eligible
and ineligible WIC client groups, wherein greater proportions of women in the eligible
WIC client group participated in each of the three programs (Medicaid: 73.2%; FIP:
21.3%; Food Stamps: 38.2%) than women in the ineligible WIC client group (Medicaid:
57.7%; FIP: 10.2%; Food Stamps: 24.7%).

The greatest proportion of women in each group participated in no more than one
of the three support programs. In particular, 69.8% of the Kent County study sample
participated in one or less programs, 61.4% of the eligible WIC client group participated
in one or fewer programs, and 76.4% of the ineligible WIC client group participated in no
more than one of the three support programs. There were statistically signiﬁcant
differences in proportions of mothers’ overall participation in any or all of the three
support programs and these differences existed between the eligible WIC client group
and the ineligible WIC client group. In particular, greater proportions of the eligible WIC
client group participated in two or more support programs (38.6%), whereas a greater
percentage of the ineligible WIC client group participated in no support programs
(40.5%). There were no considerable differences between the Kent County study sample
and these other two groups on overall program participation.

A third group of comparisons focusing on factors related to prenatal care was run
on the Kent County groups. Three factors were examined for this group of comparisons:
initiation of prenatal care, location of prenatal care, and payment for prenatal care. As
above, 26 women in the study sample were excluded from analysis due to lack of
secondary information. Results of comparisons on the study sample that participated in

the current investigation (n=83) and the two WIC client groups are presented in Table 6.

81

Table 6: Comparison of Kent County Study Participants and Eligible and Ineligible
WI C Client Groups on Characteristics of Prenatal Care

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Kent County Eligible Ineligible
Eligible WIC Client WIC Client
Study Sample Group Group
Count Percent Count Percent Count Percent
Initiation of Prenatal Care *** 83 ----- 888 ----- 2,635 -----
1St trimester 70 84.3% 757 85.2% 1,976 75.0%
2a or 3"T trimester 8 9.6% 120 13.5% 237 9.0%
No Prenatal Care/Unknown 5 6.0% 1 1 1.2% 422 16.0%
Location of Prenatal Care *** 82 ----- 879 ----- 2,590 -----
Hospital or health department clinic 9 l 1.0% 146 16.6% 359 13.9%
Private doctor’s ofﬁce/HMO 62 75.6% 670 76.2% 1,771 68.4%
Not going yet/Unknown l 1 13.4% 63 7.2% 460 17.8%
Payment for Prenatal Care *** 82 ----- 877 ----- 2,587 -----
Health insurance 10 12.2% 138 15.7% 392 15.2%
HMO 4 4.9% 15 1.7% 49 1.9%
Medicaid 63 76.8% 703 80.2% 1,808 69.9%
Other 5 6.1% 21 2.4% 338 13.1%
*** p < 0.001

As outlined in Table 6, the largest proportion of women in each group initiated

prenatal care during the ﬁrst trimester of pregnancy (SSK=84.3%; EK=85.2%;

IK=75.0%) and most received prenatal care in a private doctor’s ofﬁce or health

maintenance organization (HMO) (SSK=75.6%; EK=80.2%; IK=69.9%). Following from

results presented in Table 6, which indicated that large proportions of all three groups

were Medicaid participants, the largest percentage of each of the three groups paid for

prenatal care with Medicaid (SSK=76.8%; EK=80.2%; IK=69.9%).

However, there were statistically signiﬁcant differences between the eligible WIC

client group and the ineligible W1C client g roup o n all three 0 f t hese c haracteristics.

Further exploration of these differences suggested that larger proportions of the eligible

82

 

WIC client group initiated prenatal care during the ﬁrst trimester or the second or third
trimester (98.7%), whereas a considerably larger percentage of the ineligible WIC client
group had not initiated prenatal care or did not know when prenatal care was initiated
(16.0%). These differences mirror the differences between the eligible WIC client group
and ineligible WIC client group on location of prenatal care, wherein a greater proportion
of the ineligible WIC client group had not gone to prenatal care or went to an unspeciﬁed
location (17.8%). Finally, the eligible WIC client group possessed a greater proportion of
women that paid for prenatal care with Medicaid (80.2%). Payment for prenatal care was
the only characteristic on which the Kent County study sample differed signiﬁcantly from
the other two groups. On this characteristic, a greater percentage of women in the study
sample paid for prenatal care with an HMO (4.9%) than the other two groups (EK=1.7%;
IK=1.9%).

The last group of comparisons made on the three Kent County groups focused on
mothers’ infant feeding b ehaviors. F or t his g roup o f comparisons, four characteristics
were explored, including initiation of breastfeeding, initiation of formula feeding, current

breastfeeding status, and duration of breastfeeding. Results of these comparisons are

presented in Table 7.

83

Table 7: Comparison of Kent County Study Participants and Eligible and Ineligible
WIC Client Groups on Infant Feeding Characteristics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Kent County Eligible Ineligible
Eligible WIC Client WIC Client
Study Sample Group Group
Count Percent Count Percent Count Percent
Breastfeeding Status *** 106 ----- 888 ----- 2,640 -----
Currently breastfeeding 56 52.8% 309 34.8% 390 14.8%
Not currently breastfeeding 50 47.2% 579 65.2% 2,250 85.2%
Initiation of Breastfeeding *** 105 ----- 888 ----- 673 -----
Initiated breastfeeding 69 65.7% 516 58.1% 673 67.5%
Did not initiate breastfeeding 36 34.3% 372 41.9% 324 32.5%
Initiation of Formula Feeding *** 28 ----- 888 ----- 996 -----
Initiated formula feeding 15 53.6% 765 86.1% 865 86.8%
Did not initiate formula feeding 13 46.4% 123 13.9% 131 14.0%
Duration of Breastfeeding *** 105 ----- 888 ----- 997 -----
0 — 10 days 44 41 .9% 609 68.6% 463 46.4%
11 — 31 days 4 3.8% 211 23.8% 176 17.7%
Longer than 1 month 57 54.3% 68 7.7% 358 35.9%
*** p < 0.001

Differences between the three Kent County groups were signiﬁcant on all four
infant feeding behaviors examined. Initiation of breastfeeding included all mothers that
had ever breastfed their baby and across all three groups, the majority of mothers had
attempted to breastfeed their baby at least once (SSK=65.7%; EK=58.1%; IK=67.5%).
However, there were statistically signiﬁcant differences between the three groups on
initiation of breastfeeding (x2=18.123, df=2, p=0.000) and these differences existed
between the eligible and ineligible WIC client groups. In particular, the eligible WIC

client group had a smaller percentage of women that had not initiated breastfeeding

84

(41.9%) and the ineligible WIC client group had a larger percentage of women that had
initiated breastfeeding (67.5%).

Initiation of formula feeding was also signiﬁcantly different between the three
groups (x2=25.111, df=2, p=0.000), with the study sample possessing a larger percentage
of women that had not initiated formula feeding (46.4%) than the other two groups.
However, as discussed previously, this result should be interpreted cautiously due to the
small portion of study participants (n=28) that were accounted for in this statistic.
Because information was not available for the remaining 78 study participants in the Kent
County study sample, this statistic may not represent actual proportions of the study
sample that had initiated formula feeding. As a result, differences between the study
sample and the two W1C client groups may not have been as pronounced as the statistic
depicted.

The Kent County study sample possessed a large percentage of mothers that were
currently breastfeeding (52.8%), while percentages of mothers that were currently
breastfeeding in the other two groups were considerably smaller (EK=34.8%;
IK=14.8%). As expected, based on these proportions, current breastfeeding status
differed signiﬁcantly between all three groups (x2=230.038, df=2, p=0.000), which may
have been related to the distribution procedure for administering questionnaires in Kent
County. In this urban location, the Kent County WIC Breastfeeding Subcommittee
volunteered to administer questionnaires to WIC clients at the seven participant clinics.
Based on their orientation toward breastfeeding, members of this group may have

knowingly or unknowingly recruited more mothers that were currently breastfeeding.

85

Moreover, this result may have also been related to the mean age of infants in
each group (as outlined in Table 5), which was youngest among the study sample and
oldest among the ineligible WIC client group. Further analysis of the relationship
between infant age and current breastfeeding status within and between these three
groups indicated that, indeed, differences in current breastfeeding status were largely a
function of infant age. The three groups used for comparisons were categorized into six
by whether they were currently breastfeeding or not currently breastfeeding. Analysis of
variance was used to compare these six groups on infant age, which indicated that
differences in mean infant ages across the six groups were statistically signiﬁcant
(F=755.090, df=1,990, p=0.000).

The Tukey HSD was applied to this data in order to investigate the location of
differences between the six groups. There were not statistically signiﬁcant differences
between mean infant ages of mothers that were currently breastfeeding and mothers that
were not currently breastfeeding in the study sample. Similarly, mean infant ages
between the mothers in the eligible WIC client group that were currently breastfeeding
and not c urrently b reastfeeding were not statistically signiﬁcant. However, there were
signiﬁcant differences on all other pair-wise comparisons, indicating that infant age may
have interacted with current breastfeeding status to produce signiﬁcant differences
between the Kent County study sample, eligible W1C client group, and ineligible WIC
client group.

Lastly, breastfeeding duration was compared across the three groups and
differences were statistically signiﬁcant (12:263680, df=4, p=0.000). In particular,

analysis of the adjusted standardized residuals indicated that there was a larger

86

percentage of women in the eligible WIC client group that breastfed for one month or
less (62.3%), w hereas the s tudy s ample c ontained a c onsiderably larger proportion of
women that breastfed for longer than one month (54.3%). However, this ﬁgure may have
been inﬂated due to the method established for the current study for computing
breastfeeding duration. If the mother was still breastfeeding at the time of survey
administration, 8 he w as c onsidered to b e b reastfeeding through the c ompletion of the
current study and this may have inﬂated the percentage of women in the study sample
that breastfed longer than one month. As a result, differences between the study sample
and the other two WIC client groups may have been less pronounced than the depiction
in Table 7.
Detroit Study Sample

The Detroit study sample was composed of 94 eligible study participants. The
comparable group of WIC clients in Detroit, labeled the eligible WIC client group, was
composed of 1,510 women. The other group of women clients at the nine participating
Detroit clinics that were ineligible to participate in the study was composed of 3,807
women. As with the Kent County study sample, comparisons were run between the study
sample, the eligible WIC client group, and the ineligible WIC client group on four
categories of characteristics: socio-demographic characteristics, participation in other
support programs, factors related to prenatal care, and infant feeding behaviors. For the
purpose of reporting results, the Detroit study sample may be abbreviated as “SSD,” the
eligible Detroit WIC client group as “ED,” and the ineligible Detroit WIC client group as

“ID.” Results are reported in the tables that follow.

87

Table 8: Comparison of Detroit Study Participants and Eligible and Ineligible WI C
Client Groups on Socio-Demographic Characteristics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Detroit Eligible Ineligible
Eligible WIC Client WIC Client
Study Sample Group Group

Count Percent Count Percent Count Percent

Race/Ethnicity 89 ----- 1,510 ----- 3,805 -----
Black/African-American 77 86.5% 1,356 89.8% 3,408 89.6%
White/Non-Hispanic 4 4.5% 57 3.8% 148 3.9%
Hispanic/Latina 4 4.5% 52 3 .4% 130 3 .4%
Other 4 4.5% 45 3.0% 119 3.1%

. A 24.8 6.0 26.0 5.8 26.1 6.0

Age (m years) (Mean) (SD) (Mean) (SD) (Mean) (SD)

Marital Status 90 ----- 1,483 ----- 3,71 7 -----
Married 15 16.7% 277 18.7% 730 19.6%
Not married 75 83.3% 1,206 81.3% 2,987 80.4%
Education Level * 90 ----- 1,505 ----- 3,788 -----
Less than 126' grade, no diploma 22 24.4% 505 33.6% 1,255 33.1%
High school degree or GED 35 38.9% 614 40.8% 1,662 43.9%
Some college, no degree 27 30.0% 328 21.8% 751 19.8%
College degree or higher 6 6.7% 58 3.9% 120 3.2%
Infant Gender 94 ----- 1,510 ----- 1,560 -----
Male 53 56.4% 725 48.0% 788 50.5%
Female 41 43.6% 785 52.0% 772 49.5%
Infant Race/Ethnicity * 35 ----- 1,510 ----- 1,549 -----
Black/African-American 30 85.7% 1,364 90.3% 1,350 87.2%
White/Non-H ispanic 2 5.7% 41 2.7% 62 4.0%
Hispanic/Latina 3 8.6% 52 3.4% 56 3.6%
Other 0 0.0% 53 3.5% 81 5.2%

. B H,, 2.9 0.9 4.2 0.9 7.8 2.4

Infant Age (m months) (Mean) (SD) (Mean) (SD) (Mean) (SD)

 

A Study Sample, n=88; Eligible WIC Client Group, n=1,150; Ineligible WIC Client Group, n=3,805
B Study Sample, n=94; Eligible WIC Client Group, n=1.150; Ineligible WIC Client Group, n=1,561

* p < 0.05
*** p < 0.001

88

 

Overall, the entire Detroit WIC client group was considerably larger than the Kent
County WIC client group and was composed of 5,318 pregnant and postpartum women
(up to twelve months postpartum). As previously described, pregnant women and women
over three months postpartum were ﬁltered out of the eligible WIC client group and were
grouped together in an ineligible WIC client group that was composed of 3,807 women.
The eligible WIC client group was considerably smaller and consisted of 1,510 women.
Ninety-four eligible study participants were recruited from Detroit that made up the
Detroit study sample.

By far, the largest proportion of all three Detroit groups was Black/African-
American (SSD=86.5%; ED=89.8%; ID=89.6%). Small percentages of each of the three
groups were White/Non—Hispanic (SSD=4.5%; ED=3.8%; ID=3.9%), Hispanic/Latina
(SSD=4.5%; ED=3.4%; ID=3.4%), and other races/ethnicities (SSD=4.5%; ED=3.0%;
ID=3.1%). R acial/ethnic distributions 0 fall three g roups w ere similar and differences
were not statistically signiﬁcant across the three groups on this characteristic.

As with Kent C ounty, the Detroit study sample possessed signiﬁcantly greater
proportions of women with higher levels of education (12:15.178, df=6, p=0.019).
Across all three groups, the largest percentage of each group had earned a high school
degree or the equivalent (SSD=38.9%; ED=40.8%; ID=43.9%), with considerable
portions of each group obtaining less than a 12th grade education (SSD=24.4%;
ED=33.6%; ID=33.1%). Differences between the three groups existed between the study
sample and the WIC client groups on proportion that had obtained at least some college-

level education, wherein the study sample possessed considerably larger proportions of

89

mothers that had some college (30.0%) or a college degree or higher (6.7%) than either of
the WIC client groups from Detroit.

As with Kent County, a large percentage of the Detroit WIC clients were
unmarried across all three groups (SSD=83.3%; ED=81.3%; ID=80.4%). Differences on
this characteristic were not statistically signiﬁcant between the three groups (x2=1.042,
df=2, p=0.594). Mothers’ m ean a ge w as s lightly o lder a mong t he D etroit W IC 0 lient
groups at 26.0 years of age for the eligible WIC client group (SD=5.8) and 26.1 years of
age for the ineligible WIC client group (SD=6.0), compared to the mean age of 24.8 for
the study sample (SD=6.0). Once again, however, differences on this characteristic were
not statistically signiﬁcant across the three groups (F=1.866, df=5,402, p=0.155).

Three characteristics of mothers’ infants were compared as well, including infant
gender, infant race/ethnicity, and infant age. As would be expected, differences between
the Detroit study sample and the two WIC client groups were not statistically signiﬁcant

on infant gender (x2=3.757, df=2, p=0.153). Although differences between the three

groups were statistically signiﬁcant on infant race/ethnicity (x2=14.467, df=6, p=0.025),
this result should be treated with extreme caution due to very limited information on
infant race/ethnicity among the study sample. In particular, information on infant
race/ethnicity was only available on 35 study participants’ infants, which was
considerably smaller than the total number of eligible study participants in the Detroit
study sample (n=94).

The ﬁnal characteristic on mothers’ infants, infant age, was signiﬁcantly different
between the three groups (F=1,624.853, df=3,164, p=0.000). Based on the results of post

hoc tests, it was found that differences in mean infant age were signiﬁcant between all

90

pairs of the three groups. In particular, the study sample had a mean infant age that was
signiﬁcantly younger than both WIC client groups, while the ineligible WIC client group
had a mean infant age that was signiﬁcantly older than the eligible WIC client group and
the study sample.

Two assumptions of analysis of variance were violated on this comparison,
including homogeneity of variances and equal sample size. Although violations of the
assumptions of analysis of variance can be made with relatively minor consequence, the
combination 0 f t hese two violations c an p roduce r esults that are less robust (Howell,
2002). In this event, Howell (2002) recommends decreasing the degrees of freedom
considerably in order to obtain a more conservative critical value. Using this procedure
and limiting the degrees of freedom to 1 for the numerator and denominator, it was
concluded that the F -statistic for this comparison was sufﬁciently large to compensate for
violation of these two assumptions. As a result, mean infant age was signiﬁcantly
different between all three groups.

Next, using secondary data obtained from the Michigan WIC Division,
comparisons were performed to investigate differences in participation in other matemal-
child support programs across the three groups. As previously described, data was
available on participation three programs available to low-income mothers and their
children: (1) Medicaid, a medical assistance program, (2) the Family Independence
Program ( FIP), a c ash a ssistance p rogram, and (3) the Food Stamps Program, a food
assistance program. Data was available on a portion of the Detroit study sample (n=58) to
compare with characteristics of the eligible WIC client groups in this location. Based on

the small portion of the study sample represented in these statistics, the presence or

91

absence of signiﬁcant differences between the study sample and other W1C clients should
be treated with caution. Results of these comparisons on program participation are

presented in Table 9

Table 9: Comparison of Detroit Study Participants and Eligible and Ineligible WI C
Client Groups on Participation in Other Support Programs

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Detroit Eligible Ineligible
Eligible WIC Client WIC Client
Study Sample Group Group
Count Percent Count Percent Count Percent
Medicaid *** 58 ----- 1,510 ----- 3,794 -----
Medicaid participant 38 65.5% 998 66.1% 2,269 59.8%
Not a Medicaid participant 20 34.5% 512 33.9% 1,525 40.2%
FIP *** 58 ----- 1,510 ----- 3,794 -----
FIP participant 17 29.3% 427 28.3% 799 21.1%
Not an FIP participant 41 70.7% 1,083 71.7% 2,995 78.9%
Food Stamps *** 58 ----- 1,510 ----- 3,794 -----
Food Stamps participant 23 39.7% 622 41.2% 1,291 34.0%
Not a Food Stamps participant 35 60.3% 888 58.8% 2,503 66.0%
Overall Participation *** 58 ----- 1,510 ----- 3,794 -----
Participant in 0 programs 20 34.5% 506 33.5% 1,482 39.1%
Participant in 1 program 15 25.9% 373 24.7% 1,029 27.1%
Participant in 2 programs 6 10.3% 219 14.5% 519 13.7%
Participant in 3 programs 17 29.3% 412 27.3% 764 20.1%

 

 

 

 

 

*** p < 0.001

 

Results presented in Table 9 indicate that the largest proportions of the Detroit
three groups participated in Medicaid (SSD=65.5%; ED=66.1%; ID=59.8%), followed by
participation in the Food Stamps Program (SSD=38.7%; ED=41.2%; ID=34.0%) and the
ED=28.3%; ID=21.1%).

Family Independence Program (SSD=29.3%; Although

participation was signiﬁcantly different between the three groups on all three programs,

these differences existed primarily between the eligible WIC client group and the

92

ineligible WIC client group. Indeed, the only program on which the Detroit study sample
differed from either 0 f t he W [C client groups was the Family Independence Program
(x2=24.354, df=2, p=0.000), where the study sample contained a greater proportion of
participants that were enrolled in the program than the ineligible WIC client group.

There were also statistically signiﬁcant differences between the study sample and
the WIC client groups on overall program participation (x2=38.599, df=6, p=0.000). In
particular, a larger proportion of the Detroit study sample (29.3%) participated in all three
support programs than the ineligible WIC client group (20.1%). Additionally, differences
were signiﬁcant b etween the eligible W IC c lient g roup a nd t he ineligible W IC client
group in the proportion of women participating in none of the programs. Speciﬁcally, a
greater percentage of the ineligible WIC client group were not enrolled in any of the three
support programs (39.1%) than the eligible WIC client group (33.5%).

Comparisons were made on other characteristics available from the secondary
data set related to mothers’ prenatal care. Three characteristics of prenatal care were
examined: initiation, location, and payment. As with other characteristics contained
solely in the secondary data set, information was available for only a portion of the
Detroit s tudy s ample. As a r esult, 8 light 0 aution s hould b e exercised in assessing the
degree to which the study sample is representative of other WIC clients based on these

characteristics. Results of these comparisons are presented in Table 10.

93

Table 10: Comparison of Detroit Study Participants and Eligible and Ineligible WI C
Client Groups on Characteristics of Prenatal Care

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Detroit Eligible Ineligible
Eligible WIC Client WIC Client
Study Sample Group Group
Count Percent Count Percent Count Percent
Initiation of Prenatal Care *** 58 ----- 1,510 ----- 3,794 -----
1" trimester 43 74.1% 1,033 68.4% 2,779 73.2%
2nd or 3“r trimester 12 20.7% 404 26.8% 745 19.6%
No Prenatal Care/Unknown 3 5.2% 73 4.8% 270 7.1%
Location of Prenatal Care ** 58 ----- 1,489 ----- 3,718 -----
Hospital or health department clinic 44 75.9% 894 60.0% 2,231 60.0%
Private doctor’s ofﬁce/HMO 10 17.2% 514 34.5% 1,206 32.4%
Not going yet/Unknown 6.9% 81 5.4% 281 7.6%
Payment for Prenatal Care *** 58 ----- 1,484 ----- 3,717 -----
Health insurance 24 41.4% 402 27.1% 1,216 32.7%
HMO 5 8.6% 213 14.4% 607 16.3%
Medicaid 25 43.1% 792 53.4% 1,626 43.7%
Other 4 6.9% 77 5.2% 268 7.2%
** p < 0.01
*** p < 0.001

 

All characteristics associated with prenatal care were signiﬁcantly different
between the three Detroit comparison groups. Differences were signiﬁcant between all
three groups on the ﬁrst characteristic: initiation of prenatal care (x2=37.726, df=4,
p=0.000). Initiation of prenatal care commenced signiﬁcantly earlier for a larger
proportion of the Detroit study sample (74.1%) than the eligible WIC client group
(68.4%). Alternatively, the eligible WIC client group had a greater proportion of women
that initiated prenatal care in the second or third trimester, compared to 20.7% of the
study sample and 19.6% of the ineligible W1C client group.

The second characteristic associated with prenatal care, location of prenatal care,

was also signiﬁcantly different between the Detroit study sample and the two WIC client

94

 

groups (x2=15.011, df=4, p=0.005). In particular, a signiﬁcantly greater proportion of the
study sample received prenatal care from a hospital or health department clinic (75.9%)
than either of the WIC client groups (ED=60.0%; ID=60.0%). Conversely, much larger
proportions of the WIC client groups received prenatal care from a private doctor’s office
or HMO (ED=34.5%; lD=32.4%) than the Detroit study sample (75.9%). Payment for
these prenatal care visits also differed signiﬁcantly (x2=45.023, df=6, p=0.000), wherein
a larger percentage of the Detroit study sample paid for prenatal care with private
insurance (41.4%) than the WIC client groups (ED=27.1%; ID=32.7%). Alternatively,
these groups had greater proportions of women that paid for prenatal care with an HMO
(ED=14.4%; ID=16.3%) or, in the case of the eligible WIC client group, paid for prenatal
care with Medicaid (53.4%).

Finally, comparisons were made on the Detroit sample and its corresponding WIC
client groups on infant feeding behavior. For these comparisons, data was available on

four characteristics and results of these comparisons are presented in Table 11.

95

Table 11: Comparison of Detroit Study Participants and Eligible and Ineligible WI C
Client Groups on Infant Feeding Characteristics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Detroit Eligible Ineligible
Eligible WIC Client WIC Client
Study Sample Group Group
Count Percent Count Percent Count Percent
Breastfeeding Status *** 94 ----- 1,510 ----- 3,807 -----
Currently breastfeeding 22 23.4% 294 19.5% 489 12.8%
Not currently breastfeeding 72 76.6% 1,216 80.5% 3,318 87.2%
Initiation of Breastfeeding *** 93 ----- 1,510 ----- 1,561 -----
Initiated breastfeeding 44 47.3% 491 32.5% 744 47.7%
Did not initiate breastfeeding 49 52.7% 1,019 67.5% 817 52.3%
Initiation of Formula Feeding * 35 ----- 1,510 ----- 1,549 -----
Initiated formula feeding 26 74.3% 1,341 88.8% 1,392 89.9%
Did not initiate formula feeding 9 25.7% 169 11.2% 157 10.1%
Duration of Breastfeeding *** 88 ----- 1,510 ----- 1,560 -----
0 — 10 days 58 65.9% 1,236 81.9% 1,027 65.8%
11— 31 days 8 9.1% 228 15.1% 213 13.7%
Longer than 1 month 22 25.0% 46 3.0% 320 20.5%
* p < 0.05
*** p < 0.001

As indicated in T able 1 1, there w ere s igniﬁcant differences b etween the three

groups on all four infant feeding behaviors. Initiation of two types of infant feeding

(breastfeeding and formula feeding) were compared across the three groups. Initiation of

breastfeeding was signiﬁcantly different (XL—“74.992, dﬁ2, p=0.000), and these

differences were present between the eligible W1C client group and the other two

comparison groups. Percentages reported in Table 12 indicate that the WIC client group

possessed a signiﬁcantly smaller proportion of women that had ever breastfed (32.5%)

than the other two groups (SSD=47.3%; ID=47.7%). Initiation of formula feeding was

signiﬁcantly different at a more conservative alpha level (x2=9.010, df=2, p=0.011),

96

 

wherein a smaller percentage of the study sample had ever initiated formula feeding
(74.3%) than either of the WIC client groups (ED=88.8%; ID=89.9%). It should be noted
that this statistic was computed on slightly over one-third of the total Detroit study
sample for whom data was available at the time of the study. As a result, these
differences should be interpreted with some caution. Consistent with these results, a
signiﬁcantly larger proportion of the eligible WIC client group breastfed for the shortest
duration of time, or zero to ten days (81.9%). Conversely, larger percentages of the
Detroit study sample (25.0%) and the ineligible WIC client group (20.5%) breastfed
longer than one month.

A signiﬁcantly larger proportion of the Detroit study sample was currently
breastfeeding (23.4%) than the other two WIC client groups(ED=19.5%; ID=12.8%)
(XL—742.969, df=2, p=0.000). As in the Kent County comparisons, it was suspected that
this result was related to differences in mean infant age across the three groups, which
gradually increased from the Detroit study sample to the eligible WIC client group to the
ineligible WIC client group. Six groups were categorized by current breastfeeding status
(currently breastfeeding or not currently breastfeeding) and comparison group (study
sample, eligible WIC client group, or ineligible WIC client group). Analysis of variance
was used to compare these six groups on infant age to investigate the role of this
characteristic in apparent differences between the original three comparison groups.
Results of this analysis indicated that mean infant ages across the three groups were
statistically signiﬁcant (F=744.793, df=3,164, p=0.000), suggesting the mediating
inﬂuence of mean infant age on the apparent differences in current breastfeeding status

between the three comparison groups. Indeed, subsequent results of a post hoc test to

97

identify the location of these differences indicated that mean age differed between the
three comparison groups but, for the most part, was not statistically different within the
comparison groups. The only exception to these results was mean age on the ineligible
WIC client group, where mean infant age signiﬁcantly differed within the group as well.

Having explored the degree to which each location’s study sample represented the
population of WIC participants at these two urban locations, results of the four research
objectives that guided this study are presented in the subsections below.

Study Objective 1:
Comparison by Urban Location

The ﬁrst objective of this study was to describe the study sample of postpartum
WIC participants and to compare these characteristics across two urban areas in
Michigan. Primary data from administered questionnaires and secondary data from
Michigan W1C Division were analyzed for this objective, which included measurements
of socio-demographic characteristics, prenatal care factors, characteristics of mothers’
infant feeding choices, beliefs about breastfeeding, sources of external inﬂuence,
maternal body image, and attitudes toward motherhood and the baby.

Secondary data was available on Kent County and Detroit WIC client groups on a
selection of these characteristics, which permitted exploration of potential differences
between the larger groups of eligible study participants at each location. Where data was
available, comparisons were made between WIC client groups in tandem with
comparisons on study samples from each location. Results of these analyses are discussed

where applicable.

98

Primary data on respondents’ race/ethnicity, age of mother and infant, infant
gender, marital status, education level, employment status, and student status was
analyzed and compared across urban location. Secondary data obtained from Michigan
WIC Division was used to conﬁrm consistency between respondents’ questionnaire
answers and the secondary data client proﬁle on four of these variables: race/ethnicity,
age, marital status, and education level. Speciﬁc inconsistencies between characteristics
on the primary and secondary data sets were defaulted to the primary data characteristic
on dynamic characteristics (marital status and education level) and to the secondary data
characteristic on static characteristics (race/ethnicity, infant gender, and ages of mother
and infant). This procedure was described in Chapter 3.

Descriptive statistics were calculated for the overall sample and for study samples
by urban location. Table 12 outlines characteristics of the sample overall, as well as
comparisons between the two study locations. Statistically signiﬁcant differences
between the Kent County sample and the Detroit sample on speciﬁc characteristics are

marked with asterisks that identify level of signiﬁcance.

99

Table 12: Comparison of Study Participants by WIC Agency on Socio-Demographic

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Characteristics
Overall Kent County Detroit
Sample Sample Sample
Count Percent Count Percent Count Percent
Race/Ethnicity *** 194 ----- 105 ----- 89 -----
Black/Aﬁican-American 100 51.5% 23 21.9% 77 86.5%
White/Non-Hispanic 70 36.1% 66 62.9% 4 4.5%
Hispanic/Latina 18 9.3% 14 13.3% 4 4.5%
Other 6 3.1% 2 1.9% 4 4.5%
. A 24.9 5.7 24.9 5.6 24.8 6.0
Age 0“ years) (Mean) (SD) (Mean) (SD) (Mean) (SD)
Marital Status ** 195 ----- 105 ----- 90 -----
Manied 55 28.2% 40 38.1% 15 16.7%
Not married 140 71.8% 65 61.9% 75 83.3%
Education Level 195 ----- 105 ----- 90 -----
Less than 12lh grade, no diploma 47 24.1% 25 23.8% 22 24.4%
High school degree or GED 64 32.8% 29 27.6% 35 38.9%
Some college, no degree 63 32.3% 36 34.3% 27 30.0%
College degree or higher 21 10.8% 15 14.3% 6 6.7%
Employment Status 173 ----- 93 ----- 80 -----
Employed before and after birth 39 22.5% 26 28.0% 13 16.3%
Employed only before birth 58 33.5% 26 28.0% 32 40.0%
Employed only after birth 6 3.5% 3 3.2% 3 3.2%
Not employed before or after birth 70 40.5% 38 40.9% 32 40.0%
Student Status 154 ----- 81 ----- 73 -----
Attended before and after birth 16 10.4% 6 7.4% 10 13.7%
Attended only before birth 25 16.2% 11 13.6% 14 19.2%
Attended only after birth 3 1.9% 2 2.5% 1 1.4%
Did not attend before or after birth 110 71.4% 62 76.5% 48 65.8%

 

 

 

 

 

 

 

 

A Overall Sample, n=193; Kent County Sample, n=105; Detroit Sample, n=88
B Overall Sample, n=200; Kent County Sample, n=106; Detroit Sample, n=94

** p < 0.01
*** p < 0.001

 

(Table 12 continued on next page)

100

 

Table 12 (Continued): Comparison of Study Participants by WIC Agency on Socio-
Demographic Characteristics

 

 

 

 

 

 

 

 

 

 

 

 

Overall Kent County Detroit
Sample Sample Sample
Count Percent Count Percent Count Percent
Infant Gender 200 ----- 106 ----- 94 -----
Male 109 54.5% 56 52.8% 53 56.4%
Female 91 45.5% 50 47.2% 41 43.6%
Infant Race/Ethnicity *** 63 ----- 28 ----- 35 -----
Black/African-American 36 57.1% 6 21.4% 30 85.7%
White/Non-Hispanic 19 30.2% 17 60.7% 2 5.7%
Hispanic/Latina 5 7.9% 2 7.1% 3 8.6%
Other 3 4.8% 3 10.7% 0 0.0%
. B M". 2.5 1.0 2.3 1.0 2.9 0.9
Infant Age (1n months) (Mean) (SD) (Mean) (SD) (Mean) (SD)

 

 

 

 

 

 

 

 

A Overall Sample, n=193; Kent County Sample, n=105; Detroit Sample, n=88
B Overall Sample, n=200; Kent County Sample, n=106; Detroit Sample, n=94
** p < 0.01

*** p < 0.001

As indicated in Table 12, differences between the Kent County sample and the
Detroit sample were statistically signiﬁcant on two socio-demographic characteristics
associated speciﬁcally with the mother: race/ethnicity (x2=89.586, df=3, p=0.000) and
marital status (x2=10.989, df=], p=0.001; correction for continuity: x2=9.956, df=1,
p=0.002). Among the overall sample, large proportions were Black/African-American
(51.5%) and White/Non-Hispanic (35.6%). After dividing the sample by WIC agency,
results indicated that there were statistically signiﬁcant differences between participants’
race/ethnicity at the two urban locations (x2=89.586, df=3, p=0.000). The racial/ethnic
distribution of the Kent County sample was predominantly White/Non-Hispanic (62.9%),
but the sample also possessed substantial proportions of Black/African-American WIC

participants (21.9%) and Hispanic/Latina WIC participants (13.3%). Conversely, the

101

Detroit sample was considerably more homogenous, wherein 86.5% of the sample was
Black/Aﬁican-American.

Because secondary data was available on the WIC client group at each urban
location, comparisons were made between the Kent County and Detroit WIC client
groups to determine whether racial/ethnic differences between the two samples would be
expected based on differences between the larger groups of eligible non-participants at
each location. Indeed, there were statistically signiﬁcant differences between the two
WIC client groups on this characteristic (x2=1,115.147, df=3, p=0.000) and, similar to the
Detroit study sample, this result was likely based on the large proportion of
Black/African—American WIC participants (89.8%) in the Detroit WIC client group.

Marital status also differed signiﬁcantly b etween the K ent C ounty a nd Detroit
study samples. Among the overall sample, the majority of study participants were not
married (71.8%) and, after splitting the overall sample by urban location, this
characteristic was also present in the two smaller study samples. In particular, the Kent
County s tudy s ample w as c omposed o f 61. 9% w omen that w ere not married and the
Detroit study sample was 83.3% unmarried. Despite ﬁnding that both samples were
composed largely of women that were not married, however, there were statistically
signiﬁcant differences between the Kent County sample and the Detroit sample on this
characteristic (x2=10.989, df=1, p=0.001; correction for continuity: x2=9.956, df=1,
p=0.002). As presented in Table 7, although both study samples were largely composed
of women that were not married, the Detroit sample contained a larger percentage of
unmarried women than the Kent County sample. This difference was also present

between the larger WIC client groups from each urban location (x2=123.186, df=1,

102

p=0.000; correction for continuity: x2=122.120, df=1, p=0.000), wherein 81.3% of the
Detroit WIC client group was not married, as opposed to 60.5% of the Kent County WIC
client group.

Socio-demographic characteristics associated with the mother that were not
signiﬁcantly different between the Kent County and Detroit samples included age
(t=0.088, df=19l, p=0.930), education level (x2=4.771, df=3, p=0.189), employment
status (x2=4.517, df=3, p=0.211), and student status (12:3.068, df=3, p=0.381). Mean age
of the overall sample of mothers was 24.9 years old (SD=5.7) with a minimum age of
15.6 and a maximum age of 42.2. Mean ages of mothers in each urban location were
Similar to the overall mean age, with a mean age of 24.9 among the Kent County sample
(SD=5.6) and 24.8 among the Detroit sample (SD=6.0). A t-test was performed to
compare study samples from each urban location by age. This comparison conﬁrmed that
no statistically signiﬁcant differences existed between the Kent County sample and the
Detroit sample (t=0.088, df=l91, p=0.930). A comparison conducted between the WIC
client groups at each location corroborated with this result (t=-l .486, df=2,396, p=0.137),
indicating that no statistically S igniﬁcant differences 0 n m can a ge w ould b e expected
between clients from each urban location.

The overall sample of study participants was relatively well-educated, with 7 5.9%
of study participants having obtained at least a high school degree (or its equivalent). By
urban location, there were no statistically signiﬁcant differences between the Kent
County sample and the Detroit sample on this characteristic (x2=4.771, df=3, p=0.189).
Similarly, comparisons between the WIC client groups at each location indicated that

these larger groups were not signiﬁcantly different on education level obtained

103

(x2=1.058, df=3, p=0.787). However, referring to results presented earlier, comparisons
between the study sample from each location and the corresponding WIC client group
suggested that the study sample from each location had obtained higher levels of
education than the larger W1C client group. This difference in education level was
statistically signiﬁcant between the Kent County sample and its larger WIC client group
(x2=30.559, dﬁ3, p=0.000) and approached statistical signiﬁcance between the Detroit
sample and its larger WIC client group (x2=6.453, df=3, p=0.092). Thus, although the
Kent County and Detroit study samples did not differ signiﬁcantly on education level
obtained, these samples had obtained higher levels of education than each of their
corresponding WIC client groups.

The remaining two characteristics related Speciﬁcally to the mothers that were not
signiﬁcantly different between the Kent County and Detroit study samples were
employment status and student status. Although the largest proportion of the overall
sample was not employed before or after birth (40.5%), substantial proportions of the
overall sample were employed only before birth (33.5%) or were employed before and
after birth (22.5%). This result was similar on student status among the overall sample,
where the majority of study participants did not attend school before or after birth
(71.4%) and smaller percentages attended school only before birth (16.2%) or attended
school before and after birth (10.4%). Only small percentages of the overall study sample
initiated employment (3.5%) or started attending school after birth (1.9%).

Similarly, study samples by urban location were composed primarily of women
that were not employed or did not attend school before or after birth. Among the Kent

County sample, 40.9% were not employed before or after birth with sizeable portions

104

employed only before birth (28.0%) or employed before and after birth (28.0%).
Distribution on this characteristic was similar among the Detroit sample, where 40.0% of
Detroit study participants were employed only before birth. There were no statistically
signiﬁcant differences between the Kent County sample and the Detroit sample on
employment status (x2=4.517, df=3, p=0.211). Differences between the samples were not
statistically signiﬁcant on student status either (x2=3.068, df=3, p=0.381). Secondary data
was not available on these characteristics among the larger WIC client groups, limiting
the ability to compare the results of comparisons between the study samples to potential
differences between the larger WIC client groups.

Several socio-demographic characteristics of mothers’ infants were also compared
by urban location, including infant gender, race/ethnicity, and age. Of these three
characteristics, two were signiﬁcantly different between the Kent County sample and the
Detroit sample. As might be expected, infant gender distribution was similar between the
two urban locations, wherein 52.8% of the Kent County sample had male infants and
56.4% of the Detroit sample had male infants. However, there were differences between
the distribution of infant race/ethnicity by urban location and the mean age of infants
between the WIC agencies.

Concurrent with the statistically Signiﬁcant differences in mothers’ race/ethnicity
by urban location, compositions of infant race/ethnicity by urban location were
signiﬁcantly different ( x2=30.643, df =3, p =0.000). As e xpected b ased on racial/ethnic
distributions of mothers by urban location, a greater proportion of infants in the Detroit
study sample were Black/Aﬁican-American (85.7%), whereas a greater proportion of

infants in the Kent County study sample were White/Non-Hispanic (60.7%). These

105

differences were also present in comparisons between the eligible WIC client groups
from each urban location, suggesting that these differences should be expected. However,
based on the small portion of study respondents’ infants described by this characteristic
(n=63), ﬁgures presented on this characteristic for the study samples did not account for
infant r ace/ethnicity o n the m ajority o f s tudy p articipants. L astly, infant age was also
statistically different between the study samples by urban location (t=—4.31 1, df=198,
p=0.000). Based on comparisons of this characteristic between the two eligible WIC
client groups, this difference was expected.

Relying on information contained in the secondary data set, comparisons were
made by urban location on mothers’ participation in other matemal-child support
programs. As previously described, WIC maintains records on clients’ participation in
three support programs: Medicaid (medical assistance), FIP (cash assistance), and Food
Stamps (nutrition assistance). Although secondary information on these characteristics
was not available on all study participants, Table 13 presents results on the portion of the

study sample for which secondary data existed.

106

Table 13: Comparison of Study Participants by WIC Agency on Participation in

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Other Support Programs
Overall Kent County Detrort
Sample Sample
Count Percent Count Percent Count Percent
Medicaid 141 ----- 83 ----- 58 -----
Medicaid participant 91 64.5% 53 63.9% 38 65.5%
Not a Medicaid participant 50 35.5% 30 36.1% 20 34.5%
FIP * 141 ----- 83 ----- 58 -----
FIP participant 28 19.9% 11 13.3% 17 29.3%
Not an F 1P participant 1 13 80.1% 72 86.7% 41 70.7%
Food Stamps 141 ----- 83 ----- 58 -----
Food Stamps participant 47 33.3% 24 28.9% 23 39.7%
Not a Food Stamps participant 91 66.7% 59 71.1% 35 60.3%
Overall Participation 141 ----- 83 ----- 58 -----
Participant in 0 programs 50 35.5% 30 36.1% 20 34.5%
Participant in 1 program 43 30.5% 28 33.7% 15 25.9%
Participant in 2 programs 21 14.9% 15 18.1% 6 10.3%
Participant in 3 programs 27 19.1% 10 12.0% 7 29.3%

 

 

 

 

 

 

 

 

* p < 0.05

Comparisons on participation in other matemal-child support programs yielded
few statistically signiﬁcant differences. The Kent County and Detroit study samples did
not differ statistically on participation in Medicaid, Food S tamps, o r o verall p rogram
participation count. However, there were statistically signiﬁcant differences between the
two study samples on participation in FIP (12:5.531, df=l, p=0.019; correction for
continuity: x2=4.568, df=], p=0.033), wherein a greater proportion of the Detroit study
sample participated in this cash assistance program (29.3%) than the Kent County sample
(13.3%). This difference by urban location was supported by a comparison of the eligible

Kent County WIC client group and the eligible Detroit WIC client group, wherein

107

differences between the two larger groups on participation in FIP were also statistically
signiﬁcant (71914-330, df=1, p=0.000; correction for continuity: 12:13.996, df=l,
p=0.000). However, statistically signiﬁcant differences between the eligible WIC client
groups were also found on participation in Medicaid (x2=13.134, df=], p=0.000;
correction for continuity: x2=12.806, df=1, p=0.000) and overall program participation
count (12:56.118, df=3, p=0.000).

Several characteristics associated with study participants’ pregnancy, particularly
factors related to prenatal care, were available for analysis from secondary data. These
factors included timing of initiation of prenatal care, location of prenatal care, and
payment for prenatal care. Analysis was limited on these characteristics because
secondary data was only available on a portion of survey respondents. Nonetheless,
results are reported in Table 14 to convey general characteristics of prenatal care among a

reduced portion of the study sample.

108

 

Table 14: Comparison of Study Participants by WIC Agency on Characteristics of

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Prenatal Care
Overall Kent County Detroit
Sample Sample Sample
Count Percent Count Percent Count Percent
Initiation of Prenatal Care 141 ----- 83 ----- 58 -----
lSt trimester 1 13 80.1% 70 84.3% 43 74.1%
2’“ or 3rd trimester 20 14.2% 8 9.6% 12 20.7%
No Prenatal Care/Unknown 8 5.7% 5 6.0% 3 5.2%
Location of Prenatal Care *** 140 ----- 82 ----- 58 -----
Hospital or health department clinic 53 37.9% 9 11.0% 44 75.9%
Private doctor’s ofﬁce/HMO 72 51.4% 62 75.6% 10 17.2%
No prenatal care/Unknown 15 10.7% 11 13.4% 4 6.9%
Payment for Prenatal Care *** 140 ----- 82 ----- 58 -----
Health insurance 34 24.3% 10 12.2% 24 41.4%
HMO 9 6.4% 4 4.9% 5 8.6%
Medicaid 88 62.9% 63 76.8% 25 43.1%
Other 9 6.4% 5 6.1% 4 6.9%
*** p < 0.001

For both study samples by urban location, the majority of mothers initiated
prenatal care in the ﬁrst trimester of pregnancy (Kent County: 84.3%; Detroit: 74.1%). A
slightly larger proportion of the Detroit study sample initiated prenatal care in the second
or third trimester (20.7%), but differences between the two study samples by urban
location were not statistically signiﬁcant (x2=3.426, df=2, p=0.180).

However, differences were statistically signiﬁcant between the two study samples
on the other two characteristics of prenatal care: location of prenatal care (x2=61.632,
df=2, p=0.000) and payment for prenatal care (x2=18.835, df=3, p=0.000). Within the
Detroit study sample, a greater proportion received prenatal care in a hospital or health

department clinic (75.9%) and paid for prenatal care with health insurance (41.4%) than

109

the Kent County study sample. Alternatively, within the Kent County study sample, a
larger percentage of the sample received prenatal care in a private doctor’s ofﬁce or
health maintenance organization (7 5.6%) and paid for prenatal care with Medicaid
(76.8%) than that Detroit study sample.

Next, primary data collected on infant age and gender, mothers’ current and
previous breastfeeding and formula feeding practices, timing of the initial infant feeding
decision, and introduction of other foods were analyzed by urban location. Secondary
data obtained from Michigan WIC Division was used to conﬁrm consistency between
respondents’ questionnaire answers and the secondary data client proﬁle on ﬁve of these
variables: infant gender, infant age, initiation of breastfeeding, current breastfeeding
status, and duration of breastfeeding (if stopped). Speciﬁc inconsistencies between
characteristics on the primary and secondary data sets were defaulted to the primary data
characteristic on dynamic characteristics (characteristics associated with infant feeding)
and to the secondary data characteristic on static characteristics (infant gender and age).
This procedure was described in Chapter 3. Information on initiation of formula use was
also available in the secondary data set and was analyzed for this set of comparisons as
well. As with other factors only available in the secondary data set, analysis was limited
by the number of questionnaire respondents for whom secondary data was available.

Descriptive statistics were calculated on the characteristics outlined above for the
overall sample and for study samples by urban location. Table 15 outlines characteristics

of the sample overall, as well as comparisons between the two study locations.

110

Table 15: Comparison of Study Participants by WIC Agency on Infant Feeding

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Characteristics
Overall Kent County Detroit
Sample Sample
Count % Count % Count %

Timing of Feeding Decision ** 196 ----- 104 ----- 92 -----
Before pregnancy 82 41.8% 54 51.9% 28 30.4%
During pregnancy 58 29.6% 29 27.9% 29 31.5%
After birth 56 28.6% 21 20.2% 35 38.0%
Breastfeeding Status * 200 ----- 106 ----- 94 -----
Currently breastfeeding 78 39.0% 56 52.8% 22 23.4%
Not currently breastfeeding 122 61.0% 50 47.2% 72 76.6%

26.3% 0.40 37.7% 0.45 13.4% 0.28

- ***

Breastfeeding Percentage (Mn) (SD) (Mn) (SD) (Mn) (SD)
Initiation of Breastfeeding * 198 ----- 105 ----- 93 -----
Initiated breastfeeding 113 57. 1% 69 65.7% 44 47.3%
Had used a breast pump 65 58.6% 42 61 .8% 23 53.5%
Had never used a breast pump 46 41.4% 26 38.2% 20 46.5%
Did not initiate breastfeeding 85 42.9% 36 34.3% 49 52.7%
Initiation of Formula Feeding 63 ----- 28 ----- 35 -----
Initiated formula feeding 41 65.1% 15 53.6% 26 74.3%
Did not initiate formula feeding 22 34.9% 13 46.4% 9 25.7%
Introduction of Other Foods * 189 ----- 101 ----- 88 -----
Introduced other foods 19 10.1% 5 5.0% 14 15.9%
Did not introduce other foods 170 89.9% 96 95.0% 74 84.1%
Duration of Breastfeeding *** 193 ----- 105 ----- 88 -----
0 - 10 days 102 52.8% 44 41.9% 58 65.9%

11 — 31 days 12 6.2% 4 3.8% 8 9.1%
Longer than 1 month 79 40.9% 57 54.3% 22 25.0%
Previous Breastfeeding Experience 108 ----- 59 ----- 49 -----
Initiated breastfeeding with other children 52 49.1% 32 54.2% 21 42.9%
Did not previously initiate breastfeeding 55 50.9% 27 45.8% 28 57.1%

. 2.5 1.0 2.3 1.0 2.9 0.9

Infant A e in months ***

g ( ) (Mn) (SD) (Mn) (SD) (Mn) (SD)

 

*p<ocs
** p < 0. 01
*** p < 0.001

111

 

Of the nine infant feeding behaviors compared by urban location, differences
were statistically signiﬁcant between Kent County and Detroit on six characteristics.
Characteristics were similar by urban location on use of a breast pump (x2=0.744, df=1,
p=0.389), initiation of formula feeding (x2=2.937, df=l, p=0.087), and previous
breastfeeding experience (x2=1.387, df=2, p=0.239). Information on initiation of formula
feeding was limited by the portion of the study sample for which data was available. Of
the 63 study participants for whom data was available, 65.1% of the overall sample had
initiated formula feeding. Initiation of formula feeding was not signiﬁcantly different
between the eligible Kent County WIC client group and the eligible Detroit WIC client
group (x2=3.698, df=], p=0.054; correction for continuity: x2=3.453, df=1, p=0.063), but
differences between the larger eligible WIC client groups approached statistical
signiﬁcance

Among women that initiated breastfeeding, slightly over half of the overall study
sample (58.6%) had used a breast pump at least once. Ftuther exploration of this
characteristic indicated that 23.9% of the 113 w omen that had initiated b reastfeeding
typically fed their baby pumped breastmilk at least once per week. Secondary data on use
of a breast pump was not available to explore this characteristic among other WIC clients
at each urban location.

Study participants’ previous experience breastfeeding was assessed in an item
asking the respondent to indicate whether she had breastfed any of her other children, if
applicable. Of the 108 study participants with other children, less than half of the overall

sample had initiated breastfeeding with their other children (49.1%). To explore the

112

relationship of this characteristic with initiation of breastfeeding and current
breastfeeding status, Chi-square analysis was performed comparing these characteristics
with previous breastfeeding experience. Indeed, differences were signiﬁcant on both
comparisons. Among women with previous breastfeeding experience, a larger proportion
of women had initiated breastfeeding with their most recent infant (83.0%) than mothers
that did not (17.0%) (x2=24.335, df=1, p=0.000; correction for continuity: x2=22.441,
df=1, p=0.000). Similarly, among women with previous breastfeeding experience, a
larger proportion of the overall study sample was currently breastfeeding (62.3%) than

mothers that were not currently breastfeeding (37.7%) (x2=21.888, df=-1, p=0.000;

correction for continuity: x2=20.087, df=1, p=0.000).

Both of these characteristics, initiation of breastfeeding and current breastfeeding
status, were statistically different by urban location. Initiation of breastfeeding was much
greater among the Kent County sample (65.7%) than among the Detroit sample (47.3%)
(x2=6.817, df=l, p=0.009; correction for continuity: x2=6.087, df=1, p=0.014).
Breastfeeding status was also signiﬁcantly different by urban location (x2=18.133, df=1,
p=0.000; correction for continuity: x2=16.917, df=1, p=0.000), where 52.8% of the Kent
County study sample was currently breastfeeding compared with 23.4% of the Detroit
sample. Consistent with the above result suggesting differences between the Kent County
and Detroit study samples on current breastfeeding status, differences by urban location
were also statistically signiﬁcant on breastfeeding percentage (F=4.588, df=176.491,
p=0.000).

Comparisons were performed on the larger WIC client groups, which indicated

the presence of statistically signiﬁcant differences between the eligible Kent County WIC

113

client group and the eligible Detroit WIC client group on initiation of breastfeeding
(x2=150.340, df=1, p=0.000; correction for continuity: x2=149.292, df=1, p=0.000) and
current breastfeeding status (x2=69.787, df=], p=0.000; correction for continuity:
x2=68.976, df—‘l, p=0.000). Based on these results, differences on these characteristics by
urban location were expected.

Timing of the initial infant de cision w as m ade b efore p regnancy for a g reater
proportion of the Kent County study sample (51.9%) than the Detroit study sample
(30.4%) (x2=11.051, df=2, p=0.004). Based on previously reported differences between
these two groups on current breastfeeding status, this result was explored by comparing
timing of the infant feeding decision by current breastfeeding status. Analysis of these
characteristics indicated statistically signiﬁcant differences (x2=18.126, df=2, p=0.000),
wherein a greater proportion of mothers that were currently breastfeeding made their
decision to b reastfeed b efore ( 49.4%) o r during p regnancy (39.0%) than mothers that
were not currently breastfeeding. Conversely, a greater percentage of mothers that were
not currently breastfeeding made their initial infant feeding decision after birth (39.5%)
than mothers that were currently breastfeeding.

Differences in whether mothers had introduced other foods to the infant were
statistically signiﬁcant by urban location (x2=6.246, df=1, p=0.012). In particular, a
greater percentage of the Detroit sample had introduced other foods (15.9%) than the
Kent County sample (5.0%). However, based on statistically signiﬁcant differences
between the mean infant ages by urban location, this results was further explored for a
potential mediating inﬂuence of infant age on the introduction of other foods. Four

groups were formed by urban location and introduction of other foods, which were then

114

compared on mean infant age using analysis of variance. Results indicated that there were
statistically signiﬁcant differences between the four groups on mean infant age
(F=13.452, df=188, p=0.000). The Tukey HSD was applied to this data in order to
investigate the location of differences between the four groups. Differences were not
signiﬁcant between mothers that had introduced other foods and those that had not in
each urban location, suggesting that differences in introduction of other foods were more
likely linked to differences in infant age between the urban locations.

In addition to exploring and comparing infant feeding behaviors by urban
location, comparisons were conducted on beliefs about breastfeeding, sources of social
inﬂuence, body image, and attitudes toward motherhood. To measure beliefs about
breastfeeding, two scales of the Breastfeeding Attrition Prediction Tool (Janke, 1994)
were included on the questionnaire. The Positive Breastfeeding Sentiment Scale was
composed of fourteen items that measured mothers’ beliefs in the advantages of
breastfeeding, while the Negative Breastfeeding Sentiment Scale was composed of
sixteen items that measured mothers’ beliefs in the disadvantages of breastfeeding. All
items were rated on a ﬁve-point Likert scale that ranged from strongly disagree (rated
with a 1) to strongly agree (rated with a 5). Table 16 displays mean ratings of each item

by scale, including rank of the item among all items on the positive or negative scale.

115

Table 16: Overall Mean Item Ratings and Rankings on Scales of the Breastfeeding
Attrition Prediction Tool (Janke, 1994)

 

[ Rank] MeanA [ s.D.

 

Positive Breastfeeding Sentiment Scale

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Breastmilk is healthy for the baby. 1 4.60 0.70
Breastfeeding is natural. 2 4.36 0.76
Breastfeeding is best for the baby. 3 4.26 0.88
Breastmilk is more nutritious than infant formula. 4 4.25 0.89
Breastfeeding helps you bond with your baby. 5 4.12 0.89
Breastfeeding makes you closer to your baby. 6 4.01 1.02
Breastfeeding is more economical than formula feeding. 7 3.68 1.02
Breastfeeding is personally satisfying. 8 3.57 1.07
Breastfeeding is more convenient than formula feeding. 9 3.44 1.28
Infant formula can cause constipation. 10 3.25 0.83
Infant formula can cause allergies. 11 3.06 1.05
Formula fed babies are more fussy than breastfed babies. 12 2.90 0.89
Formula fed babies tend to get sick. 13 2.75 0.99
Formula fed babies tend to be overweight. 14 2.53 0.85
Negative Breastfeeding Sentiment Scale
Formula feeding gives you more ﬁ’eedom. l 3.71 0.93
Formula feeding is easier than breastfeeding. 2 3.66 1.05
It is embarrassing to breastfeed in public. 3 3.58 1.11
Breastfeeding is more time consuming than formula feeding. 4 3.56 1.06
Breastfeeding is painful. 5 3.54 1.06
Formula feeding lets the father become close to the baby. 6 3.53 0.97
No one else can help feed the baby when you breastfeed. 7 3.48 1.22
Breastfeeding makes returning to work difﬁcult. 8 3.42 0.96
When you breastfeed you never know if the baby gets enough milk. 9 3.41 1.05
Breastfeeding makes your breasts sag. 10 3.33 1.02
It is difficult to ﬁnd places to breastfeed outside your home. 11 3.32 1.01
Mothers who form. feed get more rest than breastfeeding mothers. 12 3.29 1.03
Mothers who formula feed get back into shape sooner. 13 3.19 1.20
Breastfeeding ties you down. 14 3.19 0.99
Formula fed babies are easier to satisfy than breastfed babies. 15 3.13 0.95
Breastfeeding is messy. 16 3.12 1.13

 

 

 

 

 

A Mean rating based on a ﬁve-point Likert scale (l=strongly disagree, 5=strongly agree)

As indicated in Table 16, mean ratings on most items of the Positive
Breastfeeding Sentiment Scale were neutral (rated with a 3) or higher. Mean ratings for

each item ranged from 2.53, indicating slight overall disagreement, to 4.60, suggesting

116

strong agreement. Among the top-ranked items, there was strong agreement with items
related to infant health and nutrition beneﬁts of breastfeeding and enhancement of
mother-child bonding. Further, there was relatively small variation in respondents’
answers among these top-ranked items.

Only three items on the Positive Breastfeeding Sentiment Scale received mean
ratings that indicated slight overall disagreement. The ﬁrst item, “Formula fed babies are
more fussy than breastfed babies,” had a mean rating of 2.90 (Rank: 12, SD=0.89), or just
under a mean neutral rating on the item. The second item read, “Formula fed babies tend
to get sick” (Rank: 13, Mean=2.75, SD=0.99) and the third item with a mean rating that
in slight disagreement stated, “Formula fed babies tend to be overweight” (Rank: 14,
Mean=2.53, SD=0.85). There was relatively small variation in respondents’ answers on
all three of these items. Further, all of these items seemed to be relatedto mothers’
assessment formula feeding as having negligible risk to infant health.

Mean ratings of items on the Negative Breastfeeding Sentiment Scale were more
uniformly in slight agreement across all items. In particular, all items had mean ratings
over 3, which would indicate a neutral sentiment to the item, but no items received a
mean rating larger than 4, which would suggest agreement with the item. Top-ranked
items on this scale seemed to suggest slight agreement with the greater freedom afforded
formula feeding mothers. Items that ranked toward the bottom of the list suggested
slightly less agreement with difﬁculties associated with the practice of breastfeeding and
impacts on the mother’s health and well-being.

Comparisons were made by urban location on the thirty items related to positive

and negative breastfeeding sentiment. Differences between mean scores on each of these

117

items were statistically signiﬁcant on seven items, which included six positive
breastfeeding sentiment items and one negative breastfeeding sentiment item. Items that
possessed statistically signiﬁcant mean scores by urban location are outlined in Table 17.

Table 1 7: Statistically Significant Differences on Positive and Negative Beliefs about
Breastfeeding by WIC Agency

 

I Kent County Detroit
Study Sample Study Sample

I Mean A I S.D. I Mean" I S.D.

 

 

Positive Breastfeeding Sentiment Scale

 

Breastmilk is healthy for the baby.

(t=2.342, df=145.692, p=0.021) 4-71 0556 4.47 0.822

 

Breastmilk is more nutritious than infant formula.

(1:2.540, df=180, p=0.012) 4.40 0.812 4.07 0,949

 

Breastfeeding helps you bond with your baby.
(t=2.417, df=182, p=0.017) 4-27 0771 3.95 0.987

 

Breastfeeding is more economical than formula feeding.

(t=3.083, df=180, 12:0. 002) 3-89 1016 3.44 0.985

 

 

Breastfeeding is personally satisfying.
(t=3.260, df=178,p=0.001) 3.81 0.976 3.30 1.107
Infant formula can cause allergies. 2 91 0 963 3 24 1 122

 

 

 

 

(t=-2.124, df=/62. 782, p=0. 035)

 

Negative Breastfeeding Sentiment Scale

 

 

No one else can help feed the baby when you breastfeed.
(t=-3. 152, df=183.925, p=0. 002) 3.22 1.274 3.77 1.097

 

 

 

 

 

 

A Mean rating based on a ﬁve-point Likert scale (1=strongly disagree, 5=strongly agree)

Results of comparisons by urban location on the thirty breastfeeding items
suggested a greater degree of positive breastfeeding sentiment among the Kent County
study sample. Among the six positive breastfeeding sentiment items with statistically
signiﬁcant differences, mean ratings of the Kent County sample were signiﬁcantly
higher, or in stronger agreement with the positive breastfeeding items, than the mean
ratings of the Detroit sample on all but one item. This one item, which stated “Infant
fomula can cause allergies,” had a higher mean score among the Detroit study sample

(Mean=3.24, S.D.=l.122). The Detroit study sample was also in stronger overall

118

agreement with the one statistically signiﬁcant negative breastfeeding S entiment item,
which read “No one else can help feed the baby when you breastfeed” (Mean=3.77,
S.D.=1.097).

Another set of comparisons were performed by study location on sources of
external inﬂuence in the infant feeding decision. Results of these comparisons are

outlined in Table 18.

Table 18: Mean Item Ratings on Sources of External Inﬂuence

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Mean A S.D.
Family and Friends 3.423 3.688
My baby’s father 3.910 5.591
My mother 4.091 4.691
My mother-in—law 3.636 4.042
My family 3.281 4.332
My best friend 2.832 4.323
My other friends 2.635 3.791
Breastfeeding Peer Counselors 5.091 4.654
Healthcare Delivery System 5.343 4.367
My doctor’s 5.525 4.576
My widwife 4.872 4.969
WIC Staff 6.071 3.871
Community Sources 2.196 3.439
Other people at my school 1.318 3.163
Other people at my work 2.163 4.029
My community 1.771 3.666
People in public places 1.275 3.013

 

 

7lrMean rating ranges from -10 (strong inﬂuence in the direction of formula feeding) to +10 (strong
inﬂuence in the direction of breastfeeding)

Among individual sources of inﬂuence, the (grand)mother was the strongest
inﬂuence in the direction of breastfeeding (Mean=4.091, SD=4.691). On systems of
inﬂuence, the healthcare delivery system was the strongest inﬂuence in the direction of

breastfeeding (Mean=5.343, SD=4.367). Differences between the Kent County study

119

sample and the Detroit study sample were statistically signiﬁcant on two individual
sources: my family (t=2.886, df=137, p=0.005) and my best ﬁiend (t=2.210, df=129, p-
0.029). On both of these sources, Kent County study participants indicated a stronger
mean inﬂuence of both of these sources in the direction of breastfeeding.

Additional comparisons were performed on the study samples by urban location
on two subscales of the Maternal Adjustment and Maternal Attitudes Questionnaire
(Kumar et al., 1984): Body Image and Attitudes toward Motherhood and the Baby. Table
19 displays items and mean ratings for each subscale, including item rank among the
twelve items. Each item was measured using a ﬁve-point Likert scale that ranged from

strongly disagree (rated with a 1) to strongly agree (rated with a 5).

120

Table 19: Overall Mean Item Ratings and Rankings on Subscales of the Maternal
Adjustment and Maternal Attitudes Questionnaire (Kumar, Robson and
Smith, I 984)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 Rank I Mean C I S.D.
Body Image Subscale A
I felt that my face was attractive. 1 3.34 1.12
I felt proud of my appearance. 2 3.22 1.25
I felt that I had gained too much weight. 3 3.14 1.30
I felt attractive. 4 3.08 1.14
I felt that my breasts were attractive. 5 3.06 1.11
I liked the shape of my body. 6 2.95 1.21
My body felt awkward and clumsy. 7 2.89 1.21
I felt that my body was soft and cuddly. 8 2.89 1.14
I felt that my breasts were too big. 9 2.76 1.21
I felt that my complexion was poor. 10 2.64 1.71
I felt that my breasts were too small. 11 1.92 0.83
I felt that my body smelled bad. 12 1.58 0.76
Attitudes toward Motherhood and the Baby Subscale B

l have felt proud of being a mother. 1 4.65 0.67
I have enjoyed caring for my baby’s needs. 2 4.61 0.64
I have enjoyed feeding my baby. 3 4.44 0.63
I have been feeling happy that I have a baby. 4 4.32 0.84
l have been wondering whether my baby will be healthy/normal. 5 3.24 1.14
I have had enough time for myself. 6 2.91 1.08
The thought of having more children appeals to me. 7 2.46 1.40
1 have felt that life has been more difﬁcult since the baby was born. 8 2.44 1.15
I have been wonying that 1 might not be a good mother. 9 1.99 1.07
I have worried about accidentally hurting my baby. 10 1.88 1.07
I have felt disappointed with motherhood. 11 1.59 0.71
l have regretted having the baby. 12 1.32 0.59

 

 

 

 

A Subscale items began “During my last 3 months of pregnancy. .
B Subscale items began “Since my baby was born...”
C Mean rating based on a five-point Likert scale (1=strongly disagree, 5=strongly agree)

As indicated in Table 19, respondents indicated stronger agreement with the body
image items that suggested a positive perception of one’s body. Only one item related to
negative body perception ranked near the top of the list, which stated “I felt that I had
gained too much weight” (Rank: 3, Mean=3.13, SD=1.30). Similarly, mothers in the

study sample were in stronger agreement with items that were positively oriented toward

121

motherhood. Once again, only one item that had a negative orientation ranked near the
top of the list of items: “I have been wondering whether my baby will be healthy and
normal” (Rank: 5, Mean=3.24, SD=1.15).

Comparisons were performed by urban location on the twelve body image items
and the twelve maternal attitudes items. There were four statistically signiﬁcant
differences on the body image items and two statistically signiﬁcant differences on the
maternal attitudes, which are displayed in Table 20.

Table 20: Statistically Signiﬁcant Differences on Body Image Items and Attitudes
toward Motherhood and the Baby Items by WIC Agency

 

 

 

 

 

 

 

 

I Kent County I Detroit
Study Sample Study Sample
I Mean C ] so. I Mean C S.D.
Body Image Subscale A
I felt that my face was attractive.
(t=-2.045, df=l65.391, p=0.042) 3'18 "0‘6 3'52 L207
I felt attractive.
(t=-2.038, df=177,p=0.043) 2.92 1.098 3.26 1.173
I felt that my complexion was poor.
(t=-2.016,df=177,p=0.045) 2.81 1.129 2.46 1.191
I felt that my body smelled bad.
(t=-1.998, df=/78, p=0.047) 1.68 0.762 1.46 0.749

 

 

 

 

 

Attitudes toward Motherhood and the Baby Subscale B

 

 

 

I have been feeling happy that I have a baby.

“:23”, df=181,p=0.019j 4.45 0.791 4.16 0.866
The thought of having more children appeals to me.

(t=2.047, df=178. 765, p=0.042) 2'66 L470 2'24 L283

 

 

 

 

 

 

A Subscale items began “During my last 3 months of pregnancy. ..”
B Subscale items began “Since my baby was born...”
C Mean rating based on a ﬁve-point Likert scale (1=strongly disagree, 5=strongly agree)

As indicated in Table 20, the Detroit study sample indicated greater overall
agreement with body image items that were positively oriented and were in greater
overall disagreement with body image items that suggested negative body image. In

particular, the Detroit study sample possessed a greater mean score on two positively

122

worded body image items: “I felt that my face was attractive” (Mean=3.52, SD=1.21) and
“I felt attractive” (Mean=2.92, SD=3.26). Further, although both study samples by urban
location indicated overall disagreement with items that indicated negative body image,
mean scores for Kent C ounty on two of these items was signiﬁcantly lower than the
Detroit sample: “I felt that my complexion was poor” (Mean=2.81, SD=1.13) and “I felt
that my body smelled bad” (Mean=1.68, SD=0.76).

Alternatively, the Kent County study sample indicated stronger overall agreement
with two items that were positively oriented toward motherhood than the Detroit sample.
The ﬁrst of these items, “I have been feeling happy that I have a baby,” possessed strong
agreement from both study samples. However, the Kent County study sample indicated
greater 0 verall a greement with the statement, with a s maller level of variance among
study participants’ rating on this item (Mean=4.45, SD=0.79). Both study samples by
urban location indicated slight disagreement with the statement “The thought of having
more children appeals to me,” however, the Detroit sample was in stronger disagreement
with the statement (Mean=2.24, SD=1.28) than the Kent County sample.

For the subsequent study objective, items from the above two subscales, in
addition to the beliefs about breastfeeding items, were assessed for measurement
reliability using exploratory factor analysis.

Study Objective 2:
Assessment of Measurement Reliability

The second objective of the study was to test measurement reliability of two
subscales of the Maternal Adjustment and Maternal Attitudes Questionnaire (Kumar et

al., 1984) and two scales of the Breastfeeding Attrition Prediction Tool (Janke, 1994)

123

among this sample of WIC participants. All items were phrased in the form of a
statement; respondents were asked to respond based on a ﬁve-point Likert scale
(1=strongly disagree, 5=strongly agree). Respondents’ answers were inspected for
patterned responses to ensure that respondents did not indicate the same answer for all
thirty items on the Breastfeeding Attrition Prediction Tool, or on all twelve items on each
subscale of the Maternal Adjustment and Maternal Attitudes Questionnaire. Respondents
that answered any set of items with patterned responses were removed from data analysis
for that scale only. After checking for patterned responses on each of these scales and
subscales, exploratory factor analysis was performed in order to determine whether sub-
factors of each concept could better explain positive and negative breastfeeding
sentiment, body image, and attitudes toward motherhood and the baby.

BreasJeeding Attrition Prediction Tool. Construct validity of each scale was
assessed separately using exploratory factor analysis. In particular, principal components
analysis was performed in order to extract one set of sub-factors related to positive
breastfeeding sentiment and a separate set of sub-factors related to negative breastfeeding
sentiment. First, the group of fourteen it ems o n the P ositive B reastfeeding S entirnent
Scale was factor analyzed using principal components analysis, which extracted two
factors with eigenvalues greater than one in the unrotated component matrix. To
encourage extraction of theoretically interpretable factors, a rotated component matrix
was produced using varimax rotation. Using this process, items separated into two factors
that focused on the advantages of breastfeeding and the disadvantages of formula

feeding. Items that clustered under each factor are outlined in Table 21.

124

Table 21: Items and Item Statistics for Factored Subscales of the Positive

Breastfeeding Sentiment Scale (Janke, I 994)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

li‘otated Cronbach’s
actor
Loadin Alpha
J
Factor I:
Belief in the benefits of breastfeeding 0‘874
Breastfeeding is best for the baby. 0.854
Breastmilk is healthy for the baby. 0.818
Breastfeeding helps you bond with your baby. 0.772
Breastfeeding is natural. 0.757
Breastmilk is more nutritious than infant formula. 0.727
Breastfeeding makes you closer to your baby. 0.674
Breastfeeding is more convenient than formula feeding. 0.554
Breastfeeding is personally satisfying. 0.543
Breastfeeding is more economical than formula feeding. 0.511
Factor II: 0 752
Belief in the infant health risks of formula feeding '
Formula fed babies tend to be overweight. 0.732
Formula fed babies are more fussy than breastfed babies. 0.690
Infant formula can cause constipation. 0.685
Infant formula can cause allergies. 0.683
Formula fed babies tend to get sick. 0.681

 

Overall, the two factors presented in Table 21 explained 52.1% of the variance in

positive breastfeeding sentiment. The ﬁrst factor had an eigenvalue 5.104, which

explained 36.5% of the variance. The second factor had an eigenvalue of 2.196 and

explained 15.7% of the variance. Based on the criteria established for removing items

from the factor structure, all of the items were retained. In particular, all items loaded at a

level of 0.30 or higher and only one item cross-loaded on both factors. This item,

“Breastfeeding is more economical than formula feeding,” had a factor loading of 0.511

on the ﬁrst factor and 0.407 on the second factor. Because the difference between the two

factor loadings was greater than 0.05 and the item had a better theoretical ﬁt with Factor

I, this item was placed on the ﬁrst factor.

125

 

In total, Factor I contained nine items with rotated factor loadings ranging from
0.511 to 0.854. The Cronbach’s alpha coefﬁcient was 0.874. These items seemed to relate
to mothers’ belief in the advantages of breastfeeding, including infant health, the mother-
infant relationship, personal satisfaction, and practical considerations. As a result, this
factor was labeled “Belief in the beneﬁts of breastfeeding.”

Factor H contained ﬁve items with rotated factor loadings ranging from 0.681 to
0.732. The Cronbach’s alpha coefficient was 0.752. Items in this factor appeared to relate
to the disadvantages of formula feeding, including infant health risks and effect of
formula feeding on the infant’s emotional disposition. Items in this category were
therefore grouped under the label of “Belief in the infant health risks of formula feeding.”

The Negative Breastfeeding Sentiment Scale (Janke, 1994) contained sixteen
items that measured mothers’ beliefs in the disadvantages of breastfeeding. Exploratory
factor analysis was performed on these sixteen items to assess construct validity. In
particular, p rincipal c omponents a nalysis w as p erformed, w hich e xtracted four factors
with eigenvalues greater than one in the unrotated component matrix. To encourage
extraction of theoretically interpretable factors, a rotated component matrix was produced
using varimax rotation. Using this process, items separated into four factors that focused
on the advantages of breastfeeding and the disadvantages of formula feeding. Items that

clustered under each factor are outlined in Table 22.

126

Table 22: Items and Item Statistics for Factored Subscales of the Negative

Breastfeeding Sentiment Scale (Janke, I 994)

 

Rotated

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Factor Cronbach’s
Alpha
. Loading
Factor I: O 739
Belief that breastfeeding limits lifestyle habits '
Mothers who FF get more rest than breastfeeding mothers. 0.751
Mothers who FF get back into shape sooner. 0.655
When you BF you never know if the baby gets enough milk. 0.602
Breastfeeding makes returning to work difficult. 0.536
Breastfeeding is more time consuming than formula feeding. 0.517
Formula fed babies are easier to satisfy than breastfed babies. 0.510
Factor 11:
' Belief that the breastfeeding mother is fully responsible 0,614
for infant feeding
7 Formula feeding is easier than breastfeeding. 0.637
Breastfeeding makes your breasts sag. 0.567
It is embarrassing to breastfeed in public. 0.551
Formula feeding gives you more freedom. 0.538
No one else can help you feed the baby when you breastfeed. 0.533
Formula feeding lets the father become close to the baby. A -----
Factor III: 0 634
Belief that it is difﬁcult to breastfeed outside the home ‘
Breastfeeding is messy. 0.798
Breastfeeding ties you down. 0.661
It is difﬁcult to ﬁnd places to breastfeed outside your home. 0.490
Factor IV: _____
Belief that it is painful to breastfeed
Breastfeeding is painful. 0.699

 

2rItem originally loaded on Factor III at 0.426

Overall, the four factors presented in Table 22 explained 49.2% of the variance in

negative breastfeeding sentiment. The ﬁrst factor had an eigenvalue 4.096, which

explained 25.6% of the variance. The second factor had an eigenvalue of 1.426 and

explained 8.9% of the variance. The third factor explained 7.7% of the variance in

negative breastfeeding sentiment and had an eigenvalue of 1.235. Lastly, although the

127

 

fourth factor only contained one item, its eigenvalue was 1.119 and the factor explained
7.0% of the variance in negative breastfeeding sentiment. Because this item explained a
considerable portion of the variance, the item was included in subsequent data analysis as
a separate characteristic

Following principal components analysis on the negative breastfeeding sentiment
items, two items warranted further exploration. One item cross-loaded on two factors:
“No one else can help feed the baby when you breastfeed.” The factor loading of this
item was 0.533 on Factor 11, whereas the factor loaded at -0.541 on Factor IV. According
to the criteria set for placing items, this item did not place deﬁnitively in either factor (as
the difference between the items was not greater than 0.05). A decision was made to
retain the item in Factor H on the basis of its better theoretical placement with items on
Factor H. A second item, “Formula feeding lets the father become closer to the baby,”
originally loaded on Factor 111. However, due to its relatively low factor loading (0.426)
and incongruence with the other three items in the factor, the item was placed in Factor 11
where it appeared to have a better conceptual ﬁt.

Consequently, in order to test the reliability of this and other factors of the
Negative Breastfeeding Sentiment Scale, Cronbach’s alpha was computed for each factor.
The Cronbach’s alpha coefficient for Factor I was 0.739. This factor contained six items
with factor loadings that ranged ﬁ'om 0.510 to 0.751. Items in the factor related to the
impacts of mothers’ decision to breastfeed on her ability to maintain characteristics of her
lifestyle. As a result, this factor was labeled “Belief that breastfeeding limits lifestyle

habits.”

128

Factor 11 included six items with factor loadings that ranged from 0.533 to 0.637.
Cronbach’s alpha coefﬁcient for this factor was 0.614. Items on this factor related to the
belief that burdens of and responsibilities for breastfeeding were borne exclusively by the
mother. This category of beliefs was therefore labeled “Belief that the breastfeeding
mother is fully responsible for infant feeding.”

Lastly, Factor III contained three items with factor loadings between 0.490 and
0.798. Cronbach’s alpha coefﬁcient for the factor was 0.634. Measurement reliability of
this factor was explored, which indicated that the Cronbach’s alpha coefﬁcient could be
increased to 0.675 with the removal of one item: “It is difficult to ﬁnd places to
breastfeed outside your home.” However, it was decided that the increase in Cronbach’s
alpha coefﬁcient was minimal compared to the theoretical removal of the item. In
particular, as the three items on Factor 111 related to issues of breastfeeding in public or
outside the home, this item was retained. Accordingly, these three items were grouped
under the label “Belief that it is difficult to breastfeed outside the home.”

The two factors that were extracted from the Positive Breastfeeding Sentiment
Scale and the three factors that were extracted from the Negative Breastfeeding
Sentiment Scale, in addition to the fourth solitary item about breastfeeding as painful,
were used in subsequent analysis predicting mothers’ infant feeding behavior.

Maternal Adjustment and Maternal Attitudes Questionnaire. Construct validity
of two subscales of the Maternal Adjustment and Maternal Attitudes Questionnaire were
investigated using exploratory factor analysis. Subscales were factor analyzed separately
in order to extract sub-factors related to the two distinct concepts. The ﬁrst subscale,

Body Image, contained twelve items that related to mothers’ perceptions of their body.

129

Principal components analysis was performed on these twelve items, which extracted two
factors with eigenvalues greater than one. In order to aid in the interpretability of the two
factors, principal components analysis was performed with varimax rotation. This
resulted in two factors that related to positive perceptions of one’s body and negative
perceptions of one’s body. The factor structure is presented in Table 23.

Table 23: Items and Item Statistics for Factored Subscales of the Body Image
Subscale A (Kumar, Robson and Smith, 1984)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Rotated Cronbach’s
Factor Alpha
Loading
Factor I: 0 856
Positive perceptions of one’s body ‘
I liked the shape of my body. 0.840
I felt attractive. 0.798
I felt proud of my appearance. 0.789
I felt that my face was attractive. 0.694
I felt that my breasts were attractive. 0.671
I felt that my body was soft and cuddly. 0.646
I felt that I had gained too much weight. -0.519
Factor I]: O 516
Negative perceptions of one’s body '
I felt that my breasts were too small. 0.744
I felt that my body smelled bad. 0.681
My body felt awkward and clumsy. 0.513
I felt that my complexion was poor. 0.504

 

 

 

 

”(Subscale items began “During my last 3 months of pregnancy. .
Ital: Item subsequently removed from factor structure

The two factors presented in Table 23 accounted for 48.6% of the variance in
body image. The eigenvalue for Factor I was 4.218, which explained 35.1% of the
variance. This factor contained seven items with factor loadings that ranged from -0.519
to 0.840. A Cronbach’s alpha coefficient was calculated for the seven items (0t=0.849);

further computations of Cronbach’s alpha coefﬁcient on a reduced set of items indicated

130

that the reliability coefﬁcient could be increased by removing “I felt that I had gained too
much weight.” This item was removed from the factor structure based on this result, in
addition to lacking conceptual fit with the remaining items. Review of the remaining
items suggested that items in Factor 1 related to positive body image; as a result, this
factor was labeled “Positive perceptions of one’s body.”

An additional 13.4% of the variance in body image was explained by Factor H,
which had an eigenvalue of 1.611. Cronbach’s alpha for this factor was 0.516, which was
maximized by retaining all four items that loaded on this factor. Items that loaded on this
factor suggested greater negativity toward one’s body; accordingly, Factor H was labeled
“Negative perceptions of one’s body.”

Of the twelve items entered for analysis, one item did not load on either factor: “I
felt that my breasts were too big.” However, two other items in the subscale addressed
breast size (“I felt that my breasts were too small” and “I felt that my breasts were
attractive”) and, as a result, this factor was removed from the factor structure and
subsequent analysis.

The other subscale of the Maternal Adjustment and Maternal Attitudes
Questionnaire (Kumar et al., 1984) that was factor analyzed for the current study
objective was the Attitudes toward Motherhood and the Baby subscale. This subscale
also contained twelve items, which were factor analyzed using principal components
analysis. This analysis extracted ﬁve factors with eigenvalues greater than one. Principal
components analysis with varimax rotation was performed to produce factors with greater

decipherability. Results of this analysis are outlined in Table 24.

131

Table 24: Items and Item Statistics for F actored Subscales of the Attitudes toward
Motherhood and the Baby Subscale A (Kumar, Robson and Smith, 1984)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

li‘otated Cronbach’s
actor A1 ha
Loading P

Factor I: 0 7 64
Personal satisfaction with motherhood '

I have regretted having the baby. 0.776

I have felt proud of being a mother. 0.762

I have felt disappointed with motherhood. 0.685

I have been feeling happy that l have a baby. 0.626
Factor 11: 0 757
Positive experiences with caring for the baby '

I have enjoyed feeding my baby. 0.856

I have enjoyed caring for my baby’s needs. 0.835
Factor III: 0 495
Worries about being a responsible mother '

I have been worrying that 1 might not be a good mother. 0.742

I have been wondering whether my baby will be healthy/normal. 0.663

I have worried about accidentally hurting my baby. 0.581
Factor IV: 0 524
Worries about impacts of motherhood on lifestyle habits '

I have had enough time for myself. 0.892

I have felt that life has been more difficult since the baby was born. 0.643
Factor V: _____
Thought of having more children appealing

The thought of having more children appeals to me. 0.899

 

 

 

 

A Subscale items began “Since my baby was born...”
Ital: Item subsequently removed from factor structure

As indicated in Table 24, the twelve items of the Attitudes toward Motherhood
and the Baby subscale loaded on ﬁve factors. Overall, the factor structure explained
68.1% of the variance in this construct. Factor I had an eigenvalue of 3.505, which
explained 29.2% of the variance. Factor loadings ranged from 0.626 to 0.726. The

Cronbach’s alpha coefﬁcient for the four items that loaded 0 n this factor w as 0.764.

132

Based on similar themes between these four items, this factor was labeled “Personal
satisfaction with motherhood.”

Factor H had an eigenvalue of 1.290, which accounted for 10.8% of the variance
in attitudes toward motherhood and the baby. The two items loaded on this factor had
factor loadings of 0.835 and 0.836, and the reliability coefﬁcient for these items was
0.757. As both items related to mothers’ experiences with taking care of the new baby,
this factor was labeled “Positive experiences with caring for the baby.”

The third factor extracted from the twelve items accounted for 10.0% of the
variance in the construct and had an eigenvalue of 1.205. Three items loaded on this
factor and based on the themes of the three items, the factor was labeled “Worries about
being a responsible mother.” Factor loadings of the three factors ranged from 0.581 to
0.742; Cronbach’s alpha coefficient for the three factors was 0.495.

Factor IV contained two items that related to mothers’ concerns about her own
routines, with factor loadings of 0.643 and 0.892. These two items explained 9.5% of the
variance in attitudes toward motherhood and the baby and had an eigenvalue of 1.139.
Cronbach’s alpha coefﬁcient for the factor was 0.525. As indicated, the items related to
the perceived impact of motherhood on mothers’ lifestyle. As a result, this factor was
labeled “Worries about impacts of motherhood on lifestyle habits.”

The ﬁnal factor only contained one item: “The thought of having more children
appeals to me.” The factor loading on this item was 0.899, which had an eigenvalue of
1.036 and explained 8.6% of the variance in the construct. Based on the amount of
variance explained by this item, the item was included as a distinct variable in subsequent

data analysis along with the four other factors extracted on this construct.

133

Study Objective 3:
Predicting Infant Feeding Behavior by Systems of External Inﬂuence

The third objective of this study was to predict infant feeding behavior on the
basis of three systems of external inﬂuence: family and ﬁiends, the healthcare delivery
system, and community sources. A regression model was hypothesized that p redicted
percentage of breastfeeding on three control variables and three systems of external
inﬂuence. The three control variables included race/ethnicity, education, and beliefs
about breastfeeding. As indicated, the three systems of social inﬂuence included family
and friends, the healthcare delivery system, and community sources. The regression
model predicting breastfeeding percentage on these characteristics was outlined in
Chapter 3.

The regression model was developed to predict infant feeding behavior. For the
model, infant feeding behavior was measured as a breastfeeding percentage. In particular,
mothers’ total number of feedings in a typical day was calculated as a sum of the number
of times the infant was fed formula, pumped breastmilk, and breastmilk from the breast in
a typical day. Next, mothers’ total number of feedings of breastmilk was calculated as a
sum of the number of times the infant was fed both pumped breastmilk and breastmilk
from the breast in a typical day. Breastfeeding percentage was computed as the number
of breastfeedings in a typical day divided by the number of total feedings in a typical day.
As a result, breastfeeding percentage could range from zero (no breastfeeding) to one
(exclusive breastfeeding).

To predict this dependent variable, twelve independent variables were entered

into a predictive model using stepwise regression. In order to control for socio-

134

demographic characteristics and beliefs about breastfeeding, these characteristics the
foremost to be entered into the model. Race/ethnicity was the ﬁrst of these characteristics
to be entered into the regression model. For the purposes of entering a categorical
variable into the predictive model, a dummy variable was created that measured
race/ethnicity. Because Black/African-American respondents composed the largest
proportion of the study sample, the race/ethnicity dummy variable was deﬁned as
Black/African-American (coded as a 1) or not Black/Aﬁican-American (coded as a 0).

Education level obtained was recoded into two dummy variables for the
regression model. The ﬁrst dummy variable indicated whether the respondent had
obtained a high school degree or the equivalent (coded as a 1) or had obtained less than a
12Lh grade education (coded as a 0). The second dummy variable indicated whether the
respondent had obtained any level of college education (coded as a 1) or had not obtained
any level of college education (coded as a 0). These two variables were entered into the
regression model following race/ethnicity in the order discussed above.

The fourth through ninth variables entered into the regression model were beliefs
about breastfeeding. As outlined in results presented above, ﬁve factors and one solitary
item related to positive and negative aspects of breastfeeding were extracted from
existing scales on positive and negative breastfeeding sentiment. Factor scores were
computed by taking the mean of all items that loaded on the factor. In the event that the
respondent did not respond to speciﬁc items on the factor, the mean factor score was
computed as the mean of all items on the factor that were answered.

Factor scores were entered into the regression model in order of highest reliability

coefﬁcient. As a result, factors were entered in the following order: (1) Belief in the

135

beneﬁts of breastfeeding (0t=0.874), (2) Belief in the infant health risks of formula
feeding (01:0.752), (3) Belief that breastfeeding limits lifestyle habits (01:0.739), (4)
Belief that it is difﬁcult to breastfeed outside the home (or=0.634), (5) Belief that the
breastfeeding mother is fully responsible for breastfeeding (01:0.614), (6) Belief that it is

painful to breastfeed (one-item a not applicable).

After controlling for socio-demographic characteristics and beliefs about
breastfeeding, three additional variables and four interaction variables were entered into
the regression model as predictors of breastfeeding percentage. The ﬁrst set of variables
included the family and friends system of inﬂuence, the healthcare delivery system of
inﬂuence, and community sources system of inﬂuence. Each system was composed of a
collection of individuals and groups. The family and friends system consisted of seven
individuals/groups that included the baby’s father, the (grand)mother, the (grand)mother-
in-law, the family, the best friend, other friends, and Breastfeeding Peer Counselors. The
healthcare delivery system contained three groups: doctor(s), the midwife, and WIC staff.
Lastly, community sources encompassed four groups that included other people at
school, other people at work, the community, and people in public places.

Inﬂuence scores were computed for each individual/group as a product of
perceived expectation and value, wherein a score of -10 indicated a very strong inﬂuence
in the direction of formula feeding and a score of +10 suggested a very strong inﬂuence
in the direction of breastfeeding. Using these inﬂuence scores, mean scores were
computed on each system of external inﬂuence by taking the mean of the inﬂuence scores
for the individuals/groups in that system. In the event that the respondent skipped any

individuals/groups or indicated “Don’t know or does not apply,” the mean score for the

136

system of external inﬂuence was computed as a mean of all items that were answered. So
long as the study participant indicated inﬂuence from at least one individual/group in the
system, a mean score was computed for that system of external inﬂuence. Mean score for
the family and friends system of inﬂuence was 3.423 (SD=3.68), 5.34 for the healthcare
delivery system of inﬂuence (SD=4.36), and 2.20 for community sources system of
inﬂuence (SD=3.44).

Mean scores for the three systems of external inﬂuence (family and ﬁiends, the
healthcare delivery system, and community sources) were entered into the regression
model, respectively. In addition, interaction variables between these three variables were
entered into the regression model to account for potential relationships between each of
these measures. Interaction between each set of variables was computed as a product of
the mean scores of each system of external inﬂuence. These variables were computed for
the interaction b etween ( 1) the family a nd friends s ystem a nd t he he althcare delivery
system, (2) the family and friends system and the community sources system, (3) the
healthcare delivery system and the community sources system, and (4) all three systems.
Interaction variables were entered into the regression model in the order presented.

Before running the regression model, a correlation matrix was constructed of the
dependent variable and independent variables entered into the regression model to
examine the strength and direction of relationships amongst the variables. The correlation

matrix is presented as Table 25.

137

.o>o. .ood o... .m .5o....:w.m 5.3.2.50 .23—
.96. .od 2.. a ESE—Em 8.3.8.50 l

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

no no no oo 3 mo No no no .o .o .o No 89.8w b22550 A
oo.. mmvd ovvd mood. Rod .o.d .N..o ohmd Emd wnmd Sad- wood- Sod .855... m.
mm. o: mv. om . 3. we. on . om. we. xv. to. a: bo>..on. 285.8: AN.
oo.. .omd Nmod mmod. mid ww..o .N.d :md no.d 2...? omod voud 6052......
3. mm. mo. om. .o. no. no. oo. oo. on. 3. mount... can DEE... A. .
oo.. omod. oood .o.d Sid mood mend om.d on..o- Sud- ohm. o 6052......
co. co. 8. no. co. co. No. mo. .0. mo. .85.». A
oo.. vo.d omod 5nd .ood wmod- wood- amod .ood mo..o- .A-o..o..om o.
m... of n: 3.. an. 2.. .2 ob. Nu. b._.£m:oam3. A
8.. 8...... 8v... 8...... 88... .8... 8.... m8..- 8.... ”3.23m a
do. of do. do. we. no. we. do. $5.50.. 0.3.5 AM.
8.. 2m... 2.... 8N... .8... c8..- 8.... 48... H3.2.3.
m: m... m... do. do. no. on. 2.03.1.
8.. .8... n2... 8.... 8.....- 28...- 8.... dream .5
3.. me. o... or. do. :— 82. 5.8:. «BEE... Ac
8.. 3...... 8.... 3.....- 8.... mum... Hanzom
o... F— .b. o: n: acacom
8.. 8N... 8...- and..- 8.8 H3.2.3. .m
on. o: m: N... amazon. 085 A
oo. . owed- hood- mm . .o 82.823. .V
e... 21. N: Bowed .028 :3... Am
oo.. om . .o- oo . .o- 62.82%“.
m .1 . :— . :3...oE<-:3_.....<\xom_m
8.. 82.. 5.6.2.3663. a
A... www.coouoa wc.coo...mmo..m A.
oo..
m. N. .. o. o w n o w v m m .

 

 

 

 

 

 

 

 

 

 

 

 

 

 

vs Acne: .536232 5 mazes: 252.2%}: 33 .5353wa 8.23: aezutteb 5.8.5 eauN "mN 933.

138

As indicated in Table 25, breastfeeding percentage was signiﬁcantly correlated
with eight of the independent variables. Of these eight statistically signiﬁcant
correlations, all correlation coefficients indicated that relationships were in the expected
direction based on previous research (Grossman et al., 1989; Kurinij et al., 1988; Park et
al., 2003; Pugh et al., 2001; Schwartz, Popkin, Tognetti, & Zohoori, 1995). Speciﬁcally,
there was a negative relationship between breastfeeding percentage and race/ethnicity,
indicating that Black/African-American mothers in the sample typically had lower
breastfeeding percentages. As well, breastfeeding percentage was negatively associated
with two factors related to negative breastfeeding sentiment. Positive associations existed
between breastfeeding percentage and two factors related to positive breastfeeding
sentiment and inﬂuence from two systems of external inﬂuence: family and ﬁiends and
the healthcare delivery system.

Among systems of external inﬂuence, bivariate associations were statistically
signiﬁcant between each pair of systems. The strength of relationships of each of these
pairs was moderate, wherein the correlation coefﬁcient for family and ﬁiends and the
healthcare delivery system was 0.591 (p=0.000), family and friends and community
sources was 0.448 (p=0.000), and the healthcare delivery system and community sources
was 0.482 (p=0.000). The strengths of these relationships between pairs of systems
supported the decision to include interactions between each system as independent
variables in the regression model.

However, based on the results presented in Table 25, the decision was made to
remove c ommunity s ources s ystem o f e xtemal influence 0 n the b asis o f its pair-wise

match with other variables in the correlation matrix. In particular, lack of responses from

139

 

 

 

 

study participants severely limited the potential sample size for the proposed regression
model, which had a minimum pair-wise match of 61 study participants on speciﬁc
independent variables. Because the inclusion of this variable in the regression model
would have limited sample size for the analysis, the community sources system of
inﬂuence and its interactions were removed from the proposed model. The revised

proposed model is presented in Figure 3.

Figure 3: Revised Regression Model A

 

 

IFB = a 'l' b) (R) + b2 (E) + b3 (BBF) '1' b4 (FF) '1' b5 (HDS) + b6 (FF'HDS) '1' 6

Where:

IFB = Current Infant Feeding Behavior
R = Race

E = Education

BBF = Beliefs about Breastfeeding

FF = Family and Friends (System of External Inﬂuence)

HDS = Healthcare Delivery System (System of External Inﬂuence)

a = Intercept

bx = Average change in infant feeding behavior associated with one unit change in
the independent variable, holding all other independent variables constant

e = Error

 

Following a n e xploration o f t he r elationships b etween the proposed dependent
and independent variables, the proposed regression model was constructed using stepwise
regression. As outlined, variables were entered to control for socio-demographic

characteristics and beliefs about breastfeeding ﬁrst, followed by systems of external

140

 

 

inﬂuence and their interactions. Table 26 displays results of stepwise multiple regression

analysis on breastfeeding percentage.

Table 26: Results of Stepwise Regression Analysis Predicting Breastfeeding
Percentage on Systems of External Inﬂuence

 

 

 

 

 

 

 

 

 

 

 

Variable 1 Variable 2 Variable 3 Variable 4 Ad
a Belief (+): Belief (-): Belief (-): Belief (-); Rf.
Beneﬁts Responsibility Painful Lifestyle
-1.09 0.343
Model 1 B 0.255
I3 0.505
-0.54 0.352 -0. 167
Model 2 B 0.301
I} 0.519 -0.237
B -0.36 0.345 -0. 150 0058
Model 3 0.317
I3 0.509 -0.213 -0. 147
B -0.37 0.317 -0.201 -0.072 0.109
Model 4 0.333
I3 0.467 -0.285 -0. 184 0.174

 

 

 

 

 

 

 

 

 

 

 

As indicated, there were four independent variables were statistically signiﬁcant
in forecasting breastfeeding percentage: belief in the beneﬁts of breastfeeding, belief that
the breastfeeding mother is fully responsible for infant feeding, belief that breastfeeding
is painful, and belief that breastfeeding limits lifestyle habits. The adjusted R2 for the
model presented was 0.333, indicating that 33.3% of the variance in breastfeeding
percentage was explained by the four signiﬁcant predictors.

Three of the beliefs variables predicted breastfeeding percentage in the expected
direction. In particular, belief in the beneﬁts of breastfeeding was a positive predictor of
breastfeeding percentage, indicating that a stronger belief in breastfeeding beneﬁts
predicted higher breastfeeding percentage. Based on the B coefﬁcients, this variable was
the strongest predictor of breastfeeding percentage in the ﬁnal regression equation

(B=0.467). Similarly, belief that the breastfeeding mother is fully responsible for infant

141

feeding and belief that breastfeeding is painful were related to breastfeeding percentage
in the expected direction. Both were negative predictors of infant feeding behavior,
suggesting that stronger beliefs in these dis advantages 0 f b reastfeeding w ould p redict
lower breastfeeding percentages.

However, the ﬁnal statistically signiﬁcant predictor of breastfeeding percentage
was not in the anticipated direction. This variable, the belief that breastfeeding limits
lifestyle habits, was expected to have predict breastfeeding percentage in a negative
direction. That is, the stronger the mother’s belief in limitations on her lifestyle, the lower
her predicted breastfeeding percentage. However, this variable had a positive coefﬁcient
in the ﬁnal model (B=0.174; B=0.109), indicating that the opposite relationship was
suggested by the data.

All four statistically signiﬁcant predictors were factors of beliefs about
breastfeeding, suggesting the role of mothers’ beliefs in predicting infant feeding
behavior. However, based on research suggesting the inﬂuence of external sources on
infant feeding behavior (Matich & Sims, 1992), it was expected that at least one of the
two systems of external inﬂuence (family and ﬁ'iends or the healthcare delivery system)
or their interaction would be a signiﬁcant predictor of infant feeding behavior. To further
explore this result, the proposed regression model was tested for multicollinearity to
examine whether any of the independent variables were perfectly correlated with one or a
combination of other independent variables. The recommended method for testing
variables for multicollinearity is to regress each independent variable on the remaining
independent variables and to inspect the adjusted R2 of each model for its proximity to

1.0 (Lewis-Beck, 1980). Using this method for testing multicollinearity, the model

142

predicting family and friends system of inﬂuence had the highest adjusted R2 at 0.839. In
the event of close proximity to 1.0, Lewis-Beck (1980) recommends proposing two
alternative models, one of which discards the offending variable and the other of which
regresses the dependent variable on the discarded variable.

F arnily and friends was removed from the original regression model and analysis
yielded the same results as presented in Table 26. Additionally, an alternative regression
model was proposed that removed the six variables associated with beliefs about
breastfeeding, as these variables composed the majority of signiﬁcant predictors of the
family and ﬁiends system of inﬂuence. For this model, socio-demographic variables were
retained to control for differences in race/ethnicity and education level 0 btained. The
three independent variables measuring family and friends system of external inﬂuence,
the healthcare delivery system of external inﬂuence, and there interaction were entered in
the alternative regression model. Results of regression analysis on the alternative
proposed model are presented in Table 27.

Table 27: Results of Stepwise Regression Analysis Predicting Breastfeeding
Percentage on Systems of External Inﬂuence (Alternative Model)

 

 

 

 

 

 

 

Variable 1 Variable 2 2
a Family and Friends Race/Ethnicity: Adl- R
System of External Inﬂuence Black/African-American
0.168 0.043
Model 1 B 0.133
I3 0.372
B 0.249 0.040 0160
Model 2 0.162
I3 0.348 -0.190

 

 

 

 

 

 

 

The alternative regression model was more consistent with literature that
emphasizes the inﬂuence of family and friends (Dix, 1991; Guttman & Zimmerman,

2000) and race/ethnicity (Baranowski et al., 1983; Blum, 1999; Oyeku, 2003; Park et al.,

143

 

2003; Weller & Dungy, 1986) on mothers’ feeding behaviors. This regressionmodel
possessed two statistically signiﬁcant predictors of breastfeeding percentage, which
included one variable associated with systems of external inﬂuence. Indeed, aﬁer
removing variables related to breastfeeding beliefs from the proposed model, the family
and friends system of external inﬂuence was a statistically signiﬁcant predictor of infant
feeding behavior. In particular, according to the ﬁnal regression equation, a one point
increase in mean inﬂuence from family and friends predicted a 4% increase in
breastfeeding percentage when race/ethnicity was held constant.

As well, race/ethnicitywas a signiﬁcant predictor int he alternative regression
equation. In particular, being Black/Aﬁican-American predicted a 16% decrease in
breastfeeding percentage when the family and ﬁiends system of external inﬂuence was
held constant.

However, this alternative model explained considerably less of the variation in
breastfeeding percentage than the original proposed model. The adjusted R2 for the
alternative model was 0.162, suggesting that 16.2% of the variance in breastfeeding
percentage was explained by these two predictors. As indicated on the original model,
33.3% of the variance in breastfeeding percentage was explained by one positive and
three negative beliefs about breastfeeding. This suggests that the original model had a
better overall ﬁt with the data, and also that the socio-demographic variables need to be

considered in concert with systems of external inﬂuence.

144

1""‘~'-‘"-""l

 

Study Objective 4:
Predicting Infant Feeding Behavior by Body Image and Attitudes toward Motherhood

The fourth and ﬁnal objective of the current investigation was to predict infant
feeding behavior by mothers’ positive and negative perceptions of their bodies and by
adjustment to motherhood and the mothering role. A regression model was hypothesized
that predicted percentage of breastfeeding on the same three control variables as
presented above, in addition to two independent variables and their interaction. These two
independent variables, as indicated, included body image and attitudes toward
motherhood and the baby. The regression model predicting breastfeeding percentage on
these characteristics was outlined in Chapter 3.

The structure of the proposed model resembled that proposed for the third
objective. For the model, the dependent variable was measured as breastfeeding
percentage. Computation of the percentage was described above; to review, the
percentage was a proportion of total times the infant was fed breastmilk in a typical day
divided by the total times the infant was fed in a typical day. Again, because it was a
proportion, this number could range ﬁ'om zero (no breastfeeding) to one (exclusive
breastfeeding). Additionally, several variables were entered into the model at the outset to
control for race/ethnicity, education level obtained, and beliefs about breastfeeding.
Computation of these variables was described when they were initially introduced in the
third study objective.

Six independent variables and one solitary item followed the control variables
into the predicted regression equation. The ﬁrst two of these independent variables were

factors extracted from the B ody Image s ubscale in t he s econd s tudy o bjective. T hese

145

 

 

factors included positive perception of one’s body and negative perception of one’s body.
The next four independent variables were factors extracted from the Attitudes toward
Motherhood and the Baby subscale in the second study objective. Factors extracted from
this subscale included personal satisfaction with motherhood, positive experiences with
caring for the baby, worries about being a responsible mother, and wonies about impacts
of motherhood on lifestyle habits. An additional item from this subscale failed to load on
any of the four factors and was included in the regression model as a separate variable:
“The thought of having more children appeals to me.” Items on each subscale were rated
on a ﬁve-point Likert scale that ranged from strongly disagree (coded as a 1) to strongly
agree (coded as a 5). Mean scores were computed for each of the factors by taking the
mean of item ratings that comprised each factor.

A correlation matrix was constructed of the dependent variable and all
independent variables entered into the regression model to examine the strength and
direction of relationships between the variables. As correlations were presented among
control variables in Table 25, Table 28 presents the correlation matrix on variables that

were unique to the fourth study objective.

146

 

 

 

Aowao :5: 5 625.259 ma 0.5.:
.05. .ood 2.. 5 58.0.5? 5.5.2.50 .522.
.05. .od 5... 5 585%? 5.5.2.50 I

 

 

 

 

 

 

 

 

 

 

..m . no. mo. 3. no. we. no. we. 3. mo. 5:5an 55...... 25.). A...
no. .o .w . .o mmod mm . .o ooNd- mmod vo. .o owod voNd. mood 65:555.).
.0. do. 00. mo. 50. me. ..o. we. do. oo. 2.05.... I 5.2.5.5 Ao.
nmod oood- mood- wwod- mmod nu . d. w. ..o- .ood m . od. .mod- 659.5503.
.0. do. do. we. do. a: we. we. 3. ..o. b._....m5..m2 1 5.2.53 Aw.
oood omod 3. d. .wod- ovod vmud. Nb . d. m. ..o m . od Nb . d. H.55....25—2
oo. mg no. no. me. ..o— oo. we. no. 3. 525.595 03:5; Av.
m . od NAN .o .o . .o Sud vo . d- voNd .w . .o ovod- . o. .o- vwod U52.2.5.2
on . no. we. oo. ..o. me. ..o. no. co. 3. 5:83.28 5522. Am.
.N . d- mm . d- m .od- 0 . od .nod- n . od- ovod- mmod oood- wood ”55:55.02
do. we. 3. ..o. ..o. ..o. do. do. no. 3. 5.58.5; 3.532 Am.
no. d- 3. . d. v. . d- 3 . d. N. ..o mm. d- nmod- mmod mo. .o- hood- .ume. Atom.
.0. do. no. .6. do. we. do. we. no. we. 5.5529. o>...mon.
oo..? mm. o .Nod- mm. o oood- .ood o.od ..od mood m. ..o- Howes: Atom 2.
o. o w n o n v m N .
l S k S w e e A e
a .1 w A... h m .m w W m
P s e .I t c c c
. . n cl .L .1 B l a r.
s.) 0 c .. w .. e o l e
“1.x m. H A; h \+1 m .m B D.
.w h M. at... m H. S x w. .m.
.Inlv \.I., .AIV. .w m h .m d
B l.\ x...) .91» m d .m .Wo .m w
or IV B 0 B m H h mu
e f F c .. a s
H .m.» \F U n H w
R ial» + w .m % r.
B Wm M R B
M m

 

 

 

 

 

 

 

 

 

 

 

M 535:. 555.5% 3. 533.5; 253.5535 3.3 2535855 8.55:. 3855.250 .595 95N .mN 53.5.

 

147

 

 

 

 

 

 

 

 

 

we. we. we. we. we. we. : we. 5:555 5.2.5 053.
. wuc... god. ecod wmo.o- weed- ee..o H.xxz..§.5$. A».
oe. we. we. be. ..e. be. 03.3,... I 8:53
_ m2... 22.. 8...... as... 82.. 685502 8.
2.. we. we. be. he. 5.5.253. I 8an
_ «3.... Rod- 8:. 83- 685282 .2
we. ee— ee. ee. 805.598 35.85
_ SN... .25- 8:. 685502 .3
we. m e . m e . cozoﬂmzmm .acoEom
_ 33. «8... 68552 .2
we. we. 5.5855 3.2502
_ .mp2... owes boom 3.
oe. 5.5855 33.55
. ”Momma—.133. 2.
h. e. e. v. m. N. ..

 

 

Motherhood: More children appealing

 

Motherhood: Worries — lifestyle

 

Motherhood: Worries - responsibility

 

Motherhood: Positive experiences

 

Motherhood: Personal satisfaction

 

Body Image: Negative Perception

 

Body Image: Positive Perception

 

 

m 5353.. aemmnmxuuﬂ a. 5335\— 3535535 35 353.55Q5 8.53: 5.55559 53.5 95N "39.—=35“! wN 033—.

148

Following a n e xploration o f t he r elationships b etween the proposed dependent
and independent variables, the proposed regression model was constructed using stepwise
regression. As outlined, variables were entered to control for socio-demographic
characteristics and beliefs about breastfeeding ﬁrst, followed by body image factors and
factors of attitudes toward motherhood and the baby. Table 29 displays results of
stepwise multiple regression analysis on breastfeeding percentage.

Table 29: Results of Stepwise Regression Analysis Predicting Breastfeeding
Percentage on Body Image and Attitudes toward Motherhood and the

 

 

 

 

 

 

 

 

 

 

 

 

 

Baby
Variable 1 Variable 2 Variable 3 Variable 4 Ad
a Belief (+): Belief (-): Race/Ethnic: Belief (-): R}
Beneﬁts Painful Black/A-A Responsibility
B -l.12 0.347
Model 1 0.279
p 0.533
B -0.87 0.348 -0.071
Model 2 0.308
[3 0.534 -0.l85
B -0.71 0.321 -0.066 -0. 137
Model 3 0.330
[3 0.494 -0.172 -0. 166
B -0.42 0.330 -0.057 0136 -0.103
Model 4 0.347
[3 0.507 -0.l49 -0. 164 -0. 147

 

 

 

 

 

 

 

 

 

As indicated in Table 29, the predicted model explained 34.7% of the variance in
breastfeeding percentage. There were four statistically signiﬁcant predictors of
breastfeeding percentage in this model: belief in the beneﬁts of breastfeeding, belief that
breastfeeding is painful, race/ethnicity (Black/African-American), and belief that the
breastfeeding mother is fully responsible for infant feeding. All signiﬁcant predictors
forecasted breastfeeding percentage in the expected direction. That is, belief in the
beneﬁts of breastfeeding was a positive predictor of breastfeeding percentage. In the ﬁnal

model predicting breastfeeding percentage, a one unit increase in mean belief in the

149

beneﬁts of breastfeeding was predicted to increase breastfeeding percentage by 33.0%,
holding all other independent variables constant.

The other three variables were negative predictors of breastfeeding percentage,
including two negative breastfeeding sentiment factors and race/ethnicity. In particular,
the belief that breastfeeding is painful predicted a 5.7% decrease in breastfeeding
percentage with every one unit increase in mean rating of the belief, when all other
predictors were held constant. The other negative breastfeeding sentiment factor, belief
that the breastfeeding mother is fully responsible for infant feeding, predicted a 10.3%
decrease in breastfeeding percentage for every one unit increase in mean rating of the
belief. Lastly, race/ethnicity was a signiﬁcant predictor of the dependent variable and was
negatively related to breastfeeding percentage such that being Black/Aﬁican-American
predicted a 13.6% decrease in breastfeeding percentage, holding all other variables
constant.

As results of Study Objective 3 suggested, beliefs about breastfeeding were once
again signiﬁcant predictors of mothers’ infant feeding behavior. However, once again,
the ﬁnal regression equation obtained did not contain any signiﬁcant predictors of
breastfeeding percentage by the factors related to body image or attitudes toward
motherhood and the baby, as hypothesized for the fourth study objective. Independent
variables were investigated for multicollinearity by regressing each independent variable
on the remaining independent variables. Using this method advised by Lewis-Beck
(1980), multicollinearity was not detected among any of the independent variables. In
particular, the maximum adjusted R2 was 0.381 (obtained when belief that breastfeeding

limits lifestyle habits was regressed on all independent variables), which was not in close

150

enough proximity to 1.0 to suggest the presence of multicollinearity. Therefore, to
explore the exclusion of body image and attitudes toward motherhood and the baby from
the ﬁnal regression equation, an alternative regression model was proposed that removed
beliefs about breastfeeding from the predicted model. This alternative model retained all
control variables, including race/ethnicity and two education variables. The model also
included the two b ody im age factors, four a ttitudes toward m otherhood a nd the baby
factors, and one solitary item related to the appeal of having more children. Results of
regression analysis on this alternative model are presented in Table 30.

Table 30: Results of Stepwise Regression Analysis Predicting Breastfeeding
Percentage on Body Image and Attitudes toward Motherhood and the

 

 

 

 

 

 

 

 

 

Baby (Alternative Model)
Variable 1 Variable 2 Variable 3
Race/Ethnicity: Motherhood: . Ad -
a .
Black/African- Worries — 55:31:62,232} R}
American Responsible Mother g
B 0.41 0250
Model 1 0.087
p -0.305
B 0.667 -0.250 0109
Model 2 0.124
[3 -0.304 -0.206
B 0.567 -0.251 -0.093 0.128
Model 3 0.142
[3 -0.305 -0.l76 0.156

 

 

 

 

 

 

 

 

The ﬁnal alternative regression model had three statistically signiﬁcant predictors
of breastfeeding percentage, as outlined in Table 30. Among these three predictors, two
independent variables were demographics and the ﬁnal predictor was one factor of
attitudes toward motherhood and the baby: worries about being a responsible mother. On
this factor, a one point increase in mean rating on the factor predicted a 9.3% decrease in
breastfeeding percentage, holding all other variables constant. The other two signiﬁcant

predictors, race/ethnicity and some college education, were both associated with

151

 

 

breastfeeding percentage in the direction that would be predicted. That is, being
Black/African-American predicted a 25.1% decrease in breastfeeding percentage while
having obtained at least some college level education predicted a 12.8% increase in
breastfeeding percentage.

Once again, the alternative regression equation explained a considerably smaller
amount of the variance in breastfeeding percentage than the original regression equation,
wherein the above alternative regression equation had an adjusted R2 of 0. 142 (explaining
14.2% of the variance). Compared to the adjusted R2 of the original regression equation
for the current study objective of 0.347, the alternative equation explained a great deal
less of the variance in the dependent variable. This result suggests the potential for
speciﬁcation error in predicting breastfeeding percentage if mothers’ beliefs about
breastfeeding are excluded from the prediction model. A detailed discussion of this and

other results of data analysis follow in Chapter 5.

152

Chapter 5
DISCUSSION

The purpose of the current investigation was to explore low-income mothers’
infant feeding behavior in the context of multiple discourses on femininity and multiple
contexts of external inﬂuence. The study relied on Foucaultian theory to examine these
relationships, in which infant feeding was argued as a discursive ﬁeld that encompassed
numerous ways of understanding infant feeding choice and referenced the authority of
multiple systems of external inﬂuence. In order to explore these associations, four
research objectives guided the study, which were explored using primary data obtained
from questionnaires administered to WIC clients at sixteen urban WIC clinics and
secondary data obtained from the WIC Division of the Michigan Department of
Community Health.

To provide a description of the study sample, the ﬁrst research objective of the
study characterized the overall study sample on several characteristics related to infant
feeding and other study-related factors and compared these characteristics by urban
location. Based on initial discussions with WIC administrators at the state and local level,
it was ascertained that differences existed between WIC participants from Kent County
and Detroit on two characteristics in particular: racial/ethnic composition and
breastfeeding rates of each WIC agency. Hence, differences in the study sample by urban
location on these two characteristics were expected prior to data collection. Nonetheless,
in o rder to de velop m odels for p redicting infant feeding behavior among low-income
women across urban locations, the ﬁrst study objective compared the study sample by

urban location to aid in subsequent study objectives.

153

 

 

A series of characteristics were compared between Kent County WIC participants
and Detroit WIC participants, including race/ethnicity and infant feeding behavior, with
the understanding that the study samples would most likely differ on these, and
potentially other, characteristics. Indeed, racial/ethnic composition and infant feeding
behaviors were signiﬁcantly different between study participants from these two
communities. In particular, the Detroit study sample was composed primarily of
Black/African—American study participants, consistent with racial/ethnic composition of
other clients at the same location. Further, characteristics associated with breastfeeding
were signiﬁcantly different between study participants in the two communities, wherein
larger proportions of study participants from Kent County had initiated breastfeeding and
were currently breastfeeding. These differences were partially substantiated by
differences between the study samples on breastfeeding beliefs, on which Kent County
study participants indicated stronger agreement with positive breastfeeding sentiment
beliefs and the Detroit study participants indicated stronger agreement with negative
breastfeeding sentiment beliefs. Thus, as expected, r esults o f t he firsts tudy o bjective
suggested there were several signiﬁcant differences between study samples by urban
location. In order to account for these differences in the subsequent predictive models,
characteristics that were statistically different between study participants in the two urban
locations, and that were suggested by previous literature to be inﬂuential in predicting
feeding behavior, were included in the regression models as independent variables.

The second study objective aimed to assess measurement reliability of two scales
of the Breastfeeding Attrition Prediction Tool (Janke, 1994) and two subscales of the

Maternal Adjustment and Maternal Attitudes Questionnaire (Kumar et al., 1984) using

154

exploratory factor analysis. As the original instruments had been tested among study
samples that differed o n s everal c haracteristics from the s tudy s ample for t he c urrent
study, this objective attempted to extract sub-factors from the original scales that could
increasingly aid in predicting infant feeding behavior among low-income populations. It
was hoped that sub-factors would emerge that might be more meaningful to the topic of
infant feeding among low-income populations.

Factor analysis on the Positive Breastfeeding Sentiment Scale extracted two
factors related to positive breastfeeding sentiment, one of which related to the advantages
of breastfeeding and the other of which related to the disadvantages of formula feeding.
On this scale then, items clustered into positively-worded items and negatively-worded
items. A considerable proportion of the variance in positive breastfeeding sentiment was
explained by these two factors, of which a large portion was explained by mothers’
beliefs in the beneﬁts of breastfeeding alone. This factor was also a strong predictor of
breastfeeding percentage in subsequent regression analyses, suggesting that mothers’
beliefs in various beneﬁts of breastfeeding may be signiﬁcant in predicting feeding
behavior.

Factor analysis of the Negative Breastfeeding Sentiment Scale extracted three
factors and one solitary item, which explained a large portion of the variance in negative
breastfeeding sentiment. Interestingly, the factors extracted on this scale seemed to relate
to mothers’ conﬂict with dominant constructions of motherhood in the context of their
feeding decisions. Dominant discourses of maternal femininity, as theorized by multiple
academics (Blum, 1993; Traina, 2000; Wall, 2001; Young, 1990) and conveyed in the

empirical work of several researchers (Bialeschki & Michener, 1994; Blum, 1999; Blum

155

 

& Deussen, 1996; Murphy, 2000; Schmied & Lupton, 2001), may be constructed to
emphasize c haracteristics of generosity, nurturance, and concern for the well-being of
others. The three factors and one solitary item extracted from the Negative Breastfeeding
Sentiment Scale (Janke, 1994) appear to indicate a degree of incompatibility between
these dominant constructions of motherhood and mothers’ ability to conform to those
expectations in the context of infant feeding.

For example, consistent with literature suggesting the constraining aspects of
breastfeeding on mothers’ lifestyle behaviors (Dennis & Faux, 1999; Guttman &
Zimmerman, 2000; Li et al., 2002), one of the factors extracted from this scale contained
a range of items related to limits on mothers’ lifestyle habits. Although dominant
constructions of motherhood may devalue mothers’ entitlement to an identity external to
motherhood and a lifestyle external to the family (Bialeschki & Michener, 1994;
Freysinger, 1994; Kelly & Kelly, 1994; Shaw, 1985, 1992), this factor indicates the
potential constraints and contradictions inherent in such a limited view of motherhood.
As indicated, other factors extracted on this scale were thematically similar to this ﬁrst
factor in addressing mothers’ conﬂict with dominant discourses of maternal femininity.

On the Negative Breastfeeding Sentiment Scale, it is also worthwhile to note that
one additional item was developed and tested that read, “It is difﬁcult to ﬁnd places to
breastfeed outside your home.” The decision to develop and include this item in data
collection was based on literature that suggested breastfeeding in public may inﬂuence
mothers’ decisions about infant feeding and experiences of infant feeding (Sheeshka et
al., 2001; Stearns, 1999). In factor analysis, the item had a relatively small factor loading;

however, this item loaded on a factor that was subsequently labeled “Belief that it is

156

 

difﬁcult to breastfeed outside the home,” suggesting that the addition of this item may
have improved reliability of the instrument. Based on this result, it may be advisable for
future research to explore the inclusion of this item to assess the contribution of the item
to measurement reliability of the instrument.

Factor analysis on the two subscales of the Maternal Adjustment and Maternal
Attitudes Questionnaire (Kumar et al., 1984) extracted two factors related to body image
that focused on positive perception of one’s body and negative perception of one’s body.
Once again, items on each factor clustered into positively-worded items and negatively-
worded items related to body image. This result, mirrored in the item patterns on factors
extracted on the Positive Breastfeeding Sentiment Scale, indicates that combining
positively— and negatively-worded items on one scale may be problematic in attempting
to extract factors of these constructs. More speciﬁcally, in employing factor analysis to
extract sub-factors of positive breastfeeding sentiment and body image, the intention was
to extract themes related to meanings—further analysis and potential revision of these
items may be advisable.

Factor analysis of items related to attitudes toward motherhood extracted a greater
number of factors that clustered by conceptually interpretable themes and one solitary
item related to mothers’ orientation toward having more children. Among the extracted
factors of this scale, two related to mothers’ satisfaction with her maternal role: “Personal
satisfaction with motherhood” and “Positive experiences caring for the baby.” These
factors were consistent with matemalist conceptions of motherhood and infant feeding
(Blum, 1999). Two other factors extracted from this subscale suggested constraining

aspects of the maternal role, including “Worries about being a responsible mother” and

157

 

 

“Worn'es about impacts of motherhood on lifestyle habits,” which mirrored themes
expressed in the factors extracted ﬁ'om the Negative Breastfeeding Sentiment Scale.
Additionally, similar themes have emerged in previous empirical research relating
mothers’ infant feeding experiences to dominant constructions of motherhood (Blum,
1999; Blum & Deussen, 1996; Murphy, 2000).

Following factor analysis, extracted factors were used in subsequent regression
analysis to predict infant feeding behavior by (1) systems of external inﬂuence and (2)
body image and attitudes toward motherhood and the baby. The theoretical ﬁamework of
the current investigation drew upon Foucaultian theory to investigate the translation of
discourses of maternal and sexual femininity into mothers’ feeding behaviors. Following
from these discourses, the Foucaultian perspective of the current study situated
knowledge in the context of power, wherein diverse ways of knowing and understanding
infant feeding were conjectured to convey authority of individuals and institutions. Based
on mothers’ perceptions of this authority, it was suggested that the viewpoints of actors
external to the mother-infant dyad may convey authority to the extent that mothers’ infant
feeding behavior was associated with the expectations of these individuals. To explore
these theoretical relationships, two regression models were proposed that predicted
mothers’ feeding behavior on discourses of femininity and external authority.

However, based on the results of both regression analyses, signiﬁcant predictors
of breastfeeding percentage were neither discourses of maternal femininity or sexual
femininity, nor were systems of external inﬂuence signiﬁcant predictors of feeding
behavior. On the contrary, results indicated that signiﬁcant predictors of mothers’ infant

feeding behavior were primarily positive and negative beliefs about breastfeeding. More

158

 

 

speciﬁcally, on both models predicting breastfeeding percentage, belief in the beneﬁts of
breastfeeding was the strongest predictor of the dependent variable. In particular, stronger
belief in the beneﬁts of breastfeeding predicted a larger breastfeeding percentage. The
strength of this characteristic in predicting infant feeding behaviors has been supported
by previous empirical research (Jones, 1986, 1987; Losch et al., 1995).

Two other beliefs about breastfeeding were significant predictors of breastfeeding
percentage on both models and both were negatively associated with the dependent
variable. These negative breastfeeding s entiment factors, the b elief t hat the m other is
fully responsible for infant feeding and the belief that breastfeeding is painful, were both
negatively associated with breastfeeding percentage. Literature supports the signiﬁcance
of these predictors, suggesting that mothers’ concerns about themselves and their
lifestyles are important, although potentially overlooked, factors in mothers’ decisions
about their infant feeding behaviors (Arafat et al., 1981; Murphy, 2000). Further, as
indicated, such factors related to negative breastfeeding sentiment seemed to relate to
dominant constructions of motherhood in the context of infant feeding behavior.
Therefore, although these factors were categorized as negative beliefs about
breastfeeding, they may suggest that discourses of maternal femininity were in fact
signiﬁcantly associated with mothers’ feeding behavior.

Consistent with the literature (Baisch et al., 1989; Baranowski et al., 1983; Park et
al., 2003), infant feeding behavior had a negative correlation with race/ethnicity such that
lower breastfeeding percentage was associated with being Black/African-American. This
was also a signiﬁcant predictor of infant feeding behavior in the second regression

equation predicting feeding behavior on maternal body image and attitudes toward

159

motherhood and the baby, wherein being Black/African-American was a negative
predictor of breastfeeding percentage. Based on the differences between the Kent County
study sample and the Detroit study sample on both of these characteristics, it is likely that
differences in breastfeeding rates between the larger WIC client groups at each location
are at least partially related to the differences in racial/ethnic composition at each
location.

The combination of positive and negative predictors of infant feeding behavior on
both regression models supports the Foucaultian contention that discourses and bodies of
knowledge on infant feeding are not necessarily complementary. Quantitative exploration
of such concepts provided a unique approach to Foucaultian theory; relying on
quantitative methodology, this approach yielded several ﬁndings that support his
theoretical framework. Indeed, although mothers’ belief in the beneﬁts of breastfeeding
was a signiﬁcant positive predictor of breastfeeding percentage, mothers’ understandings
of the limitations of breastfeeding on personal lifestyle and independence indicate that
some bodies of knowledge surrounding breastfeeding, motherhood, and independence
may be contradictory and constraining. Foucaultian theory supports such a divergence of
bodies of knowledge surrounding a topic, speciﬁcally with reference to the diversity of
individuals and institutions that espouse these bodies of knowledge. This is a particularly
relevant ﬁnding in relation to the authority of the medical institution in constructing
knowledge on the topic of breastfeeding. As an authority on medical knowledge of infant
health, the ﬁnding suggests the necessity of medical programs that not onlypromote
these beneﬁts of breastfeeding, but also acknowledge and give credence to mothers’

concerns about the limitations of breastfeeding.

160

Managerial Recommendations

Following from this, managerial implications of the current investigation strongly
emphasize critical evaluation of personal and institutional assumptions about responsible
motherhood. Across two unique models of breastfeeding percentage, three beliefs about
breastfeeding were consistently signiﬁcant predictors of this characteristic. The ﬁrst,
belief in the beneﬁts of breastfeeding, was the strongest predictor of greater breastfeeding
percentage on both models. Beliefs associated with this factor related to beliefs in the
infant health beneﬁts of breastfeeding, greater bonding between the mother and the
infant, personal satisfaction with the experience of breastfeeding, and the convenience of
breastfeeding. As a strong predictor of greater breastfeeding percentage, fostering beliefs
in these potential beneﬁts of breastfeeding may have the potential to increase the
proportion of feedings that mothers select to breastfeed their infants rather than formula
feed their infants. However, previous empirical research on this belief has indicated that
typically this belief is strong even among mothers that decide not to initiate breastfeeding
(Brownell et al., 2002; Gabriel et al., 1986; Schmied & Lupton, 2001). Thus, it may be
advisable for healthcare initiatives to redirect a portion of their efforts in a direction with
some promise for fostering greater understanding between the messages of the healthcare
delivery system and the needs of mothers.

If the healthcare delivery system is to continue to promote the beneﬁts of
breastfeeding to its clients, it may be useful to note that, for the current investigation,
belief in the beneﬁts of breastfeeding was strongly correlated with the inﬂuence of family
and friends, the healthcare delivery system, and community sources. This result suggests

that, in addition to the healthcare delivery system as a staunch promoter of breastfeeding

161

and positive breastfeeding sentiment, two other resources exist for promoting positive
beliefs about breastfeeding. In particular, involvement of family and friends whose
opinions a re v alued b y t he m other dur ing p renatal a nd postpartum care may promote
positive breastfeeding sentiment among family, friends, and the mother. This tactic for
increasing breastfeeding initiation and duration has been recommended in numerous
studies, particularly with reference to the baby’s father (Humphreys et al., 1998; Isabella
& Isabella, 1994; Matich & Sims, 1992). However, in developing approaches to
involving these individuals, equal emphasis and consideration should be placed on the
recruiting and involving the range of individuals that are most signiﬁcant to the mother.
These individual may or may not include the baby’s father and, based on research that
suggests greater reliance on social networks of friends and extended family among
minority populations (Blum, 1999; Blum & Deussen, 1996; Dominguez & Watkins,
2003), efforts to involve family and friends should be sensitive to involving not only the
baby’s father, but also the (grand)mother, close friends, and extended family members.
Two other beliefs about breastfeeding were also consistently signiﬁcant predictors
of breastfeeding percentage in the opposite direction. Stronger agreement with two
beliefs, the belief that the breastfeeding mother is fully responsible for infant feeding and
the belief that breastfeeding is painful, predicted lower breastfeeding percentage. Items
on these beliefs related to mothers’ exclusive responsibility for breastfeeding, burdens of
breastfeeding on the mother, and mothers’ experiences of physical discomfort while
breastfeeding. Consistentthroughout all of these it ems was a focus on the needs and
desires of the mother, wherein the items suggested that mothers’ needs and desires were

in opposition to breastfeeding the infant. These results suggest a promising direction for

162

future healthcare initiatives aimed at promoting infant and maternal health, most notably
through the affirmation of mothers’ needs and wants in her infant feeding decisions.

In order to alleviate mothers’ negative sentiment toward breastfeeding in the
context of constraints on her own needs and desires, it may be advisable for healthcare
delivery p rofessionals to c oach ne w m others 0 n the beneﬁts of breastfeeding that are
exclusive to the mother. In particular, fostering greater belief in the beneﬁts of
breastfeeding might reference the beneﬁts derived from breastfeeding that are exclusive
to the mother. Two practical considerations, convenience and economic advantages, were
associated with a belief in the beneﬁts of breastfeeding. The economic advantages of
breastfeeding has relatively exclusive beneﬁts for the mother; however, acknowledging
thatthe W IC p rogram p rovides free formula to non-breastfeeding mothers, it may be
challenging to promote the ﬁnancial beneﬁts of breastfeeding to mothers that participate
in this program and who would not be required to pay for formula in any case.
Nonetheless, the strength of mothers’ beliefs in the beneﬁts of breastfeeding in predicting
infant feeding behavior suggests that it may be advisable to counsel mothers on the
practical beneﬁts of breastfeeding as exclusive to the mother. In the speciﬁc context of
WIC participants, it may be important to focus on the advantage of convenience over
economics.

Additional emphasis on the exclusive beneﬁts to the mother may focus on
emotional motivations, including personal satisfaction and closeness to the infant. Indeed,
an unscientiﬁc review of informational materials provided to mothers at both of the
participating WIC agencies indicated that this approach is emphasized. However, it

should be noted that these beneﬁts are closely bound to constructions of motherhood that

163

 

 

equate maternal satisfaction with attending to the needs of others, which, as noted, may
conﬂict with mothers’ concern for their own needs and desires. If the emotional beneﬁts
are promoted delicately, that is, with careﬁil sensitivity to dominant ideologies about the
care-taking roles of mothers, emotional beneﬁts derived exclusively by the mother may
be another approach to promoting the beneﬁts of breastfeeding.

Along these same lines, based on the strength of mothers’ concerns for
themselves in negatively predicting infant feeding behavior, it is important that mothers’
concerns for themselves are not discounted in counseling mothers in their infant feeding
behaviors. In particular, healthcare professionals should be careﬁJl not to displace
mothers’ concerns about their own needs and desires with automatic praise of the beneﬁts
of breastfeeding, regardless of whether the beneﬁts are exclusive to the mother, the
infant, or shared between both. Wall (2001) has noted the abundance of breastfeeding
promotion materials that address mothers’ concern for others, with relative ignorance of
mothers’ concern for themselves. Following from this research and results of the current
study, it is important that healthcare delivery professionals give credence to the very real,
and predictive, concerns of mothers regarding impacts of breastfeeding on themselves
and their lifestyle.

Research Implications

Contrary to the proposed models of infant feeding behavior for the current
investigation, discourses of maternal and sexual femininity and the authority of external
inﬂuences were not signiﬁcant predictors of mothers’ infant feeding behavior. However,
extracted factors of the Positive and Negative Breastfeeding Sentiment Scales appeared

to address themes related to maternal femininity and concerns about mothers’ autonomy.

164

Based on this result, measurement of the constructs warrants ﬁirther assessment in future
studies. For the current investigation, instrumentation developed by Janke (1994) was
used to measure external inﬂuence on mothers’ feeding decisions, while instrumentation
developed by Kumar, Robson and Smith (1984) was used to measure body image and
attitudes toward motherhood and the baby. Factor analysis was performed on all of these
instruments to extract factors of each of these constructs which, when entered into
predictive models, were not signiﬁcant predictors of feeding behavior. The statistical
insigniﬁcance of these measures in predicting feeding behavior contradicts previous
research suggesting the inﬂuence of external actors (Humphreys et al., 1998; Isabella &
Isabella, 1994; Matich & Sims, 1992), sexuality (Jones, 1987; Maehr et al., 1993; Shaw,
2003; Stearns, 1999; Young, 1990), and maternal femininity (Arafat et al., 1981; Murphy,
2000) on mothers’ feeding decisions.

Based on these results, it is recommended that future research qualitatively
examine this instrumentation to enhance reliability of the measures and validity of the
constructs. In particular, qualitative review of items on the existing instruments is
recommended to determine the r eliability o f e xisting it ems in relation to more recent
ﬁndings on infant feeding behavior. Relying on qualitative assessment of infant feeding
issues relevant to populations of low-income women, future research may attempt to
modify items on these existing scales to contain items and issues supported in qualitative
literature.

Quantitative exploration of the existing scales may also be advisable using
experimental design, wherein inﬂuence, body image, and attitudes toward motherhood

and the baby are explored during pregnancy (potentially at several intervals) and after

165

 

 

birth to temporally track changes in expectant mothers’ transition into motherhood.
Existing subscales of the Maternal Attitudes and Maternal Adjustment Questionnaire
(Kumar et al., 1984) already contain these two item formats that facilitate asking
questions during pregnancy and after birth. Formats of the external inﬂuence items would
need to be revised to facilitate this exploration.

Also in relation to the wording of questionnaire items, it is recommended that
future research examine the extent to which instrumentation that combines positively-
worded and negatively-worded items affects the reliability of the instrument. For the
current investigation, factor analysis of the existing instrumentation on body image and
attitudes toward motherhood and the baby almost exclusively extracted factors that
contained items with either positively-worded items or negatively-worded items. As
indicated, the intention of factor analysis 0 n these it ems w as to e xtract s ubfactors o f
meanings—not grammatical structure—related to these constructs. Revision of these
existing instruments might therefore incorporate an alignment of all questionnaire items
so that they are either positively-worded or negatively-worded to explore whether factor
analysis extracts similar categories of items.

This recommendation also extends to the items on beliefs about breastfeeding,
wherein items almost exclusively grouped on the Positive Breastfeeding Sentiment Scale
by whether they were oriented toward breastfeeding or away from formula feeding. The
recommendation to further assess instrumentation using qualitative techniques also
applies to this scale, as existing items may be worded to measure something altogether
different from the intention of the item or whole categories of beliefs about breastfeeding

supported in the literature may not yet exist on the current instrument.

166

 

It should also be noted that beliefs about breastfeeding were included in the study
to control for mothers’ personal beliefs about the advantages and disadvantages of
breastfeeding. However, based on the results of data analysis, beliefs about breastfeeding
were the strongest predictors of infant feeding behavior. Based on these results, it is
advised that future research conduct further exploration of the antecedents of beliefs
about breastfeeding, including how and when beliefs about breastfeeding are shaped, the
individuals that have the potential to inﬂuence beliefs, experiences and observations that
affect beliefs, and other factors that are associated with the development and change in
beliefs about breastfeeding.

Study Limitations

Stemming from the above issues related to measurement, it must be noted that
individual measures and complete scales used for the current study were limited in their
ability to capture the immense complexity of concepts such as sexual femininity,
maternal femininity, social inﬂuence, and beliefs about breastfeeding. Admittedly, issues
surrounding body image provide only one slice of information about women’s feelings of
sexual femininity, just as maternal adjustment can only illuminate a small portion of
mothers’ interpretations of their own maternal femininity. Acknowledging this limitation
in the development of the study instrument, it was nonetheless decided that the
information that could be provided from the items and scales included on the instrument
were indicators of these concepts. Suggestions for use of these scales in future studies has
been discussed among the research implications of the current investigation.

Additionally, s everal factors associated with the process and outcomes of data

collection were also limiting characteristics in this study. The ﬁrst of such characteristics

167

 

was sample size, which was considerably smaller than proposed in the original study
proposal. Following discussions in April/May 2003 with WIC administrators and staff at
the state and local level, it was anticipated that a sample of 100-200 study participants
could be recruited from each of the participating urban locations in a span of three weeks
of questionnaire administration in J une/July 2003. During this three week data collection
period, discussions were conducted approximately once every week with the WIC
coordinator at each urban location to determine progress of data collection. In these
discussions, it became apparent that three weeks of data collection was insufﬁcient to
recruit the desired number of study participants from each urban location. Accordingly,
the study timeline was extended by a period of four weeks to obtain larger study samples
from each site.

Throughout the data collection timeline, conversations were held with the WIC
coordinators to explore reasons for low participation. There were several reasons cited by
coordinators for the apparent discrepancy between anticipated study participation and
actual study participation, which ultimately led to an extended timeline for survey
administration. First, the coordinators expressed concern with the amount of internal
WIC paperwork required from WIC participants that were recently postpartum and that
were also being recruited for the study. As expressed by the coordinators, in order to
obtain WIC certiﬁcation for herself and her newborn infant, new mothers were obligated
to complete a Postpartum Health History for herself and an Infant Health History for her
newborn infant. Each of these documents consisted of approximately six pages of
questions about mothers’ and infants’ demographics and household characteristics,

general health, nutrition and environmental characteristics, and infant feeding behaviors.

168

 

Based on this amount of intensive information required from new mothers for internal
WIC records, WIC coordinators were uncertain about mothers’ willingness to complete
an additional six pages of questions for the study. Regardless of this internal WIC
paperwork, mothers may have been intimidated by sheer length of the instrument, which
is contained in Appendix B.

Second, the study made use of self-administered questionnaires to collect primary
data on study participants. Questionnaires were only made available in English and, in
most cases, WIC clinics did not have staff resources available to read the questionnaire
orally to study participants. This required that potential study participants were literate in
English and that they were able to independently read and understand items on the
questionnaire. Acknowledging these factors, attempts were made to adjust the language
of the instrument to an appropriate reading level for WIC populations that were literate in
English and to minimize the effect of education on ability to participate in the study. The
instrument w as reviewed by WIC staff and administrators prior to administration and
approved for WIC participants. Nonetheless, the nature of the methodology may have
limited participation by some potential study participants.

Third, WIC coordinators’ observations of new mothers in the waiting area
included the reﬂection that many new mothers attended their WIC appointment with their
newborn infant and potentially other children. Although some mothers attended their
appointment with another adult, WIC coordinators noted that the majority of women
were caring for their newborn infant, and potentially other children, independently in the
clinic waiting area. Mothers’ priority to attend to her children during the waiting period,

which could potentially last for an hour or longer, may have been a factor in deterring

169

some mothers from participating in the study. These observations were conﬁrmed by the
researcher during a visit to one of the W1C clinics managed by the Detroit City Health
Department.

An additional issue related to the small obtained sample size was related to
sensitive management of staff and management at the participating WIC clinics by the
researcher. Throughout the data collection period, minimal direct contact was maintained
between the researcher and WIC clinic staff. Speciﬁcally, most study-related information
was mediated through a third party, namely the WIC coordinator for each urban location.
Considering the large amount of responsibility placed on WIC clinic staff for data
collection, neglect of this speciﬁc relationship may have induced negative sentiment
toward the study or limited widespread knowledge of the study among staff employed
throughout the sixteen participating WIC clinics. In retrospect, greater attention would
have been directed to WIC staff and this relationship would have been treated much more
sensitively. The researcher’s experience administering questionnaires at the WIC clinic in
Detroit conﬁrmed the need for additional resources to assist staff in administering
questionnaires and to ease the burden on the WIC clientele. However, owing to the
extensive costs of placing research personnel at all sixteen clinics, this was outside the
scope of the current investigation.

Based on the extension of the data collection timeline, the date of secondary data
extraction was another limiting characteristic of the current investigation. Speciﬁcally,
this limited the ability for the current investigation to draw conclusions about the two
urban WIC populations based on results of the study. Relying on the original study

timeline of three weeks, secondary data extraction was conducted at the end of June 2003

170

to capture WIC clients that were active WIC participants at the same time of data
collection. However, as indicated, this timeline was extended by four weeks owing to the
small number of completed questionnaires collected after the initial three weeks of data
collection. As a result, approximately four weeks of new enrollees were not captured in
the secondary data set and secondary data on several study participants was unmatched.

Further, despite the ability to compare characteristics of the study sample with
other WIC clients at the same location, sampling procedures for the current study were
not random. Survey administration commenced in mid-June and concluded at the end of
July 2003, accounting for only a portion of the yearly WIC client base. All WIC clients
that met the eligibility requirement of zero to three months postpartum and that were
willing to participate in the study were given the opportunity to complete a questionnaire.
Absence of random sampling procedures limited the ability to generalize results to the
larger WIC client groups at each of these locations.

However, one particular strength of the current investigation was the availability
of secondary data on the study sample and the larger group of active WIC clients enrolled
at each WIC clinic, which enhanced the researcher’s ability to ascertain the degree to
which the study sample was representative of other WIC clients at the same location.
Three groups were compared to investigate the degree to which the study sample was
representative of other WIC clients that were eligible to p articipate int he s tudy ( 0-3
months postpartum) and other WIC clients that were ineligible to participate in the study
(pregnant or greater than 3 months postpartum). Overall, comparisons between the three
groups indicated that the study sample from each urban location was moderately

representative of other WIC clients that would have been eligible to participate in the

171

study and slightly less representative of ineligible WIC clients ﬁom each urban location.
More speciﬁcally, in the context of socio-demographic characteristics, the study sample
recruited for the current investigation was, for the most part, representative of other WIC
clients in the same urban location.

Most notably, differences between the study sample and the eligible and ineligible
WIC client groups from each location were statistically signiﬁcant on education level
obtained, w hereby t he s tudy s ample at e ach location possessed a larger proportion of
women that had obtained at least some college-level education. As the literature suggests
a positive relationship between breastfeeding initiation and duration and level of
education (Grossman et al., 1989; McLeod et al., 2002; Quarles et al., 1994), this
difference in education level obtained between t he s tudy s ample a nd the eligible a nd
ineligible WIC client groups at each location was one limitation of the current study. In
particular, based on the signiﬁcant role of education suggested by these previous studies,
the ability for the current investigation to make predictions about other WIC clients at
each urban location may have been limited by this result.

Other signiﬁcant socio-demographic predictors of infant feeding behavior cited in
the literature (Kurinij et al., 1988; Maher, 1992; Oyeku, 2003; Quarles etal., 1994),
including race/ethnicity and mother’s age, were similar between each location’s study
sample and its corresponding eligible and ineligible WIC client groups. In the context of
other characteristics of the study sample and the eligible and ineligible WIC client groups
at each urban location, the study sample was representative of other WIC clients to a fair
extent. Although there were relatively few signiﬁcant differences between the study

sample and the WIC client groups on participation in other maternal-child support

172

programs or characteristics associated with prenatal care, there were signiﬁcant
differences on infant feeding behavior between these three groups. However, upon
further analysis, these differences were more likely attributable to differences in infant
age across the three groups. Based on these results, the study sample from each urban
location was similar to the larger eligible WIC client group on more characteristics than
they were different.

Conclusion

This study commenced with the purpose of exploring breastfeeding in its social
context, both with reference to mothers’ systems of social inﬂuence and mothers’ beliefs
about their maternal and sexual femininity. One of the most prominent results in this
exploration was the signiﬁcance of mothers’ concerns about themselves in p redicting
infant feeding behaviors. From a Foucaultian perspective, this result suggests the
potential for decision-making conﬂict in the presence of maternal discourses on infant
feeding. Other discourses of femininity and resulting bodies of knowledge on infant
feeding, although supported in the literature, were not supported in the current study.

This apparent incongruence with previous literature underscores the need for
further research on discourses of femininity, sources of external inﬂuence that espouse
diverse discourses, and resulting bodies of knowledge surrounding infant feeding. One of
the most fundamental directions in which future research may move is toward enhanced
measurement of these constructs using both qualitative and quantitative methods.
Existing quantitative instrumentation on these constructs, some of which was used in the

current investigation, remains in early stages of testing and reﬁnement. The results of this

173

 

 

study stress the importance of even further theme development and testing to attempt to
enhance measurement of these complex constructs.

Based on the results obtained in the current investigation, this research also has
multiple implications for practice, particularly with reference to maternal discourses on
femininity and mothers’ beliefs about breastfeeding. In practice, mothers may outwardly
express concerns about their personal lifestyle or, on the basis of prevailing norms
regarding motherhood, may inwardly struggle with standards of responsible motherhood
that necessitate breastfeeding. This may signal considerable conﬂict for some mothers in
assessing their infant feeding options. For healthcare professionals, these concerns may
require particular sensitivity to mothers’ beliefs about the constraining effects of
breastfeeding on their lifestyles and personal autonomy.

Indeed, the results of this study underscore the importance of greater sensitivity
toward mothers from multiple members of their social networks. It is apparent that
mothers’ infant feeding choices are not made in the isolated context of the mother-infant
dyad. Rather, mothers’ understandings of responsible motherhood suggest that their
behaviors and beliefs are shaped in a larger social context. Consistent with the
Foucaultian perspective taken in this study, it is paramount that mothers’ infant feeding

choices are treated as complex decisions that reference these multiple social contexts.

174

Appendix A

MEASUREMENT OF FACTORS RELATED TO
MOTHERS’ INFANT FEEDING BEHAVIOR

Several scales have been developed that measure factors related to motherhood
and infant feeding choices, outlined in Table A1.

Table A1: Measurement of F actors Related to Motherhood and Infant Feeding

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

# of Items Reliability l'f
Maternal Adjustment and Maternal Attitudes 60 ,
Kumar, Robson and Smith (1984)
1) Body image 12 0.72
2) Somatic symptoms 12 0.58 " ".
3) Marital relationship 12 0.74
4) Attitudes to sex 12 0.82
5) Attitudes to pregnancy/motherhood and the baby 12 0.73
Infant Feeding Questionnaire 37
Dusdieker, Booth, Seals and Ekwo (I 985)
l) Worries about own health and family’s health 4 0.92
2) Maternal beliefs in preventive health behavior 4 0.71
3) Maternal breastfeeding beliefs 6 0.87
4) Maternal beliefs in personal satisfaction from breastfeeding 3 0.70
5) Maternal breastfeeding worries 6 0.75
6) Maternal anxiety before breastfeeding began 2 0.78
7) Maternal worries about lack of psychosocial support 2 0.91
8) Inﬂuence of resource materials 2 0.74
9) Inﬂuence of breastfeeding women 2 0.65
10) General support for breastfeeding 2 0.84
11) Public inﬂuences on decision to breastfeed 4 0.7 5
Social Support for Breastfeeding Scale 17
Matich and Sims (1 992)
1) Tangible support 7 0.88
2) Emotional support 5 0.94
3) Informational support 5 0.93
Breastfeeding Attrition Prediction Tool 51
Janke ([994)
1) Negative breastfeeding sentiment 15 0.83
2L Positive breastfeeding sentiment 14 0.79
3) Social and professional support 12 0.85
4) Breastfeeding control 10 0.81
Breastfeeding Self-Efficacy Scale 39
Dennis and F our (1999)
1) Technique 2 l 0.94
ZLIntrapersonal thoughts 13 0.93
3) Support 5 0.75

 

 

 

 

 

175

 

 

AQum
Snﬁﬂr
inﬁun
nﬂe r
iunns
educz

mate:

item'
(0 p]
Past
fain
Dart
(an
(an
ﬂlr

ne‘

Sr

in

(I

Mothers’ ability to adjust to maternal roles was the topic of the Maternal
Adjustment and Maternal Attitudes Questionnaire, developed by Kumar, Robson and
Smith (1984). Although the scale was not developed with speciﬁc emphasis on mothers’
infant feeding choices, speciﬁc subscales may contribute to understanding the maternal
role negotiation processes that inﬂuence mothers’ decision-making processes. Survey
items were developed from unstructured interviews with mothers involved in prenatal
education and childbirth classes. The instrument, containing 91 items related to the
maternal role, was tested among two samples of clients at an antenatal clinic in London.

Following item analysis, ﬁve subscales were extracted that each contained twelve
items: body image, somatic symptoms, marital relationship, attitudes to sex, and attitudes
to pregnancy/motherhood and the baby. Sample items ask about mothers’ feelings in the
past four weeks and included: “Have you felt attractive?” (body image), “Have you felt
faint or dizzy?” (somatic symptoms), “Have you found it easy to show affection to your
partner?” (marital relationship), “Have you found your partner sexually desirable?”
(attitudes to sex), and “Have you Have you enjoyed caring for your baby’s needs?”
(attitudes to pregnancy/motherhood and the baby). Items were randomly scattered
throughout the 60-item instrument and were measured on two four-point Likert scales:
never, rarely, often, very often and not at all, a little, a lot, and very much.

Although several items on the questionnaire developed by Kumar, Robson and
Smith (1984) related to infant feeding, the questionnaire was not developed as an
instrument speciﬁc to this discursive ﬁeld. As one of the ﬁrst attempts to standardize
measures related exclusively to infant feeding choices, the Infant Feeding Questionnaire

(Dusdieker et al., 1985, p. 695) was a scale that added complexity to “oversimpliﬁed”

176

 

 

models of the speciﬁc discursive ﬁeld of infant feeding. Although previous studies had
attempted to quantify inﬂuences on infant feeding choices, the researchers critiqued these
studies for exploring relationships of distinct constructs with infant feeding choices. As a
result, the Infant Feeding Questionnaire was developed for two purposes: (1) to quantify
and standardize concepts involved in women’s initial infant feeding decision and (2) to
use multivariate analysis to examine primary and secondary inﬂuences on infant feeding
choices.

Survey items were developed by the researchers following semi-structured
interviews with women that had chosen to breastfeed their infants. The ﬁnal instrument
contained 49 items that addressed personal and familial health concerns, maternal beliefs
in preventive health behaviors, positive attitudes toward breastfeeding, worries about
breastfeeding, social inﬂuences on the initial infant feeding decision, and availability of
sources of social support. The researchers surveyed 157 ﬁrst-time mothers that had
chosen to exclusively breastfeed (n = 100) or exclusively formula feed (n = 57). The
sample was recruited from four pediatric practices in two Midwest towns.

Factor analysis extracted ll subscales involved in the initial infant feeding
decision: worries about their own and family’s health (4 items, or = 0.92), maternal
beliefs in preventive health behavior (4 items, or = 0.71), maternal breastfeeding beliefs
(6 items, a = 0.87), maternal beliefs in personal satisfaction from breastfeeding (3 items,
a = 0.70), maternal breastfeeding worries (6 items, a = 0.75), maternal anxiety before
breastfeedingbegan(2 items, CL = 0.78), maternal worries about lack of psychosocial

support (2 items, or = 0.91), inﬂuence of resource materials (2 items, or = 0.74), inﬂuence

of breastfeeding women (2 items, 0. = 0.65), general support for breastfeeding (2 items, at

177

= 0.84), and public inﬂuences on the decision to breastfeed (4 items, or = 0.75). Among
the twelve items removed following factor analysis, two items were removed from the
instrument that would appear particularly inﬂuential in mothers’ infant feeding choices.
“The inﬂuence of the baby’s father” and “advice from health care personnel” clustered
together and both were removed because they appeared to measure different concepts.
Nonetheless, these items were included separately as independent variables in regression
analysis and causal diagram due to their face validity.

The instrument’s multidimensional treatment of mothers’ infant feeding choices
allowed researchers to move beyond bivariate analyses of factors related to infant feeding
decisions. However, the sampling method for the study may have limited the content
validity of the instrument. The researchers recommended further testing of the instrument
with ethnically and racially diverse populations, as participants for the study were
recruited from towns with predominantly white populations. Even so, the Infant Feeding
Questionnaire (Dusdieker et al., 1985) was a foregoing attempt at synthesizing multiple
constructs into a model of mothers’ infant feeding choices.

Subsequent scales added complexity to constructs measured in the Infant Feeding
Questionnaire (Dusdieker et al., 1985). Responding to the alleged absence of a model
addressing the complexity of the social support construct alone, Matich and Sims (1992)
developed a social support instrument to measure the sources, types and amounts of
mothers’ perceived social support. The researchers argued that the social support
construct in particular “is complex and appears multidimensional in nature” (Matich &
Sims, 1992, p. 920). Indeed, measures of social support in previous research focused on

one aspect of social support with little attention to multiple aspects of the construct.

178

 

 

Accordingly, the model developed by Matich and Sims (1992) attempted to identify
sources of social support with particular focus on aspects and amounts of perceived social
support from each source. Questionnaire items were developed to reﬂect three aspects of
support — tangible, emotional, and informational — which the researchers suggested may
have independent effects on physical and mental health outcomes (Matich & Sims, 1992).

Respondents were asked to consider 23 items for each of nine sources of social
support. Sample items included: “Takes me to the doctor’s or other places I need to go”
(tangible), “Cares about how I’m feeling and how the pregnancy is going” (emotional),
and “Teaches me how to do some things like getting prepared for breastfeeding”
(informational) (Matich & Sims, 1992). For each item, respondents were presented with
ﬁve Likert-type responses: no help at all, helps a little, fairly helpful, very helpful, and
extremely helpful.

Questionnaires were administered to 159 women in their third trimester enrolled
in prenatal classes and clinics, including a WIC clinic, in a rural area of Pennsylvania. A
greater number of respondents indicated an intention to breastfeed their baby (breastfeed,
n = 85; formula feed, n = 74), and only women that indicated an intention to breastfeed
were asked to complete a second questionnaire at four weeks postpartum. Following
factor analysis, three subscales composed of seventeen total items were included in the
revised social support scale (tangible: 7 items; emotional: 5 items; informational: 5
items).

Just as the Social Support for Breastfeeding Scale (Matich & Sims, 1992)
enhanced the complexity of the social support construct in predicting women’s infant

feeding choices, the scale developed by Dennis and Faux (1999) added complexity to

179

 

maternal conﬁdence as a predictor of these choices. The Breastfeeding Self-Efﬁcacy
Scale (Dennis & F aux, 1999) was developed to explore the relationship between maternal
conﬁdence in ability to breastfeed and breastfeeding duration. The instrument was based
on self-efficacy theory, which locates individuals’ decisions and endurance in the context
of individuals’ evaluation of their own ability to perform those behaviors. As a result, the
instrument was developed as a predictor of premature weaning among breastfeeding
mothers.

The researchers identiﬁed and developed measures for three dimensions of self-
efﬁcacy: technique, intrapersonal thoughts, and support. Following a review of literature,
expert review and an initial pilot-test, the ﬁnal instrument contained 43 items. Items
began with the phrase, “I can always...” and listed speciﬁc experiences or feelings
related to breastfeeding. Sample items included: “Manage the breastfeeding situation to
my satisfaction” (technique), “Keep feeling that I really want to breastfeed my baby for at
least six weeks” (intrapersonal thoughts), and “Express my breastmilk if I need to”
(support). Respondents were asked to reply using a ﬁve-point Likert scale that ranged
from “not at all conﬁdent” to “always conﬁdent.”

The Breastfeeding Attrition Prediction Tool (Janke, 1994) was developed to
examine predictors of premature weaning among breastfeeding mothers. The instrument
relied on the assumptions of the Theory of Reasoned Behavior, which situates
individuals’ decisions and behaviors within the context of three constructs: desired
outcomes (attitude), social reaction (subjective norm), and perceived control (control).

Focusing on measures that were largely social, questionnaire items were developed from

180

 

 

a review of literature, clinical experience, and personal interviews with postpartum
women.

Following an initial pilot of the instrument to reﬁne items, a sample of 201
breastfeeding mothers was recruited from postpartum units of two urban hospitals to test
scale reliability (Janke, 1994). Study participants were eligible if they had indicated an
intention to breastfeed for at least eight weeks. The instrument contained 51 unique items
measuring mothers’ attitudes toward breastfeeding (attitude), inﬂuence of others
(subjective norm), and perceived ability to overcome constraints to breastfeeding
(control). The attitude and subjective norm constructs were each measured in two sets of
questions to explore perception and importance of each item.

Factor analysis extracted four subscales: negative breastfeeding sentiment (15
items, or = 0.83), positive breastfeeding sentiment (14 items, or = 0.79), social and
professional support (12 items, a = 0.85), and breastfeeding control (10 items, or = 0.81).
These subscales drew on the strengths of the Infant Feeding Questionnaire (Dusdieker et
al., 1985) by treating mothers’ infant feeding choices as a multidimensional construct.
Further, the instrument drew on the strengths of the Social Support for Breastfeeding
Scale (Matich & Sims, 1992) in that it operationalized the social support variable as a
multidimensional variable composed of perception of support and motivation to comply.

Nonetheless, as an evolving instrument, Janke (1994) acknowledged that study
limitations warrant further testing of the instrument. All items remained in the instrument
following factor analysis due to a low 0 riterion for removal ( factor loading less than
0.30). Only one item loaded below this level (social and professional support: “the baby’s

father”), but remained in the instrument due to the item’s face validity. Further, the

181

 

 

sample for this instrument was adult women that were primarily white (88%), married
(80%), educated beyond high school (66%), and middle- to upper-income (69%) (Janke,
1994, p. 101). As a result of these limitations, the researcher advised testing the
instrument among larger samples and diverse populations to further reﬁne and shorten
subscales, emphasizing the need to maintain subscales’ reliability and validity (Janke,
1994, p. 104).

Based on a review of these studies that used these scales, a decision was made
based on three criteria to include two subscales of the Maternal Adjustment and Maternal
Attitudes Questionnaire (Kumar et a 1., 1984) a nd three 5 cales from the B reastfeeding
Attrition Prediction Tool (Janke, 1994) in this study. The decision criteria in order of
priority was: (1) the applicability of the scale to the objective of this study; (2)
description and discussion of the reliability of the measure; (3) applicability to the sample
in this proposed study.

From the Maternal Adjustment and Maternal Attitudes Questionnaire (Kumar et
al., 1984), two subscales of Body Image and Attitudes toward Motherhood and the Baby
were used to measure respondents’ feelings of sexual attractiveness and maternal
femininity. Reliability for the original subscales was established through test-retest and
split-half methods; reliability reported in Table A1 is based on the split-half test of
reliability. However, the sample drawn to test reliability of the original instrument was
predominantly married or stably cohabiting with the baby’s father (Kumar et al., 1984).
For the current investigation, marital status among the sample was considerably more
diverse and, based on extracted factors of these two subscales, added to the reliability of

the subscales among a more diverse sample on the basis of the marital status

182

 

 

characteristic. Consistent with the objectives of this study, the decision was made to test
the reliability of these two subscales among this sample and to evaluate the predictability
of mothers’ feeding behavior on these two measures.

From the Breastfeeding Attrition Prediction Tool (Janke, 1994), three scales were
selected to measure beliefs about breastfeeding (Negative Breastfeeding Sentiment Scale
and Positive Breastfeeding Sentiment Scale) and sources of external inﬂuence (Social
and Professional Support Scale). Scales were developed separately and tested using
exploratory factor analysis. However, as with the scale above, the sample drawn to test
this instrument was predominantly married, and the majority of the sample was white,
educated beyond high school, and middle- to upper-income. As a result, the decision was
made to test the reliability of instrument scales with a more diverse sample, particularly
with reference to race, education and marital status. The scales were used to accomplish
Study Objective 2, which aimed to test instrument reliability among the current sample,
and Study Objective 3, which aimed to predict mothers’ feeding behavior on sources of

external inﬂuence.

183

 

 

Appendix B

SURVEY INSTRUMENT

184

 

 

 

Infant Feeding Decision Study

Dear WIC Client,

Our WIC Clinic is working with a student at Michigan State
University. We want to learn how women make choices about
infant feeding. You can help us by ﬁlling out a survey while
you wait for your appointment. The answers you provide can
help WIC staff better serve WIC clients.

Women who take part in this study will be entered into a prize
drawing. The prize is one of eight $25 gift certiﬁcates to a
local grocery store. If you Wish to be included in this drawing,
please enter your name and contact numbers on the card
attached to this survey.

Your opinions are very important to us! Thank you for
participating in this study!

Sincerely,
Nancy Erickson Allison Churilla
WIC Coordinator Graduate Student

Detroit City Health Department Michigan State University

Please write your WIC identiﬁcation number:

 

185

 

 

Please answer the following questions about your baby and how you feed your baby. Your
name will not appear in any reports and your answers will o_nly be reported by group.

 

 

1) When was your baby born? / /
Month Day Year

2) Is your baby? (Please [21 only one.)

 

 

 

 

 

 

 

 

 

 

 

Cl Male CI Female
3) Have you ever—at any time—fed your baby breastmilk? (Please [21 only one.)
D Yes D No
0 0
If 19.5: If M:

a) When did you first decide that you a) When did you ﬁrst decide that you
would feed your baby breastmilk? would not feed your baby breastmilk?
(Please 121 only one.) (Please [21 only one.)

D Before I became pregnant El Before I became pregnant

[:1 In the ﬁrst 3 months of my El In the ﬁrst 3 months of my
pregnancy pregnancy

El 1n the second 3 months of my [:1 In the second 3 months of my
pregnancy pregnancy

I] In the third 3 months of my D In the third 3 months of my
pregnancy pregnancy

13 After my baby was born Cl After my baby was born

b) Have you ever used a breast pump to
pump breastmilk for your baby?
(Please [21 only one.)

D Yes Cl No

c) Has your baby been fed breastmilk in
the past week? (Please [21 only one.)

D Yes Cl No
0
lfgg, how old was your baby when
you stopped feeding breastmilk?
weeks old
4) Currently, how many times is your baby usually fed...
a) Formula? __ times each day
( number)
b) Pumped breastmilk? times each day
( number)
c) Breastmilk from the breast? times each day
( number)
5) Other than breastmilk or formula, do you feed your baby anything else? (Please [21 only one.)
D Yes [:1 No

0

 

If _Y_e_s, please describe any other things you feed your baby:

 

 

 

 

 

186

 

6) Please respond to these items about whether certain individuals thought you should deﬁnitely
breastfeed. (Please circle only one.)

 

 

 

When I first decided how to feed my baby, E; g g § u E“, w Dong)?“
these people thought that I should definitely g g ,§ § g g E Does Not
breastfeed my baby... "’ Q Q 2 ‘5 '4 '5 Apply

a) My baby’s father 1 2 3 4 5 E]

b) My mother 1 2 3 4 5 Cl

C) My mother-in—law l 2 3 4 5 D

(1) My family 1 2 3 4 5 El

e) My best friend 1 2 3 4 5 [:1

f) My other friends 1 2 3 4 5 [:1

g) My doctor(s) l 2 3 4 5 D

h) My midwife l 2 3 4 5 D

i) WIC staff 1 2 3 4 5 D

j) Breastfeeding Peer Counselors 1 2 3 4 5 [:1

k) Other people at my school 1 2 3 4 5 1:1

1) Other people at my work 1 2 3 4 5 D
m) My community 1 2 3 4 5 C]

n) People in public places 1 2 3 4 5 El

 

7) How much do you value others’ opinions about how you should feed your baby?
(Please circle only one.)

 

 

 

b 3 3 ~ 3‘ Don ’t Know
:0 to to 3 \o g“ \a OR
I value these people’s opinions on how I g g ,§ § E E '3, Does Not
should feed my baby... '° Q Q 2 V ‘5 ‘5 Apply
a) My baby’s father 1 2 3 4 5 Cl
b) My mother 1 2 3 4 5 [:1
c) My mother-in-law l 2 3 4 5 D
(1) My family 1 2 3 4 5 C]
e) My best friend 1 2 3 4 5 El
f) My other friends 1 2 3 4 5 Cl
g) My doctor(s) l 2 3 4 5 [:1
h) My midwife l 2 3 4 5 El
i) wrc staff 1 2 3 4 5 D
j) Breastfeeding Peer Counselors 1 2 3 4 5 El
k) Other people at my school 1 2 3 4 5 Cl
1) Other people at my work 1 2 3 4 5 [I
m) My community 1 2 3 4 5 [:1
n) People in public places 1 2 3 4 5 El

187

 

 

a)
b)
C)
d)
e)

h)
i)
J)

k)
l)

m)
n)
0)

S)
0

U)
v)

x)
y)
2)
bb)

cc)
dd)

8) Please respond to these items about infant feeding. For each item below, please circle the
number that shows your opinion. (Please circle only one for each item.)

 

Breastfeeding is more convenient than formula feeding.
Breastfeeding is painful.

F orrnula feeding gives you more ﬁeedom.

Infant formula can cause allergies.

Breastmilk is healthy for the baby.

No one else can help feed the baby when you breastfeed.
It is embarrassing to breastfeed in public.

Formula fed babies tend to get sick.

Breastmilk is more nutritious than infant formula.
Breastfeeding makes your breasts sag.

Formula feeding is easier than breastfeeding.

Formula fed babies are more fussy than breastfed babies.
Breastfeeding makes you closer to your baby.
Breastfeeding makes returning to work difficult.

Formula fed babies are easier to satisfy than breastfed babies.

Formula fed babies tend to be overweight.

Breastfeeding is more economical than formula feeding.
When you breastfeed you never know if the baby is getting
enough milk.

Mothers who formula feed get more rest than breastfeeding
mothers.

Breastfeeding is natural.

Breastfeeding is more time consuming than formula feeding.
Formula feeding lets the father become close to the baby.
Infant formula can cause constipation.

Breastfeeding is best for the baby.

Breastfeeding is personally satisfying.

Breastfeeding is messy.

Breastfeeding ties you down.

Breastfeeding helps you bond with your baby.

Mothers who formula feed get back into shape sooner.

It is difficult to ﬁnd places to breastfeed outside your home.

188

Strongly
Disagree

pip—I-‘p—Iy—d nil—ﬂ—ﬂl

--_

ﬂH-~

N N N N N I Disagree

NNNNN

N MN NNNNN

NNNNN

NNIQNN

w w w w W I Neutral

wwwww

U) DJ wuwww

MWWWM

wwwww

Abbcb-hlAgI-ee

AJX-IB-D-A

J} Jib AAA-Ab

hhtﬁ-lk-b

AAA-Ssh

Strongly
Agree

MMMMUI UIMMMMI

Lit U’t MMU'MM

MMLAMUI

MMMMM

 

 

 

 
   

23

v; 9) These questions ask about how you felt during your last 3 months of pregnancy (your

third trimester). For each item below, please circle the number that shows your opinion.

(Please circle only one for each item.)

 

 

 

3.5 :3 8 - i?
i“ '22 ‘2? § 8 e a:
e -= a u '51: .3. a
Durigmy last 3 months of pregnancy... :3 i i i Z:
a) I felt proud of my appearance. 1 2 3 4 5
b) I liked the shape of my body. 1 2 3 4 f
c) I felt that my face was attractive. l 2 3 4 5
d) I felt that my breasts were too big. 1 2 3 4 5
e) I felt that my breasts were attractive. l 2 3 4 5
f) I felt that my breasts were too small. I 2 3 4 f
g) I felt that my body was soﬁ and cuddly. l 2 3 4 f
h) My body felt awkward and clumsy. l 2 3 4 5
i) I felt that my complexion was poor. 1 2 3 4 f
j) I felt that I had gained too much weight. 1 2 3 4 5
k) I felt attractive. 1 2 3 4 5
I) I felt that my body smelled bad. 1 2 3 4 5
z} 10) These questions ask about how you have been feeling since your baby was born. For
J 3 each item below, please circle the number that shows your opinion. (Please circle only

one for each item.)

 

Since my baby was born...

a)
b)
C)
d)
e)
t)

g)
h)
i)
J)
k)
l)

I have enjoyed caring for my baby’s needs.
I have felt that life has been more difﬁcult since the baby was born.
I have enjoyed feeding my baby.

I have been wondering whether my baby will be healthy and normal.

I have felt disappointed with motherhood.
I have been wonying that I might not be a good mother.

I have regretted having the baby.

I have felt proud of being a mother.

I have worried about accidentally hurting my baby.
I have had enough time for myself.

I have been feeling happy that I have a baby.

The thought of having more children appeals to me.

189

Strongly
Disagree

H ﬂ H — — _ l

_~—~_—

M N M N M N I Disagree

NKINK)NK)

(A) w r.» w to w I Neutral

1,.) r.) murmur

AAAAAAIAgree

Angngh

Strongly
Agree

'hUi'i-Utthutl

MIDMIAMIA

 

 

 

 

‘ Please answer the following questions about you and the people that you live with. Your name
will n_ot appear in any reports and your answers will only be reported by group.

 

 

 

V

‘1

What year were you bom?

 

year

Are you? (Please El only one.)
1:]

White CI Black or African-American
I] American Indian and Alaska Native 1:] Asian
1:] Native Hawaiian and Other Paciﬁc Islander [:1 Hispanic or Latino(a)
Cl Bi-Racial or Multi-Racial D Other (Please specify )

What is your highest level of education? (Please IE! only one.)
Less than 9"1 Grade
[:1 9‘h to 12Xh Grade (no diploma)
1] High School Degree or GED
El Some College (no degree)
Cl College Degree or higher

Are you? (Please [21 only one.)

 

 

 

 

U Married 1:1 Divorced or Separated

13 Single and Never Married El Widowed

D Other (Please describe: )
How many adults including yourself ( 18 years and older) live in your household?

number

Who are the other adults that live in your household? (Please [21 all that apply.)

Cl My mother Cl My father

1:] My mother-in-law Cl My father-in-law

U My best friend D Other friend(s)

Cl My baby’s father 1:] Other family member(s) (e.g., sibling, etc.)

1:] Other (Please describe: )

[:1 There are no other adults that live in my household.
Other than your newborn baby, do you have any other children? (Please [21 only one.)

Yes 1:] No
0
If leg:
a) How many other children do you have? child(ren)
number
b) Did you ever—at any time—feed your other child(ren) breastmilk? (Please
121 only one.)
I] Yes [I No

 

 

 

I90

 

 

 

 

 

 

 

 

 

\ Please answer the following questions about your work and school schedule. Your name will
/ _r_r_o_t appear in any reports and your answers will orgy be reported by group.
18) Do you currently work for pay? (Please [21 only one.)
D Yes [:1 No
0 0
If _Y__e§: If 1112:
a) Where do you work? (Please [21 a) Did you work for pay while you
only one.) were pregnant with your baby?
[I Only outside of my home (Please [21 only one.)
1:] Mostly outside of my home 1:] Yes [I No
C] Equally inside and outside
of my home

1] Mostly inside my home
D Only inside my home

b) How many hours do
you work each week?
number

c) Did you work for pay while you
were pregnant with your baby?

 

 

 

 

 

 

(Please [21 only one.)
El Yes C] No
19) Do you currently attend school? (Please [21 only one.)
C] Yes Cl No
0 0
If leg: If £9:

a) How many hours do a) Did you attend school while you
you attend school each were pregnant with your baby?
week? (Please 12'] only one.)

number D Yes [:1 No

 

 

 

b) Did you attend school while you
were pregnant with your baby?
(Please [21 only one.)

D Yes C] No

 

 

 

20) Interviews: Interviews may be held at a later date with some women that ﬁll out this survey. If
you participate in an interview, you will be given a $10 gift certiﬁcate to a local grocery store.
These 30 minute interviews will explore the factors that inﬂuence women’s infant feeding
choices. Interviews will be done at the WIC clinic or at another location convenient to you.

If you are chosen. you would be contacted by telephone and may refuse to participate at any time.

Would you be willing to participate in an interview if selected? (Please [21 only one.)

 

1:] Yes 1:] No
0
If jig, may we contact you? Please write a phone number where you can be reached.
PhoneNumber: (_______)____-—_______

 

 

 

Thank you for taking the time to participate in this survey!

191

 

 

Appendix C

APPROVAL LETTER FROM
MICHIGAN STATE UNIVERSITY
COL/IMITTEE ON RESEARCH INVOLVING HUMAN SUBJECTS

 

 

192

REFERENCES

Arafat, 1., Allen, D. E., & Fox, J. E. (1981). Maternal practice and attitudes toward
breastfeeding. JOGN Nursing, 91-95.

Arlotti, J. P., Cottrell, B. H., Lee, S. H., & Curtin, J. J. (1998). Breastfeeding among low-
income women with and without peer support. Journal of Community Health
Nursing, 15(3), 163-178.

Baisch, M. J ., Fox, R. A., Whitten, E., & Pajewski, N. (1989). Comparison of
breastfeeding attitudes and practices: Low-income adolescents and adult women.
Maternal-Child Nursing Journal, 18(1), 61-71.

Baranowski, T., Bee, D. E., Rassin, D. K., Richardson, C. J ., Brown, J. P., Guenther, N.,
& Nader, P. R. (1983). Social support, social inﬂuence, ethnicity and the
breastfeeding decision. Social Science and Medicine, 1 7(21), 1599-161 1.

 

Beal, A. C., Kuhlthau, K., & Perrin, J. M. (2003). Breastfeeding advice given to African
American and White women by physicians and WIC Counselors. Public Health
Reports, 118(4), 368-376.

Bevan, M. L., Mosley, D., Lobach, K. S., & Solimano, G. R. (1984). Factors inﬂuencing
breastfeeding in an urban WIC program. Journal of the American Dietetic
Association, 84(5), 563-567.

Bialeschki, M. D., & Michener, S. (1994). Re-entering leisure: Transition within the role
of motherhood. Journal of Leisure Research, 26(1), 57-74.

Blum, L. M. (1993). Mothers, babies, and breastfeeding in late capitalist America: The
shifting contexts of feminist theory. Feminist Studies, 19(2), 291-311.

Blum, L. M. (1999). At the breast: Ideologies of breastfeeding and motherhood in the
contemporary United States. Boston: Beacon Press.

 

Blum, L. M., & Deussen, T. (1996). Negotiating independent motherhood: Working-class
African American women talk about marriage and motherhood. Gender &
Society, 10(2), 199-211.

193

Brownell, K., Hutton, L., Hartman, J., & Dabrow, S. (2002). Barriers to breastfeeding
among African American adolescent mothers. Clinical Pediatrics, 41(9), 669-673.

Carter, P. (1995). Feminism, breasts and breast-feeding. New York: St. Martin's Press.

Chezem, J., Montgomery, P., & Fortman, T. (1997). Maternal feelings after cessation of
breastfeeding: Inﬂuence of factors related to employment. Journal of Perinatal
and Neonatal Nursing, 11(2), 61-70.

Dennis, C.-L., & Faux, S. (1999). Development and psychometric testing of the
Breastfeeding Self-Efficacy Scale. Research in Nursing & Health, 22, 399-409.

Dennis, C.-L., Hodnett, E., Gallop, R., & Chalmer, B. (2002). The effect of peer support
on breastfeeding duration among primiparous women: A randomized controlled
trial. Canadian Medical Association Journal, 166(1), 21-28.

Dix, D. N. (1991). Why women decide not to breastfeed. Birth, 18(4), 222-225.

Dominguez, S., & Watkins, C. (2003). Creating networks for survival and mobility:
Social capital among African-American and Latin-American low-income
mothers. Social Problems, 50(1), 111-135.

Dusdieker, L. B., Booth, B. M., Seals, B. F ., & Ekwo, E. E. (1985). Investigation of a
model for the initiation of breastfeeding in primigravida women. Social Science
and Medicine, 20(7), 695-703.

Food and Nutrition Service (FNS). (2003, April 22, 2003). Women, Infants, and Children
[WWW document]. Food and Nutrition Service, US. Department of Agriculture.
Retrieved June 1, 2003, 2003, from the World Wide Web:
http://www.fns.usda.gov/wic/

Foucault, M. (1977/1995). Discipline and punish: The birth of the prison (A. Sheridan,
Trans). New York: Vintage Books.

Foucault, M. (1978/1990). The history of sexuality, Volume 1: An introduction (R. Hurley,
Trans). New York: Vintage Books.

Foucault, M. (1984). The Foucault reader. New York: Pantheon Books.

194

 

is
.1-

 

Freed, G. L., Fraley, J. K., & Schanler, R. J. (1992). Attitudes of expectant fathers
regarding breast-feeding. Pediatrics, 90(2), 224-227.

Freysinger, V. J. (1994). Leisure with children and parental satisfaction: Further evidence
of a sex difference in the experience of adult roles and leisure. Journal of Leisure
Research, 26(3), 212-226.

Gabriel, A., Gabriel, K. R., & Lawrence, R. A. (1986). Cultural values and biomedical
knowledge: Choices in infant feeding. Social Science and Medicine, 23(5), 501-
509.

Giugliani, E. R. J ., Caiaffa, W. T., Vogelhut, J ., Witter, F. R., & Perman, J. A. (1994).
Effect of breastfeeding support from different sources on mothers' decisions to
breastfeed. Journal of Human Lactation, 10(3), 157-161.

Grossman, L. K., Fitzsimmons, S. M., Larsen-Alexander, J. B., Sachs, L., & Harter, C.
(1990). The infant feeding decision in low and upper income women. Clinical
Pediatrics, 29(1), 30-37.

Grossman, L. K., Larsen-Alexander, J. B., Fitzsimmons, S. M., & Cordero, L. (1989).
Breastfeeding among low-income, high-risk women. Clinical Pediatrics, 28(1),
38-42.

Guttman, N., & Zimmerman, D. R. (2000). Low-income mothers' views on breastfeeding.
Social Science and Medicine, 50, 1457-1473.

Howell, D. C. (2002). Statistical methods for psychologv (5th ed.). Paciﬁc Grove, CA:
Duxbury.

Humphreys, A. S., Thompson, N. J ., & Miner, K. R. (1998). Intention to breastfeed in
low-income pregnant women: The role of social support and previous experience.
Birth, 25(3), 169-174.

Isabella, P. H., & Isabella, R. A. (1994). Correlates of successful breastfeeding: A study
of social and personal factors. Journal of Human Lactation, 10(4), 257-264.

Jacobson, S. W., Jacobson, J. L., & Frye, K. F. (1991). Incidence and correlates of
breastfeeding in socioeconomically disadvantaged women. Pediatrics, 88(4), 728-
736.

195

 

 

J anke, J. R. (1994). Development of the Breast-Feeding Attrition Prediction Tool.
Nursing Research, 43(2), 100-104.

Jones, D. A. (1986). Attitudes of breastfeeding mothers: A survey of 649 mothers. Social
Science and Medicine, 23(11), 1151-1156.

Jones, D. A. (1987). The choice to breastfeed or bottle feed and inﬂuences upon that
choice: A survey of 1525 mothers. Child: Care, Health and Development, 13, 75-
85.

Kaplan, E. B. (1997). Not our kind of girl: Unraveling the myths of black teenage
motherhood. Berkeley, CA: University of California Press.

Kelly, J. R., & Kelly, J. R. (1994). Multiple dimensions of meaning in the domains of
work, family, and leisure. Journal of Leisure Research, 26(3), 250-274.

Kessler, L. A., Gielen, A. C., Diener-West, M., & Paige, D. M. (1995). The effect of a
woman's signiﬁcant other on her breastfeeding decision. Journal of Human
Lactation, 11(2), 103-109.

Kumar, R., Robson, K. M., & Smith, A. M. R. (1984). Development of a self-
administered questionnaire to measure maternal adjustment and maternal attitudes
during pregnancy and aﬁer delivery. Journal of Psychometric Research, 28(1),
43-51.

Kurinij, N., Shiono, P. H., & Rhoads, G. G. (1988). Breastfeeding incidence and duration
in black and white women. Pediatrics, 81(3), 365-371.

Law, J. (2000). The politics of breastfeeding: Assessing risk, dividing labor. Signs:
Journal of Women in Culture and Society, 25(2), 407-450.

Lewis-Beck, M. S. (1980). Applied regression: An introduction. Newbury Park, CA:
Sage Publications, Inc.

Li, R., Fridinger, F ., & Grummer-Strawn, L. (2002). Public perceptions on breastfeeding
constraints. Journal of Human Lactation, 18(3), 227-235.

196

 

Libbus, M. K., & Bullock, L. F. C. (2002). Breastfeeding and employment: An
assessment of employer attitudes. Journal of Human Lactation, 18(3), 247-251.

Lindberg, L. D. (1996). Women's decisions about breastfeeding and maternal
employment. Journal of Marriage and the Family, 58, 239-251.

Losch, M., Dungy, C. 1., Russell, D., & Dusdieker, L. B. (1995). Impact of attitudes on
maternal decisions regarding infant feeding. The Journal of Pediatrics, 126(4),
507-514.

Maclean, H. (1990). Women's experience of breastfeeding. Toronto: University of
Toronto Press.

Maehr, J. C., Lizarraga, J. L., Wingard, D. L., & Felice, M. E. (1993). A comparative
study of adolescent and adult mothers who intend to breastfeed. Journal of
Adolescent Health, 14(6), 453-457.

Maher, V. (1992). Breast-feeding in cross-cultural perspective: Paradoxes and proposals.
In V. Maher (Ed.), The anthropology of breast-feeding: Natural law or social
construct (pp. 185). Providence, Rhode Island: Berg Publishers Limited.

Matich, J. R., & Sims, L. S. (1992). A comparison of social support variables between
women who intend to breast or bottle feed. Social Science and Medicine, 34(8),
919-927.

McLeod, D., Pullon, S., & Cookson, T. (2002). Factors inﬂuencing continuation of
breastfeeding in a cohort of women. Journal of Human Lactation, 18(4), 335-343.

Murphy, E. (2000). Risk, responsibility, and rhetoric in infant feeding. Journal of
Contemporary Ethnography, 29(3), 291-325.

Nadesan, M. H., & Sotirin, P. (1998). The romance and science of 'breast is best‘:
Discursive contradictions and contexts of breastfeeding choices. Text and
Performance Quarterly, 18, 217-232.

Oyeku, S. O. (2003). A closer look at racial/ethnic disparities in breastfeeding. Public
Health Reports, 118(4), 377-378.

197

 

Park, Y. K., Meier, E. R., & Song, W. O. (2003). Characteristcs of teenage mothers and
predictors of breastfeeding initiation in the Michigan WIC Program in 1995.
Journal of Human Lactation, 19(1), 50-56.

Pugh, L. C., Milligan, R. A., & Brown, L. P. (2001). The breastfeeding support team for
low-income predominantly minority women: A pilot intervention study. Health
Care for Women International, 22, 501-515.

Pugh, L. C., Milligan, R. A., Frick, K. D., Spatz, D., & Bronner, Y. (2002). Breastfeeding
duration, costs, and beneﬁts of a support program for low-income breastfeeding
women. Birth, 29(2), 95-100.

Quarles, A., Williams, P. D., Hoyle, D. A., Brimeyer, M., & Williams, A. R. (1994).
Mothers' intention, age, education and the duration and management of
breastfeeding. Maternal-Child Nursing Journal, 22(3), 102-108.

Robinson, J. B., Hunt, A. E., P0pe, J ., & Garner, B. (1993). Attitudes toward infant
feeding among adolescent mothers from a WIC population in northern Louisiana.
Journal of the American Dietetic Association, 93(1 1), 131 1-1313.

Schafer, E., Vogel, M. K., Viegas, S., & Hausafus, C. (1998). Volunteer peer counselors
increase breastfeeding duration among rural low-income women. Birth, 25(2),
101-106.

Schmied, V., & Lupton, D. (2001). Blurring the boundaries: Breastfeeding and maternal
subjectivity. Sociology of Health & Illness, 23(2), 234-250.

Schwartz, J. B., Popkin, B. M., Tognetti, J ., & Zohoori, N. (1995). Does WIC
participation improve breastfeeding practices? American Journal of Public
Health, 85(5), 729-731.

Seijts, G. H. (2002). Milking the organization? The effect of breastfeeding
accommodation on perceived fairness and organizational attractiveness. Journal
of Business Ethics, 40, 1-13.

Shaw, E., & Kaczorowski, J. (1999). The effect of a peer counseling program on
breastfeeding initiation and longevity in a low-income rural population. Journal
of Human Lactation, 15(1), 19-25.

198

 

Shaw, R. (2003). Theorizing breastfeeding: Body ethics, maternal generosity and the giﬁ
relation. Body & Society, 9(2), 55-73.

Shaw, S. M. (1985). Gender and leisure: Inequality in the distribution of leisure time.
Journal of Leisure Research, 1 7, 266-282.

Shaw, S. M. (1992). Dereifying family leisure: an examination of women's and men's
everyday experiences and perceptions of family time. Leisure Sciences, 14(4),
271-286.

Sheeshka, J ., Potter, B., Norrie, E., Valaitis, R., Adams, G., & Kuczynski, L. (2001).
Women's experiences breastfeeding in public places. Journal of Human
Lactation, 1 7(1), 31-38.

Stack, C. (1975/1997). All our kin. New York: Basic Books.

Stearns, C. A. (1999). Breastfeeding and the good maternal body. Gender & Society,
13(3), 308-325.

Sullivan, J ., & Jones, L. C. (1986). Breastfeeding adoption by low-income black women.
Health Care for Women International, 7, 295-309.

Tiedje, L. B., Schiffrnan, R., Omar, M., Wright, J ., Buzzitta, C., McCann, A., & Metzger,
S. (2002). An ecological approach to breastfeeding. MCN, The American Journal
of Maternal Child Nursing, 27(3), 154-162.

Traina, C. L. H. (2000). Maternal experience and the boundaries of Christian sexual
ethics. Signs: Journal of Women in Culture and Society, 25(2), 369-405.

US. Department of Health and Human Services. (2000). HHS Blueprint for Action on
Breastfeeding. Washington, DC: US. Department of Health and Human
Services, Ofﬁce on Women's Health.

US. Department of Health and Human Services (HHS). (197 9). Healthy People: The
Surgeon General 's report on health promotion and disease prevention.
Washington, DC: US. Government Printing Ofﬁce.

199

US. Department of Health and Human Services (HHS). (1990). Healthy People 2000.
Washington, DC: US. Government Printing Ofﬁce.

US. Department of Health and Human Services (HHS). (2000a). Healthy People 2010:
Understanding and Improving Health. Washington, DC: US. Government
Printing Office.

US. Department of Health and Human Services (HHS). (2000b). HHS Blueprint for
Action on Breastfeeding. Washington, DC: US. Department of Health and
Human Services, Ofﬁce on Women's Health.

Wall, G. (2001). Moral constructions of motherhood in breastfeeding discourse. Gender
& Society, 15(4), 592-610.

Weller, S. C., & Dungy, C. I. (1986). Personal preferences and ethnic variations among
Anglo and Hispanic breast and bottle feeders. Social Science and Medicine, 23(6),
539-548.

Whelan, A., & Lupton, P. (1998). Promoting successful breastfeeding among women
with a low income. Midwifery, 14, 94-100.

Yalom, M. (1997). A history of the breast. New York: Ballantine Books.

Young, 1. M. (1990). Throwing like a girl and other essays in feminist philosophy and
social theory. Bloomington, Indiana: Indiana University Press.

200