a:
5..
in

mum.

g”

2?]...

a». V .

J 6. ft; \uu .
$12.23..
2...: 2.01.:
.12...

‘bhnflfinfi

a...

n “an”:

,-A .u.
km”

1...!
11.: z."
X

 

 

x; F, ‘ Illllllllllllllllllllll‘ll
I J LIBRARY 4

Michigan State
Unlverslty

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

This is to certify that the

thesis entitled
IDENTIFICATION OF MATERNAL DETERMINANTS OF LOW BIRTHWEIGHT

AND GROWTH RETARDATION IN THE MICHIGAN LOW-INCOME
POPULATION.

presented by
Andrea Jane Padgitt

has been accepted towards fulfillment
of the requirements for

M.S. degree in Human Nutrition

u/mtmw/

Major professtg

Date 7/“)1 ‘3 C

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

,I Av '

PLACE IN RETURN 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

 

m

 

 

 

 

 

 

 

 

 

 

 

 

 

1100 6mm.“

 

IDENTIFICATION OF MATERNAL DETERMINANTS OF LOW BIRTHWEIGHT
AND GROWTH RETARDATION IN THE MICHIGAN LOW-INCOME
POPULATION.

By

Andrea Jane Padgitt

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Food Science and Human Nutrition

2000

ABSTRACT

IDENTIFICATION OF MATERNAL DETERMINANTS OF LOW BIRTHWEIGHT
AND GROWTH RETARDATION IN THE MICHIGAN LOW-INCOME
POPULATION.

By

Andrea Jane Padgitt

The low-income population is at high-risk for low birthweight (LBW) and growth
retardation. The Special Supplemental Food Program for Women, Infants and Children
(WIC) provides services to health improve the nutrition and health status of the low-
income population. The goals of this study were 1) to evaluate the Michigan WIC
program using Healthy People 2000 goals and 2) to develop explanatory models for low
birthweight (LBW) and growth retardation in infants and young children. The subjects
were mothers participating in Michigan WIC in 1995 and their offspring from birth up to
age 5 in 1995 to 1998. The Michigan WIC program is performing similar to the national
WIC population and the entire US. population with respect to Healthy People 2000
Objective goals evaluated. Modifiable maternal predictors of birthweight identified
included prepregnancy weight, weight gain, smoking and iron status. Predictors of
stunting at 0-5 months were predominantly related to maternal health status during
pregnancy. LBW was a significant predictor of stunting in infants. Stunting at 36-48
months was predicted by stunting at prior age-groups. The findings of this study are
applicable to the current infrastructure of the Michigan WIC program and will benefit the

low-income population.

Copyright by
Andrea Jane Padgitt
2000

ACKNOWLEDGMENTS

I would like to extend my greatest gratitude to my advisor, Dr. Won Song. She spent
endless hours with me planning, reviewing and discussing this project. I appreciate all of
her thoughtful comments and suggestions throughout the process. I would also like to
thank my committee members Dr. Peg Barratt and Dr. Lorraine Weatherspoon for their
guidance. To my lab mates: Debra Keast, Jean Kerver, Saori Obayashi, Yikyung Park,
Sikhoya Wabuyele, Drs. Chung, Tang and Yang: you all are the most supportive and
encouraging group of women I have ever worked and played with. It has been a true
privilege working alongside each of you for the past two years. Finally, to my parents

whose unconditional cheerleading has brought me to where I am.

iv

TABLE OF CONTENTS

List of Tables
List of Figures
List of Appendices

List of Abbreviations

Chapter One Introduction

Chapter Two Review of Literature

Chapter Three Methods

Chapter Four Results

Chapter Five Discussion

Chapter Six Strengths and Limitations

Chapter Seven Recommendations for Future Studies
Bibliography

ix

52

73

105

113

116

127

LIST OF TABLES

Chapter Two

Table 1. Odds ratios associated with risk of smoking during pregnancy for certain
birth outcomes.

Table 2. Birthweight associated with smoking during pregnancy.

Chapter Three

Table 3. The raw and final sample sizes of subjects included in the study.

Table 4. Variables deleted from records or exclusion criteria for PNSS 1995.
Table 5. Percent of subjects excluded for each error code.

Table 6. Characteristics of women participating in Michigan WIC at the time of

enrollment, PNSS 1995.

Table 7. The sample sizes of subjects in the merged longitudinal dataset, PNSS
1995 and PedNSS 1995-1998.

Chapter Four

Table 8. Public health indicators of Healthy People 2000 Objectives of Michigan
WIC population in 1995.

Table 9. Prevalence of prepregnancy overweight in Michigan WIC women aged 20
and older in 1995 in comparison with the goal for Healthy People 2000
Objective.

Table 10. Prevalence of anemia in pregnant and postpartum women in Michigan

WIC program in 1995 vs. Healthy People 2000 Objective.

Table 11. Trimester of pregnancy when prenatal care began at the time of WIC
enrollment among Michigan WIC women in 1995 vs. Healthy People
2000 Objective.

Table 12. Tobacco use during third trimester of pregnancy by Michigan WIC
women (PNSS 1995) vs. Healthy People 2000 Objective.

Table 13.

Table 14.

Table 15.

Table 16.

Table 17.

Table 18.

Table 19.

Table 20.

Table 21.

Table 22.

Table 23.

Table 24.

Table 25.
Table 26.
Table 27.
Table 28.

Table 29.

Alcohol use during third trimester of pregnancy by Michigan WIC women
in 1995 vs. Healthy People 2000 Objective.

Incidence of low birthweight (LBW) and very low birthweight (VLBW)
among infants born to Michigan WIC women in 1995 vs. Healthy People
2000 Objective.

Prevalence of growth retardation in Michigan WIC infants and children by
age-groups in 1995-1998 vs. Healthy People 2000 Objective.

Prevalence of growth retardation in Michigan WIC infants and children by
race/ ethnicity in 1995-1998 vs. Healthy PeOple 2000 Objective.

Correlation coefiicients for independent variables with birthweight.

Multiple linear regression model to predict birthweight by maternal
characteristics.

Predictors for low birthweight identified by discriminant analysis.

Odds ratios and confidence intervals for low birthweight by maternal
characteristics.

Specificity, sensitivity, negative predictive value (PV-) and positive
predictive value (PV+) of discriminant function and logistic regression
model in Michigan WIC population, 1995.

Pearson’s (r) correlation coefficients of predictors an height-for-age
percentile of Michigan WIC infants and children, 1995-1998.

Spearman’s (rank) correlation coefficients of predictors and height-for-age
percentile of Michigan WIC infants and children, 1995-1998.

Pearson’s (r ) correlation coeflicients for birthweight and height-for-age
percentile of Michigan WIC infants and children, 1995-1998.

Predictors for stunting at 0-5 months identified by logistic regression.
Predictors for stunting at 6-11 months identified by logistic regression.
Predictors for stunting at 12-23 months identified by logistic regression.
Predictors for stunting at 24-35 months identified by logistic regression.

Predictors for stunting at 36-48 months identified by logistic regression.

vii

Table 30.

Table 31.

Chapter Five

Table 32.

Appendix A

Table 33.

Classification tables for logistic regression models predicting stunting in
the Michigan WIC population, 1995-1998.

Specificity, sensitivity, negative predictive value (PV-) and positive

predictive value (PV+) of logistic regression models for predicting
stunting in the Michigan WIC population, 1995-1998.

Michigan WIC program vs. US. population in progress toward
achievement of Healthy People 2000 Objectives.

Hemoglobin and hematocrit cut-off values for anemia in pregnant and
postpartum women.

viii

LIST OF FIGURES

Chapter Three

Figure 1. Merging the datasets to create a longitudinal dataset.

ix

APPENDICES

Appendix A Definitions and standards of variables

Appendix B Correlation coefficients between predictors

LIST OF ABBREVIATIONS

AFDC ............................................................................................................. Aid for Dependent Children
BMI ................................................................................................................................. Body mass index
CDC ......................................................................................... Centers for Disease Control and Prevention
DHHS ................................................................. United States Department of Health and Human Services
EPSDT ....................................................................... Early Periodic Screening, Diagnostic and Treatment
g ......................................................................................................................................................... gram
10M ........................................................................................................................... Institute of Medicine
IU GR ......................................................................................................... Intrauterine growth retardation
LBW ................................................................................................................................. Low birthweight
kg ................................................................................................................................................. kilogram
mcg ........................................................................................................................................... micrograrn
mg ............................................................................................................................................... milligram
MCH ............................................................................................................ Maternal and Child Programs
MMWR ........................................................................................ Morbidity and Mortality Weekly Reports
NBW ............................................................................................................................ Normal birthweight
NCHS ................................................................................................ National Center for Health Statistics
ng ................................................................................................................................................ nanogram
NHANES ..................................................................... National Health and Nutrition Examination Survey
OR ............................................................................................................................................. Odds ratio
PedN SS ......................................................................................... Pediatric Nutrition Surveillance System
PNSS .......................................................................................... Pregnancy Nutrition Surveillance System
SES ............................................................................................................................ socioeconomic status
SGA .................................................................................................................... Small-for-gestational age
U.S. ...................................................................................................................... United States of America
USDA .......................................................................................... United States Department of Agriculture
VLBW ....................................................................................................................... Very low birthweight
WIC ....................................... Special Supplemental Nutrition Program for Women, Infants and Children
y ........................................................................................................................................................... year

Chapter 1

INTRODUCTION

1.1 Background

The Special Supplemental Nutrition Program for Women, Infants, and Children
(WIC) provides services to help improve the nutrition and health status of eligible low-
income pregnant, breastfeeding and postpartum women; and infants and children up to
five years of age. The mission of the WIC program is “to safeguard the health of low-
income women, infants and children by providing nutritious foods to supplement diets,
information on healthy eating and referrals to healthcare” (www.fns.usda.gov/wic).

Since the start of the program in 1974, WIC has grown in the number of
participants, from 344,000 participants per year in 1975 to 7.4 million people per month
in 1997 (Owen and Owen, 1997; www.fns.usda.gov/wic). The Michigan WIC program
provides services to more than 230,000 women and children aged 0-5 years monthly at
230 WIC clinics in all 83 counties. One out of two children born to Michigan women
receives WIC services. Therefore, the effectiveness of WIC to improve the health status
of Michigan low-income women and children is of great magnitude.

WIC has aided in substantial improvement of the health status of our nation’s
low-income, nutritionally at-risk families (Devaney et al., 1997; Rose et al., 1998).
Specifically, WIC participants have been shown to have a significantly lower prevalence

of small-for-gestational age (SGA) and low birthweight (LBW) infants (Ahluwalia et al.,

reduce the incidence of LBW. First, to identify the subgroups that were at highest risk of
LBW and to concentrate on reducing the LBW rate in these groups. Second, to identify
those predictors of LBW, which were modifiable and develop specific programs to
alleviate them where feasible. Reduction in the incidence of LBW will be cost beneficial
for the entire US. population. Distinguishing the characteristics of LBW and NEW was
based on gestational history through examination of the lifestyle, behavioral and health
status characteristics of the mother during pregnancy, while considering the infant

characteristics such as gestational age and sex.

1.3 Growth retardation

Growth retardation is an indicator of long-term health and nutritional history of a
child or a population. On an individual level, stunting can reflect the normal variation of
growth. On a population level, stunting generally reflects poor socioeconomic, health,
and nutritional conditions. Birthweight was a strong predictor of childhood growth
(Binkin et al., 1988). Poverty was closely related to the prevalence of growth retardation
(Yip et al., 1992). Therefore, the low-income population is at increased risk for stunting.
Stunted growth can result from poor nutrition and/or an increased number of infections.
Stunting had detrimental consequences on cognitive and psychosocial development
lasting into late childhood (Miller and Korenman, 1994).

Identification of the predictors of stunting in the Michigan WIC population was
important because this could facilitate program evaluation and the development of

targeted intervention. Stunting is frequently studied in developing countries where the

prevalence is higher than the US. (Adair and Guilkey, 1997; Diaz et al., 1995; Frongillo
et al., 1997; Mendez and Adair, 1999). Limited research is available for determinants of
stunting in the US. One of the Healthy People 2000 Objectives is directed toward low-

income children, reflecting the understood disparities in this population.

1.4 Program evaluation

The low-income population is at a greater risk for nutritional and health
disparities than individuals with middle or upper incomes. The WIC program serves this
underserved population with the goal of reducing nutritional disparities and improving
birth outcomes and growth status of its participants. The Healthy People 2000 Objectives
were designed to promote healthier longer-living US. population. It is important to
assess the Michigan W1C population by Healthy People 2000 Objectives to determine
which goals have been met, which have not, to what extent this population is moving in
the right direction, and which are moving away from the target. By answering these
questions, the Michigan WIC program will have a foundation to develop programs and
seek greater amounts of firnding to be used to develop these specific targeted
interventions.

All programs, especially those receiving finds from the federal government,
require continuous and critical evaluations to ensure the efficacy of the program for the
population being served. The WIC program is evaluated and information is reported by
the Centers for Disease Control and Prevention (CDC) in Morbidity and Mortality
Weekly Reports (MMWR) (Yip et al., 1992) and other CDC-sponsored publications.

Cost benefit analyses of specific interventions or applications are reported in

peer-reviewed journals (Riordan, 1997; Tuttle and Dewey, 1996). The Michigan WIC
program is evaluated and reported through Executive Summaries provided to Michigan

Department of Community Health oflicials.

1.5 Rationale

The overlying goal of this project was to evaluate the Michigan WIC program
through critical analysis of the longitudinal data collected by this program. One way of
evaluating programs is by setting goals and developing standards. The US. Department
of Health and Human Services (DHHS) released in September, 1990 the Healthy People
2000: National Health Promotion and Disease Prevention Objectives. The purpose of the
Healthy People 2000 Objectives, 319 specific objectives in 22 categories, was to provide
a goal for “increasing the span of healthy life, reducing health disparities, and achieving
access to preventative services for all Americans by the end of the century” (Public
Health Service, 1991). No published research was available comparing the low-income
population with respect to the Healthy People 2000 Objectives. It was important to
remedy this lack of information, so that public officials know how well the low-income
population is meeting national health objectives. Knowledge will lead to better focused
intervention and intervention to improved health and nutritional status.

To the best of knowledge, no efforts have been made at the state level to create or
evaluate the longitudinal data set developed, nor investigators considered the situation
holistically, incorporating all aspects of the perinatal period and develop predictive
models. The findings from this secondary data analysis can be utilized for subsequent

interventions and important public health policy decisions by the national and state WIC

program administrators, and health and nutrition screeners and educators. The ultimate
goals of the project were to enhance efficiency and effectiveness of data management of
the Michigan and national WIC program, by merging techniques and information
available today.

Limited resources, time and, sometimes, expertise at the state level of WIC
programs led to extensive amounts of data collected, and then primarily used to fulfill
federal requirements. Local public health monitoring efforts are important to modify local
programs that are best tailored to the population being served. In order for the data to be
comparative, local public health officials must be involved in the development of data
collection and surveillance systems, data analysis and finally interpretation of results.
Closer evaluation of existing data provides in-depth information that has not yet been
fully assessed while supporting the efl’ort of collecting the data.

Descriptive statistics on WIC data have been performed in the past and some
disparate results reported in state-limited publications. Critical evaluation of the current
system to find explanatory models in order to reinforce its strengths and identify possible
remedies for the weaknesses is viewed of importance to enhance the success of the WIC

education programs.

1.6 Goals

The goals of this study were: 1) to evaluate the Michigan WIC program using
Hedthy People 2000 Objective goals and 2) to develop explanatory models for LBW
infants and growth retardation in infants and young children, through systematic and

innovative approaches for evaluating longitudinal and cross-sectional community-based

data sets generated fiom low-income, nutritionally at-risk pregnant women and their
children (aged 0-4 years) being served by the Michigan WIC program.

The findings of this study will serve two main purposes: 1) program evaluation of
the successes and limitations of the Michigan WIC program in achieving the goals
previously stated by comparing the Michigan WIC program with regard to Healthy
People 2000 Objectives; 2) the focus of targeted intervention to reduce LBW and grth
retardation will be determined. Similarly, the predictors of LBW and growth retardation
identified in this study are to guide program omcials to develop new or modify

interventions or to increase efficiency with concentrated efforts.

1.7 Objectives
The objectives of this study in the Michigan WIC population were:

1) To assess the percentage of the population who achieve nine selected Healthy
People 2000 Objectives that are related to maternal and child health.

2) To predict the birthweight of offspring by maternal health status, lifestyle and
sociodemographic characteristics in the women. Maternal health status will be
measured by iron status, prepregnancy BMI, weight gain during pregnancy.
Lifestyle factors will be measured by tobacco and alcohol use during pregnancy,
when prenatal care was obtained and participation in federal government
programs such as Food Stamps, Medicaid and Aid for Dependent Children
(AFDC). Demographic characteristics include race/ ethnicity, age, level of

education, household monthly income, and household size.

3) To identify the factors difl’erentiating LBW and NBW infants in the population
based on maternal health, lifestyle, and sociodemographic characteristics during
pregnancy.

4) To identify the factors associated with growth retardation among infants and

young children in the Michigan WIC population.

1.8 Hypotheses

l) The Michigan WIC population achieved the nine selected Healthy People 2000
Objectives evaluated in this study.

2) Birthweight of infant is predicted by specific maternal factors during pregnancy,
such as prepregnancy BMI, weight gain during pregnancy, tobacco use during
pregnancy, and when prenatal care began.

3) There are distinguishing maternal characteristics of LBW infants when compared
to NBW infants. The mother’s of LBW infants will have a higher incidence of
anemia, gain less weight during pregnancy, and have later entry into prenatal care.

4) Growth retardation in children is predicted by maternal factors during pregnancy,
birth outcomes and stunting at prior age-group categories. LBW infants are more

likely to be stunted than NBW infants.

Chapter 2

REVIEW OF LITERATURE

2.1 WIC

The WIC program began in 1972 as a pilot program and became a permanent
program in 1974. Headed by the United States Department of Agriculture (USDA) Food
and Nutrition Service, WIC provides supplemental foods, health care referrals and
nutrition education for its participants. (www.fns.usda.gov/wic). This includes providing
nutritious foods to mothers during pregnancy and formula to infants through the formula
rebate system. WIC also provides nutrition education and counseling.

There are certain eligibility requirements, based on income and nutritional status.
Only pregnant or postpartum women, infants and children up to their fifth birthday are
eligible (wwwfirsusdagov/wic). The income requirement is that the applicant’s gross
income must be at or below 185 percent of the poverty level. For example, for a family
of four, the annual income must be less than $30,895 (effective July 1, 1999 - June 30,
2000). Nutritionally at-risk is defined by two criteria: first, medically based risks such as
anemia, underweight, or poor pregnancy outcomes; second, nutrition based risks include
inadequate dietary patterns. A physician, nutritionist or nurse evaluate the nutritional
status of an applicant.

After eligibility has been established a priority system is put in place because it is
not possible to serve all eligible persons. Priority levels for the national WIC program
are: 1) pregnant women, breastfeeding women, and infants determined to be at

nutritional risk because of serious medical problems, 2) infants up to six months of age

whose mothers have participated in WIC or could have participated and had serious
medical problems, 3) children up to age five years at nutrition risk because of serious
medical problems, 4) pregnant or breastfeeding women with any nutritional risk, 5)
children up to five years of age at nutritional risk because of dietary problems, 6) non-
breastfeeding, postpartum women with any nutritional risk, 7) individuals at nutritional
risk only because they are homeless or migrants and current participants who without
WIC foods could continue to have medical and/ or dietary problems. Approximately

81% of all eligible persons are served by the US. WIC program (www.fns.usda.gov/wic).

2.2 Background information on datasets used in this study

The CDC coordinates the collection of extensive data at the community-based
level from expecting mothers, for the Pregnancy Nutrition Surveillance System (PNSS),
and fiom their children, for the Pediatric Nutrition Surveillance System (PedNSS). Data
in these two systems are used for multiple purposes: 1) to assist health professionals to
identify and reduce pregnancy related health risks, which contribute to adverse pregnancy
outcomes; 2) to furnish useful and timely data to participating states to allow them to
monitor trends in the prevalence and incidence of major factors for LBW, infant mortality
and morbidity, and to monitor infant feeding practices; 3) to monitor trends and patterns
in key indicators of nutritional status of children birth to five years of age; 4) to allow
each state to plan educational interventions and make referrals; and 5) to monitor the
progress toward Healthy People 2000 Objectives. PNSS and PedNSS are convenience
samples and provide valuable information on the majority of low-income women, infants

and children.

 

2.2.1 Pregnancy Nutrition Surveillance System

Since the inception of PNSS in 1979, the emphasis and design have changed in
order to improve quality of data collection. The current emphases are “to better quantify
preventable risk behaviors and to examine the relationship of nutrition and behavioral
risks during pregnancy to birth outcomes” (www.cdc.gov/nccdphp/dnpa/PNSS.htm).
Data is collected from trained interviewers at local WIC clinics, this is done through a
questionnaire and measurements taken by standard procedures. Information collected for
mothers in PNSS are demographics, anthropometrics, nutritional risk factors, health
behaviors and birth outcomes. (Appendix A provides detailed information on variables).
The CDC has developed a list of targeted nutritional and behavioral interventions for the
PNSS: 1) quality data collection, 2) promotion of adequate iron intake, reduction in
alcohol intake, and appropriate weight gain through nutrition education sessions and
materials, 3) breastfeeding promotion in the low-income population, 4) outreach activities
promoting early identification of pregnancy and entry into comprehensive prenatal care,
including WIC services, 5) smoking cessation for all pregnant women, especially women
who are underweight and women who are older and underweight, and 6) referrals of
clients to alcohol treatment services to reduce drinking behaviors as appropriate

(www.cdc.gov/nccdphp/dnpa/PNSS.htm).
2.2.2 Pediatric Nutrition Surveillance System

The PedNSS was developed out of findings from the USDA’s Ten State Nutrition

Survey (www.cdc.gov/nccdphp/dnpa/PedNSS.htm). This survey described the

10

nutritional status of low-income families as unsatisfactory. In 1973, PedNSS began
monitoring growth, iron status and breastfeeding of nutritionally at-risk children.
PedNSS is designed to serve as a program-based surveillance system. State and local
health departments work in conjunction with the CDC to analyze data and evaluate
programs, such as W1C, Early Periodic Screening, Diagnosis and Treatment (EPSDT)
and clinics funded by Maternal and Child Health Program (MCH) block grants (PedNSS
User Manual, 1994). Information is collected by a trained interviewer with questionnaire
answered by caregiver and measurements taken by standard procedure. Data collected
for children in PedNSS include demographics, anthropometric measurements, iron status

and infant feeding method. (Appendix A provides detailed information on variables).

2.3 Healthy People

The Healthy People 2000 Objectives serve as a vision for America, "characterized
by significant reductions in preventable death and disability, enhanced quality of life and
greatly reduced health disparities in the health status of populations within our society."
The objectives serve as a call to action for national, state and local public health officials
to strive to attain (Public Health Service, 1991). The National Center for Health Statistics
(NCHS) is responsible for monitoring the progress. DHHS and NCHS recognize that
although these are national objectives, state and local officials will do the majority of the
work towards the achievement of the goals.

Healthy People 2010 was released at the Partnerships for Health in the New
Millennium in January, 2000 (www.health.gov/healthypeople/default.htm). There are 28

focus areas. The Healthy People 2010 Objectives were developed through collaborative

11

efforts of 350 national membership organizations and 250 state health, mental health,
substance abuse and environmental agencies. The goal is to "provide the objectives in a
format that enables diverse groups to combine their efforts and work as a team." The
leadership understands the importance of the involvement of the local and state officials.
Examples of maternal and child health Healthy People 2010 Objectives:

1) Increase to at least 60 % the prevalence of healthy weight (defined as a BMI 219.0 and
s 25.0) among all people 20 years and older. Reduce to less than 15 % the prevalence of
BM] 2 30.0 among people aged 20 and older.

2) Increase the proportion of mothers who achieve a weight gain consistent with the
Institute of Medicine guidelines during their pregnancies.

3) Increase abstinence from tobacco use by pregnant women to 95 percent.

4) Increase abstinence from alcohol use by pregnant women, including any use in the past
month to 95% and binge drinking in the past month to 99%.

5) Increase to at least 90 percent the proportion of all pregnant women who begin
prenatal care in the first trimester. Increase to at least 90percent the proportion of all
live-born infants whose mothers receive prenatal care that is adequate or more than
adequate according to the Adequacy of Prenatal Care Utilization Index.

6) Reduce iron deficiency to 5 % or less among children aged 1 and 2 years; to less than

1 % among children aged 3 and 4 years; and to 7 % or less among females of
childbearing age. Reduce anemia among low-income pregnant women in their third
trimester of pregnancy to 23%.

7) Reduce LBW to an incidence of no more than 5 percent of live births and VLBW to no

more than 1 percent of live births.

12

8) Increase to at least 75 percent the proportion of mothers who breastfeed their babies in
the early postpartum period; to at least 50 percent the proportion of mothers who
breastfeed until their babies are 6 months old; and to at least 25 percent the proportion
who breastfeed until their infants are 1 year old.

9) Reduce growth retardation among low-income children aged 5 years and young to 5%

or less.

2.4 Maternal prepregnancy weight status

2.4.1 Measurements

Maternal height is an indicator of genetic potential and early environmental
conditions while maternal weight is an indicator of current environmental conditions,
including nutritional status (Kirchengast and Hartmann, 1998). Body mass index (BMI)
is an indicator of energy stores (Cogswell and Yip, 1995). It is important to realize that
once conception occurs, prepregnancy BMI and the risk factors associated with it cannot
be changed (Copper et al., 1995).

Due to the fact that the majority of studies reviewed and the present study rely on
self-reported prepregnancy weight, it is critical to determine the validity of these
measurements. Schieve et al., (1999) performed analysis of this issue using the 1988
National Maternal and Infant Health Survey. Previous studies of adult non-pregnant
women found that women tend to underreport their weight and the degree of
underreporting increases with increasing BMI (Rowland, 1990). Similar results were

found in the analysis 3,518 postpartum women. On average, women reported their

13

delivery weight (weight at the time of delivery) to be 1.28 kilogram (kg) less than the
measured delivery weight retrieved from hospital records. When measured weight was
used to classify weight gain, 30-40% of cases were classified in a different category than
when self-reported weight was used. The authors cite that this may attenuate the results
of epidemiological studies and self-reported weight should be interpreted with caution.
The authors also noted certain patterns: women who were underweight prepregnancy and
had low weight gain during pregnancy over-reported while women with high BMI

prepregnancy and high weight gain during pregnancy underreported.

2.4.2 Effects of prepregnancy weight on pregnancy health, complications

There has been an increase in prepregnancy overweight (BMI 2 26.0 to .<_ 29.0) in
the low-income population monitored through PNSS, (19.4% in 1979 and 32.6% in 1993)
(Perry et al., 1995). There are complications associated with being overweight and obese
(2 20% and 2 40% desirable body weight, respectively), including diabetes,
hypertension, dyslipidemia, cardiovascular disease, stroke, gout, sleep apnea,
osteoarthritis, menstrual irregularities, and some forms of cancer (Crane et al., 1997).
Complications during pregnancy can also result from obesity (BMI 2 29.0), such as
gestational diabetes (Odds ratio (OR)=2.00) and hypertension (OR=2. 15). (Crane et al.,
1997) Edwards et al. (1996) found higher odds ratios associated with complications of
obesity (BMI 229.0) during pregnancy: gestational diabetes (OR=6.8), hypertension
(OR=3.4) compared to normal weight women in a study of 683 obese women and 660

normal weight women. Obesity (BMI 2 29) was also associated with a greater risk of

14

cesarean delivery (OR=3 .2) in an analysis of the Central New York Regional Perinatal
Data System (Edwards et al., 1996).

Russell et al. (1995) performed a simulation study using the First National Health
and Nutrition Examination Survey (NHANES-I) and the Nutrition Examination Survey
Epidemiology Follow-up Study with an overweight baseline of 24.7% in the general US.
population. The simulated three interventions: 1) no weight gain among non-overweight
individuals, 2) weight loss among overweight individuals, and 3) a combination of the
two simulations. Results showed that no weight gain by non-overweight individuals
would achieve the goal of 20% in the overall population over a six-year period. The third
intervention including a 3 .8 kg weight loss by overweight individuals will achieve the
Healthy People 2000 Objective goal to reduce prevalence of overweight in all race-sex
strata. Therefore, it may be possible to achieve the goal to reduce the prevalence of

overweight in the general U. S. population through the aforementioned measures.

2.4.3 Effects of prepregnancy weight on birth outcomes

There is a significant and linear relationship between prepregnancy weight and
infant birthweight that is independent of gestational age (Wilcox and Marks, 1994; Perry
et al., 1995). A ISO-gram (g) increase in birthweight is found in women with a BMI>29,
conversely a 162-g reduction in birthweight is found in women with a BMI<19.8
(Cogswell and Yip, 1995). The incidence of macrosomia (birthweight >4000g) increases
with increasing prepregnancy weight (W rlcox and Marks, 1994).

A study of 18,519 women in Vienna, Austria found through regression analysis

that prepregnancy BMI, height, weight gain during pregnancy and maternal age were the

'15

most important predictors of birthweight. High prepregnancy weight was linearly and
positively related to birthweight, birth length, and head circumference. The authors
stressed that maternal nutrition status both prior to and during pregnancy are important
predictors of fetal growth, and that poor nutrition status of the mother may diminish the
genetic effects on growth. In addition, results of this study showed that a poor nutrition
status prior to pregnancy cannot be overcome by excessive weight gain during gestation
(Kirchengast and Hartmann, 1998).

The association between prepregnancy underweight and LBW has been
documented since the 19505 (Perry et al., 1995). Among low-income women in PNSS,
there is a greater percentage of LBW in prepregnancy underweight compared with
normal weight and overweight prepregnancy (Perry et al., 1995). A study of 9,651
pregnant, low-income women found an odds ratio of 1.98 for the association between
BMI < 19.8 and preterm birth. Siega—Riz et al. (1996) cite the reason for this association
in thin women being due to the fact that small body size reduces the capacity for fluid
expansion during pregnancy.

The Nutrition Collaborative Research Support Program is a cross-country study of
the relationships between malnutrition and human function in Egypt, Kenya and Mexico.
The average weight gain in these countries was one-half (approximately 7 kg) of the U. S.
recommendations. This study found an inverse relationship between maternal
prepregnancy BMI and weight gain during pregnancy. Interestingly, the mothers of all of
the LBW infants born in Kenya had a prepregnancy BMI <21. In a logistic regression
model to predict LBW, maternal BMI, parity, maternal hemoglobin, and SES correctly

classified 85% of cases. Weight retention patterns differed between obese and

16

underweight mothers in this study population. Women with a low prepregnancy BMI
gained more fat during gestation measured by skinfold thickness (biceps, triceps,
subscapular, and suprailiac) and lost more fat and weight postpartum through lactation.
Women with a high prepregnancy BMI lost fat during pregnancy, however regained it
during lactation. Longitudinal analysis showed maternal prepregnancy BMI was
positively correlated with infant weight and length at 3-6 months of age (Allen et al.,

1994)

2.5 Weight gain during gestation

2.5.1 Recommendations (past and present)

During the first part of the twentieth century, weight gain guidelines were
restrictive based on beliefs that excessive weight gain was associated with obstetrical
complications; In 1970, the National Academy of Sciences recommended women gain
20-25 pounds during pregnancy. The restrictive guidelines of the 19705 were related to
concerns about preeclarnpsia, labor and delivery complications, and maternal weight
retention associated with large amounts of weight gain during pregnancy (Scheive et al.,
1998). In 1983, the recommendations increased slightly. The American Academy of
Pediatrics and the American College of Obstetricians and Gynecologists adopted
guidelines that women gain 22-27 pounds during pregnancy. The Institute of Medicine
(IOM) released revised guidelines for weight gain during pregnancy in 1990. Institute of
Medicine’s (1990) standards of minimum recommended weight gain based on the

prepregnancy weight status: 28-40 pounds for underweight women (BMI 3 19.70);

17

25-35 pounds for normal and overweight women, (BMI 19.8 — 29.0); 15-25 pounds for
very overweight women (BMI 2 29.1). The revised guidelines are based on the woman’s
prepregnancy weight. The current guidelines reflect an increase over previous
recommendations (Witter et al., 1995). The guidelines were based on the review of
epidemiological data on fetal death, prematurity, and LBW. The expert panel concluded
that poor maternal weight gain is a predictor of poor pregnancy outcomes. (Parker and
Abrams, 1992).

Johnson and Yancey (1996) dispute the literature reviewed by the IOM citing
methodological flaws and no provision of causal relationships and lack of causal
biological mechanisms. Johnson et al., (1992) believe that the increase in weight gain
recommendations may be more harmful to the mother and override the benefits of
reducing LBW cited by the IOM. The authors felt that if these recommendations were
followed, it would be a major contributing factor to postpartum obesity. Excessive weight
gain and subsequent weight retention predisposes women to associated complications of
obesity.

Conversely, Bracero and Byrne (1998) evaluated the IOM recommendations and
found that optimal pregnancy outcomes occurred over a range of weight gain, and were
higher than the IOM recommendations. Overall, the best perinatal outcomes were
associated with weight gains of 3 1-40 pounds in their study. Gains of 36-40 pounds were

associated with the highest infant survival rate, regardless of prepregnancy BMI.

18

2.5.2 Measurement and components of weight gain

Maternal weight gain is a simple and noninvasive technique to monitor the health
of the pregnancy (Parker and Abrams, 1992). In the majority of papers reviewed, a
prepregnancy self-reported weight and a measured weight within one week of delivery
were used to calculate gestational weight gain. Selvin and Abrams (1996) discussed the
implications of using total weight gain versus net wet gain. Total weight gain may
overestimate the relationship between the mother and birth outcomes because the weight
of the infant accounts for more than 25% of gestational weight gain. The authors suggest
using the net weight gain, which accounts for infant birthweight (net weight gain = total
weight gain — infant birthweight). Net weight gain removes the “artificially induced
structural bias.”

There are four major components of weight gain during pregnancy: 1) maternal
reproductive tissue involving the growth of breast and uterine tissues, 2) increase in
maternal blood volume, 3) growth of fetus, placenta, and amniotic fluid, 4) maternal fat
stores, which contain the energy needed for fetal growth (Kirchengast and Hartmann,
1998; Wilcox and Marks, 1994). Approximately one-half of gestational weight gain is
accounted by fetal and placental weight gain in the third trimester of pregnancy (Strauss

and Dietz, 1999).

2.5.3 Efl’ects of weight gain on birth outcomes
Adequate weight gain is influenced by a variety of factors, including
prepregnancy weight, age, race and SES (Perry et al., 1995; Caulfield et al., 1996). The

effects of weight gain are often modified by prepregnancy BM] (Cogswell and Yip,

19

1995). A study of 20,971 pregnancies found that weight gain was inversely correlated
with prepregnancy weight, prepregnancy BMI, maternal age, and positively correlated
with maternal height (Bracero and Byrne, 1998). Similarly, weight gain is strongly and
negatively correlated with maternal triceps skinfold thickness (Allen et al., 1994). The
effects of prepregnancy BMI on weight gain were demonstrated in a study of 668 obese
women and 660 normal weight women, where obese women gained one-third less weight
than normal weight women (Edwards et al., 1996).

Weight gain during pregnancy is related to pregnancy outcomes, such as
' birthweight. In a study of 20,97 1 pregnancies in Brooklyn, New York, maternal weight
gain was correlated with birthweight, but accounted for only 4% of the variation in
birthweight. 56.9% of women in this study gained within the IOM recommendations for
ideal weight gain (Bracero and Byme, 1998). Birthweight and weight change during
pregnancy were linearly correlated. Weight gain during pregnancy was more strongly
associated with infant outcomes, such as birthweight, than maternal complications such
as cesarean delivery, gestational diabetes, and hypertensive disorders (Edwards et al.,
1996)

Fourteen percent of LBW in the US. has been attributed to inadequate gestational
weight gain during pregnancy (IOM, 1990). Gaining less than the IOM recommended
amount of weight is associated with a 120-g reduction in birthweight. In addition, LBW
rates were doubled in women who gained less than recommended amount (Cogswell and
Yip, 1995). In a study of the PNSS 1995 at the national level, the incidence of LBW was
highest among infants born to women who were underweight prior to pregnancy and

gained less than the recommended amount during pregnancy (Perry et al., 1995).

20

Gaining excessive amounts (greater than IOM recommended amount) of weight
during pregnancy was associated with fetal macrosomia (birthweight >4000 g), delivery
complications, and excessive weight retention (Perry et al., 1995; Johnson et al., 1992).
In addition, gaining more than the recommended amount is associated with 150-g
increase in birthweight. The rates of high birthweight (>4000 g) double with weight gain
exceeding the recommendation (Cogswell and Yip, 1995). Other complications with
excessive weight gain include increased incidence of prolonged and difficult labor,
postdatism, fetal heartrate anomalies, meconium staining, postpartum hemorrhage, fetal
trauma, neonatal resuscitation, and operative complications (Hickey et al., 1995). In an
analysis of the national PNSS 1995, gaining more than recommended amount of weight
decreased with increasing age (Perry et al., 1995).

Cesarean delivery is a complication of pregnancy related to increased weight gain.
Cesarean delivery is associated with greater maternal age, higher prepregnancy BMI,
short stature, preeclampsia, previous cesarean delivery, and high infant birthweight
(>3 591 g). Cesarean delivery may be performed in order to protect the fetus in cases of
fetal distress, fetal growth restriction, malpresentation, and multiple gestation (\Vrtter et
al., 1995).

In an analysis of 3,870 deliveries at Johns Hopkins Hospital, Caulfield et al.
(1996) examined the predictors of weight gain outside the IOM recommendations.
Women gaining less than guidelines were younger, shorter, thinner, less educated, more
likely to be black, smokers and less likely to be hypertensive. Women who gained more
than the recommended amount were taller, heavier, primiparous, and more likely to be

white and hypertensive. Only one-third of the women in this study gained the IOM

21

recommended amount. The authors noted a 2 kg weight gain difference between black
and white women.

Maternal weight gain has been implicated with other birth outcomes, including
preterm delivery. A study of 9,651 low-income, pregnant women in West Los Angeles
found that low weight gain during the third trimester increased the risk of preterm
delivery by 90% (Siega-Riz et al., 1996). Weight gain during the first and second
trimesters was not associated with preterm birth in a low-income sample of 1,015
mothers. However, weight gain during the third trimester was related to preterm delivery
in this sample (Hickey et al., 1995). Conversely, Johnson and Yancey (1996) found a
lack in relationship between inadequate weight gain and preterm delivery citing the fact
that in the past twenty years, there has been an increase in prepregnancy weight and
weight gain during pregnancy and no decrease in the rate of preterm birth.

In addition to total weight gain over the entire gestation, patterns of weight gain
during each trimester offer evidence of the importance of timing of adequate weight gain
on birth outcomes. Low weight gain in the first trimester is not associated with the
prevalence of intrauterine grth retardation (IUGR) (birthweight <2500 g and full
term). Low weight gain in the second and third trimesters significantly increased the risk
of IUGR. The authors cited these results to be significant in that the majority of mothers
with low weight gain in a particular trimester, but had overall normal weight gain
(Strauss and Dietz, 1999). Data from the Dutch famine cohort provide a unique natural
experiment for examining the effects of timing of weight gain on birth outcomes. Famine

during mid-pregnancy was associated with reduced birth length with little effect on

22

 

ponderal index (kg/m3) of the infant. Famine during late pregnancy was associated with
reduced ponderal index and birth length. (Stein et al., 1995).

Copper et al. (1995) evaluated mother’s attitudes toward weight gain during
pregnancy. The author’s developed a Likert scale to describe the attitudes and feelings
towards weight gain, perceived impact of weight gain on the health of the baby, and how
weight gain influences the mother’s attitude. The questionnaire was administered to
1,000 mothers of live birth, singletons. Women with low prepregnancy BMI (< 19.6) had
a high attitude score and women with high prepregnancy BMI (> 26.6) had poor attitudes
toward weight gain during pregnancy. However, the correlation between attitude score
and weight gain did not reach statistical significance. Attitude was not a significant
predictor of birthweight by regression analysis. In this high-risk, low-income population,
attitude serves as a poor predictor of weight gain and birthweight. The researchers found
that other variables including, body size, tobacco use, and race/ ethnicity were significant
predictors of weight gain.

Siega-Riz and Hobel (1997) examined the psychosocial effects related to weight
gain during pregnancy in a Hispanic population. Psychosocial factors were not
significant in a regression model, even though some were correlated with weight gain.
For example, a family member’s death was positively associated with weight gain while
depression during pregnancy and physical abuse by the baby’s father were negatively
associated with weight gain. The results of this study emphasize the myriad of factors
influencing weight gain, and the duty of the clinician to look at all aspects of a woman’s

health.

23

A mail survey was administered to 500,000 postpartum women in order to assess
the prenatal advice given regarding weight gain during pregnancy and the actual amount
weight gained. 27% of women reported receiving no advice about weight gain. Reported
advice given was strongly associated with actual (measured) weight gain (Cogswell et
al., 1999). The results of this study suggest the opportunity to influence gestational

weight gain and hence, birth outcomes through appropriate interventions.

2.5.4 Effects of weight gain on weight retention

Weight retention in the postpartum period is part of the concern related to the
increase in IOM standards for weight gain. Analysis of the 1988 National Maternal and
Infant Health Survey found that 56% of white women and 37% of black women were less
than four pounds over their prepregnancy weight at 10-18 months postpartum. However,
25% of white women and 45% of black women retained more than nine pounds over
their prepregnancy weight. The median difference between prepregnancy weight and

postpartum weight was 2.6 pounds (Keppel and Tafl‘el, 1993).

2.6 Iron Status

Iron deficiency is the most common known form of mineral deficiency. Its
prevalence is highest among young children and women of childbearing age. Iron is
present in all cells in the body and firnctions to: I) carry oxygen from the lungs to tissues
in the form of hemoglobin, 2) facilitate muscle use and storage in muscles is in the form
of myoglobin, 3) serves as a transport medium for electrons within cells in the form of

cytochromes, and 4) is an important part of numerous enzyme reactions. Iron deficiency

24

can be defined over a continuum. First, iron depletion occurs and this causes no
physiological impairments. In this stage, the amount of stored iron is reduced, however,
the amount of functional iron may not be afl’ected. Iron deficient erythropoeisis is the
second stage where stored iron is depleted and transport iron is reduced. Red blood cell
products are limited and erythrocyte protoporphyrin concentration is increased. The final
stage is iron deficiency anemia. There is an underproduction of iron-containing
compounds, like hemoglobin. The red blood cells of individuals with iron deficiency

anemia are microcytic and hypochromic (MMWR, 4/3/98).

2.6.1 Measurements of iron

Hematological tests such as, hemoglobin, hematocrit, mean corpuscular volume
and red blood cell distribution width, describe the characteristics of red blood cells.
Biochemical tests such as erythrocyte protoporphyrin concentration, serum ferritin
concentration, and transferrin saturation, detect earlier changes in iron status (MMWR,
4/3/98). Diagnostic tests related to iron status occur in a two-step process. First, assess
iron status by prevalence of anemia based on hemoglobin and hematocrit. Second,
perform additional tests to establish cause of anemia (Scholl and Heidiger, 1994).

Hemoglobin and hematocrit reflect the amount of functional iron in the body and
serve as late indicators of iron status. Hemoglobin is a more direct and sensitive measure
than hematocrit. Hematocrit indicates the proportion of whole blood occupied by the red
blood cell and falls only after the hemoglobin concentration declines. The cut-off values
for anemia differ between children, men, non-pregnant women and pregnant women and

by trimester of pregnancy. The cut-off values also differ by altitude, cigarette use and

25

race. For example, living at high altitudes and/or smoking can cause an upward shift in
hemoglobin and hematocrit values WWR, 4/3/98). There are small diurnal variations
in hemoglobin and hematocrit, however, neither are biologically or statistically
significant (Borel et al., 1991).

There is a direct relationship between serum ferritin concentration and the amount
of stored iron. It can serve as an early indicator of iron stores and is the. most specific
indicator available of depleted iron stores (MMWR, 4/3/98). Ferritin is considered the
"gold standard" for diagnosis of iron deficiency anemia in pregnancy. Serum ferritin
measurements are often used to confirm the cause of anemia. Anemia (low hemoglobin
or hematocrit) accompanied by low serum ferritin (less than 12 mcg/dl) reflects iron
deficiency anemia (Institute of Medicine, 1990; Puolakka et al., 1980). A relatively new
test based on transferrin receptors may be a better index of iron deficiency anemia during
pregnancy when the prevalence of iron deficiency anemia in the population is low (<
10%) (Carriaga et al., 1991).

Mean corpuscular volume is the average volume of red blood cells. Low mean
corpuscular volume indicates microcytic anemia. Erythrocyte protoporphyrin is the
immediate precursor of hemoglobin. Infection, inflammation and lead poisoning can
elevate the erythrocyte protoporphyrin concentration. Transferrin saturation measures the

extent to which transferrin have vacant binding sites (MMWR, 4/3/98).

2. 6. 2 Iron status during pregnancy

Iron absorption is increased during pregnancy in order to support the growth of

the fetus. Therefore, iron requirements are increased during pregnancy. Iron deficiency

26

anemia during pregnancy is associated with increased prevalence of preterm delivery and
LBW in some studies (Scholl et al., 1992; Scholl and Heidiger, 1994; Klebanofl’ et al.,
1991). Other birth outcomes, such as Apgar score are also negatively affected by anemia
(Rusia et al., 1995). An observational study found that iron deficiency was associated
with increased risk of complications during pregnancy, such as urinary tract infections,
pyelonephritis and preeclampsia (Kitay and Harbort, 1975).

Other studies show iron deficiency is not related to birth outcomes. Lu et al.
(1991) found that low hematocrit was not statistically associated with increased risk of
preterm birth (adjusted OR=1.0-1.25) after confounding variables were controlled for.
Meanwhile, other studies found low hematocrit in the third trimester of pregnancy may
have a protective effect on birthweight. The Collaborative Perinatal Project, a study of
50,000 consecutive pregnancies, found that low hematocrit (<29%) and high hematocrit
(>3 9%) during pregnancy was associated with fetal death, preterm delivery and LBW
(Higgins et al., 1982).

Scholl and Heidiger (1994) recommend the early diagnosis and treatment of
anemia, either prior to pregnancy or during the first trimester. They believe the factors
increasing risk of preterm delivery with iron deficiency anemia may be set in place before

the anemia is diagnosed and treated.

2.6.3 Iron status in infants and children
Children nine to eighteen months of age are at the highest risk of any age-group
because iron stores of full-term infants are sufiicient to meet iron requirements until the

age of four to six months, earlier for preterm infants (Yip et al., 1987). This is a time of

27

rapid growth and therefore, iron requirements are higher. Early introduction of cow's
milk, a poor source of iron, is also a contributing factor (Pizarro et al., 1991; Ozki, 1993;
Boutry and Needlman, 1996). After two years of age, the growth rate slows and the diet
is more diverse, thus the risk of iron deficiency declines (Dallman et al., 1996; Looker et
al., 1997).

Poor iron status has been associated with decreased mental, physical and
behavioral performance in children, as well as interference with the immune system
(Dallman, 1987; Chandra and Saraya, 1975). Lozofl‘ et al., (1991) examined the long-
term effects of iron deficiency in infants on clinical, nutritional and psycho-educational
assessments at five years of age. Infants with moderate iron deficiency performed lower
in areas of mental and motor development than infants of normal iron status. Five-year
olds who were moderately iron deficient as infants but received iron therapy treatment,
received lower scores on tests of mental and motor development than other five-year olds
who were not iron deficient. The results of this study suggest that iron therapy may not
overcome the lasting effects of iron deficiency during infancy.

In a similar study, Walter et al., (1989) examined the effects of iron deficiency
anemia on infant psychomotor development through a double-blind, placebo-control,
prospective, cohort study. The results were the same, iron deficiency anemia had an
adverse effect on psychomotor development, and this effect appeared to be dose-
dependent in proportion to the severity and duration of anemia. The detrimental affects
persisted beyond three months after iron therapy treatment, suggesting long-term

consequences of iron deficiency on growth and development.

28

 

The WIC program has been attributed to the noted decline of anemia among US.
low-income children (Miller et al., 1985; Vasquez-Scone et al., 1985). Iron depletion
(<10 ng/ml serum ferritin) declined fi'om 20.5% to 0.7% in six month-old children and
from 43.4% to 1.6% in nine-month-old children (Miller et al., 1985). The iron status of
children during the pre-WIC era is from two samples of children, first those enrolled in a
longitudinal study and second, children attending the Children and Youth clinic from
1973 to 1974. Post-WIC data was from children enrolled in WIC program from birth in
1977.

PedNSS data from six states showed a continuous decline in anemia between
1975 and 1985 in all race, age, and SES groups. The authors explained this decline by
the provision or iron-fortified formula and iron-fortified cereal to WIC participants. The
authors also suggest that a decline in iron deficiency has additional benefits, such as
decrease in other impairments related to iron, including developmental or behavioral
problems and increased susceptibility to lead absorption and toxicity (Yip et al., 1987).

The benefits of the US. WIC program have been demonstrated in studies
comparing low-income Canadian children with US. WIC participants. Canada does not
have a food supplementation program like WIC. The prevalence of anemia among low-
income children in Canada is nearly 25%, which is similar to the US. prevalence in the
19703. Multiple regression showed the two most important predictors of iron deficiency
anemia among low-income one year olds in Canada were the introduction of whole cow’s
milk before the age of six months (OR= 3.56, 95% confidence interval 107-1126) and
the use of iron-fortified cereal for less than six months (OR= 3.15, 95% confidence

interval 1.25-7.96) (Lehman et al., 1992).

29

Primary prevention should focus on ensuring adequate iron intake. This is
imperative for children less than two years of age because they are at highest risk for iron
deficiency because of inadequate intake. Secondary prevention should focus on

screening, diagnosing and treating iron deficiency (MMWR, 4/3/98).

2.7 Prenatal Care

2.7.1 Definition and measurement of prenatal care utilization

At the turn of the century, prenatal care was focused on preventing maternal
mortality and morbidity and fetal death (Alexander and Howell, 1997). Prenatal care is
not easily measured because it is not a single intervention. Rather it consists of a series of
assessments and treatments. Prenatal care can be measured qualitatively and
quantitatively. Measuring the quantity of visits can be misleading because the number of
visits can be determined the by the timing of entry, the fi'equency of visits recommended
by the medical care provider, the woman’s compliance, presence of complications and
the gestational age at time of delivery (F iscella, 1995; Misra and Guyer, 1998).

The quality of prenatal care varies and the types of services rendered also cover a
range. Homan and Korenbrot (1998) examined the components of prenatal care and their
impact on birth outcomes. The Medicaid program has adapted its prenatal care services
to focus on the needs of low-income women. Support services provided at clinics
include: nutrition, psychosocial and health promotion assessments and counseling, along

with referrals to other resources. As with all programs, it is important to tailor the

30

intervention to the targeted audience and monitor the quality and effectiveness of the
program.

Researchers have developed indices in order to assess prenatal care utilization.
The Institute of Medicine Index (Kessner Index) is used to assess the adequacy of
prenatal care based on number of visits, timing of first prenatal care visit and the length
of gestation. Care is classified as adequate, intermediate or inadequate and does not
measure the quality of care or consider relative risk of the mother. The Kessner index
has been criticized as “painting an incomplete or inaccurate picture of prenatal care
utilization” especially for women who had a pregnancy of more than 36 weeks gestation
because women to have “adequate” care must receive only nine prenatal care visits, while
other indices require a greater number of visits for care to be considered adequate (Kogan
et al., 1998). Two new indices of prenatal care utilization have been developed: R-
GINDEX proposed by Alexander et al. (1987) and the Adequacy of Prenatal Care
Utilization (APNCU) Index proposed by Kotelchuck (1994) both of which incorporate an
“intensive use” component which is for women who have extremely large numbers of
prenatal care visits.

The most frequent birth outcomes measured as related to prenatal care are preterm
delivery and LBW. It is important to recognize that prenatal care can serve as a proxy
measure for other predictors of poor birth outcomes. Other benefits of prenatal care are
more difficult to measure, including, maternal self-esteem, nesting behaviors, improved
attachment between parent and child, identification of parents with problems and

interventions to help them develop new behaviors, connection to the health care system,

31

q‘ anal-hu- f - Efrain Ff .‘

connection to social services, such as housing and drug treatment, maternal physical and
mental health and utilization of family planning and breastfeeding (Rosenberg, 1998).
Summers and Price (1993) identified the importance of preconception care.
Although this may be difficult, due to the fact that up to 50% of pregnancies in the US.
are unplanned, it is important that health professionals are aware of its benefits. The
Public Health Service convened an expert panel to address the components of
preconception care, including appropriate and ongoing risk assessment, health promotion
and medical and psychological interventions. The group identified some specific topics
to address with potential parents during preconception care: helping women evaluate
their psychological and physical readiness, examine of concerns of the father, evaluate
need for genetic counseling, create a positive environment for conception, discontinue
family planning methods and timing conception, and choose a healthcare provider and

birth place.

2.7.2 Benefits of prenatal care on birth outcomes

Although there have been noted trends in the increase of earlier and more
intensive utilization of prenatal care during the last fifteen years, research has not shown
consistency among the expected improvements in birth outcomes. Misra and Guyer
(1998) identified two issues that give insight as to why preterm delivery and LBW have
not declined. First, does the increase in the utilization of prenatal care include the women
who are most at-risk for poor birth outcomes? Second, should early and more intensive

utilization of prenatal care alone be expected to make a difi‘erence in the birth outcomes?

32

{br- m' ain‘t-AI . i 1:...“

F iscella (1995) performed a critical review of the benefits of prenatal care on birth
outcomes using specific criteria. Causal evidence guidelines incorporate four criteria:
temporal relationship, biological plausibility, consistency, and alternative explanations.
Temporal relationship states that the intervention must precede the outcome to be
considered effective. In this case, prenatal care precedes the birth. Second, biological
plausibility refers to the existence of a biologically plausible mechanism that explains the
effect of the intervention. Fiscella (1995) points out that through the modification of risk
factors improvements in birth outcomes are seen. However, the magnitude of this affect
is limited by the rate of modifiable risk factors in a population and by the existence of
effective interventions. Fiscella (1995) found consistency in the Medline search
performed. Eleven out of 14 reports analyzed reported improved birth outcomes.
Finally, alternative explanations dealt with the adequacy of control for confounding
factors and selection bias. Confounding refers to a spurious relationship between and
intervention (e. g. prenatal care) and outcome (e. g. LBW or preterm delivery) created by
an extrinsic factor that is associated with both intervention and outcome. Selection bias
“occurs when comparisons are made between groups of patients that difl’er with respect
to determinants of the outcome other than those under study (page 472).” The author
notes that this is a problem for prenatal care studies because prenatal care is both an
intervention and an indicator of maternal behavior.

It is also important to consider the benefits of prenatal care on maternal health.
Prenatal care should be considered as a vital link for continuous health care for women
(Fiscella, 1995). Prenatal care provides the opportunity to begin intervention and counsel

women on factors that contribute to poor birth outcomes, including smoking cessation,

33

nutrition, and physical and psychological stress. Research has also shown the long-term
benefits of prenatal care utilization, such as immunization compliance and well-baby
clinic attendance. Other noted improvements in health related to prenatal care use
include increased breastfeeding, and decreased morbidity and mortality related to specific
maternal complications (Alexander and Howell, 1997).

Kotelchuck (1994) found a U-shaped relationship between the recommended

visits received and the rate of LBW in a nationally representative sample.

2.7.3 Prenatal care among low-income women

Kogan et al., (1998) examined all live births between 1981 and 1995 in the US.
Analysis showed the proportion of women beginning prenatal care early in pregnancy
increased from 1981 to 1995. The low-income population is a special subgroup, which
requires further examination of the barriers and benefits of prenatal care utilization. A
case-control study of 252 inner-city, pregnant women examined the effects of inadequate
prenatal care (defined by Kessner Index) compared with intermediate or adequate
prenatal care (defined by Kessner Index). Women with inadequate care identified the
following barriers to early prenatal care: lack of transportation or childcare,
homelessness, no insurance, and fear or shame related to the pregnancy or seeing the
physician. Women with inadequate care had a significantly higher percentage of smokers
and the mean birthweight was 405 g lower when compared to women with intermediate
or adequate care. The results of this study provide areas for future program development

to decrease barriers and improve care (Melnikow and Alemagno, 1993).

34

One in five women do not receive prenatal care during the first trimester of
pregnancy. Lewis et al. (1996) measured trends in the timing of prenatal care through
birth certificates and the 1988 National Maternal and Infant Survey. In 1994, 80% of
women in the US. received prenatal care in the first trimester. 90% of Cuban mothers
and 89% of Japanese mothers received early prenatal care (Healthy People 2000 goal:
90%). The lowest percentages were among blacks (68%), Puerto Ricans (67%) and
American Indians (65%). The authors identified financial, scheduling and transportation
problems as barriers to early prenatal care, along with mistimed or unwanted pregnancy.

There are disparities among race/ethnicity groups in utilizing prenatal care.
Hispanic women are less likely to seek prenatal care than white women. Byrd et a1.
(1996) examined 300 postpartum, low-income, Hispanic women to determine the reasons
why this race/ ethnicity group is not receiving adequate prenatal care. The researchers
found that seeking prenatal care in the first trimester was associated (p< 0.05) with being
older (>24 years), having insurance, and having a planned pregnancy. Barriers were also
identified in this study, including the most common complaint of “you have to wait too
long at the clinic.” Other barriers related to late entry to prenatal care in this population
were embarrassing examination, little contact with the physician, and lack of childcare
for other children. Similar racial disparities were found in the analysis of 1980 National
Natality Survey. Using the APNCU Index, black women had significantly poorer overall
utilization of prenatal care and nearly twice as many black women (21.1%) delayed
initiation of prenatal care to after the fourth month of pregnancy than white women

(10. 5%) (Kotelchuck, 1994).

35

Handler and Rosenberg (1992) examined the site of prenatal care and preterm low
birthweight (<37 weeks gestational age and birthweight < 2500 g) among low-income,
inner city women. The results showed that women who received prenatal care fiom the
public health department were less likely to have preterm low birthweight infants than
women who received prenatal care by private physicians. This study underscores the
importance of local health departments.

Several researchers examined the benefits of Medicaid expansion on birth
outcomes, prenatal care utilization, and cost-benefits analysis. The Medicaid expansion
extended coverage to one million women who were previously ineligible. Ray et al.,
(1997) examined the effect of Medicaid expansion in Tennessee and found that the
expansions resulted in significant increase in maternal Medicaid enrollment. In addition,
there was a significant decrease in inadequate use of prenatal care, along with increase in
first trimester enrollment in Medicaid. However, no improvement was seen in the
preterm delivery rate and rate of VLBW. Long and Marquis (1998) performed a study
examining the effects of Medicaid expansion in Florida. This study yielded similar
results as the Tennessee study: increased enrollment, improved access to prenatal care,
fewer women delaying entry to prenatal care, and increased number of visits during
gestation. However, this group found a significant reduction in the incidence of LBW
when comparing pre- and post- Medicaid expansion. Baldwin et al. (1998) observed the
same reduction in rate of LBW in Missouri. A cost benefit analysis of Missouri’s
Medicaid program found 89-g and 47-g increase in birthweight associated with adequate

and intermediate, respectively, prenatal care compared to inadequate (Schrarnm, 1992).

36

A meta-analysis estimated the savings of prenatal WIC services through reduction
of the first-year medical costs for their infants. WIC participation is significantly related
to decreased odds of LBW. On average, women with WIC had 25% fewer LBW than
demographically similar women without WIC. The researchers determined that for every
federal dollar spent on WIC, there is a total savings of $3.07 in Medicaid costs. The
major limitation of this research is that it was not determined which elements of WIC
contribute to the improvement in birthweight (Avruch and Cackley, 1995). A similar
study examined the Missouri Medicaid program to evaluate the expense of prenatal care
compared to Medicaid newborn and postpartum maternal care costs. Results of 12,023

births found that for every $1 spent on prenatal care $1.49 was saved (Schramm, 1992).

2.8 Tobacco Use During Pregnancy

2.8.1 Introduction

Smoking during pregnancy is associated with poor birth outcomes, such as LBW, SGA,
and preterm birth (Cnattingus and Haglund, 1997; Hellerstedt et al., 1997; Secker-Walker
et al., 1997; ‘Horta et al., 1997; Windham et al., 1999; Zaren et al., 1996; Jedrychowski et
al., 1998; Wen et al., 1990; Nordentoft et al., 1996; Wang et al., 1997) (Tables 1 and 2).
The body of literature is consistent on the deleterious effects of tobacco use during
pregnancy and some researchers have suggested that reduction in the number of
cigarettes and/ or smoking cessation during pregnancy may reduce these negative effects
(Secker-Walker et al., 1998; Mainous and Hueston, 1994). The long-lasting effects of

maternal smoking on an infant and child are less well documented, but worthy of

37

 

discussion (”Horta et al., 1997; Fried et al., 1999; Weitzman et al., 1992; Day et al.,
1994). Much of the research related to smoking during pregnancy and adverse birth
outcomes, including the present study rely on self-reported tobacco use. Therefore, it is

important to validate self-reported smoking with biomarkers.

2.8.2 Confounding factors associated with smoking

Confounding factors must be accounted for when examining the relationship between
birth outcomes and maternal smoking. A retrospective study of 1,343 pregnant women
analyzed the effects of maternal smoking and prepregnancy obesity simultaneously, and
their contributions to decreased birthweight. Prepregnancy obesity increased the risk of

Table 1. Odds ratios associated with the risk of smoking during pregnancy for certain
birth outcomes.

 

 

Subjects Birth Odds Ratio Author
Outcome
1,048,139 live births SGA ‘ 2.1 for Cnattingus and Haglund, 1997
moderate b
2.9 for heaving
LBW d 1.7 for
moderate b
2.2 for heavy °
5,166 live births LBW d 1.59 'Horta et al., 1997
IUGR 2.07

 

" < -2 standard deviations below the mean birthweight for gestational age
b 1-9 cigarettes per day

° >10 cigarettes per day

" <2500 g

° birthweight less than 10th percentile for gestational age and sex

38

 

Table 2. Birthweight associated with smoking during pregnancy.

 

 

Subjects (live births, n) Birthweight change ‘ Author

5,166 -142 g aHorta et al., 1997
3,083 -226 g Ellard et al., 1996
15,539 434g“, -301g" Wen et al., 1990

 

‘ mothers <17 years old
b mother >36 years old

LBW in smoking mothers, OR=5. 1. However, the researchers determined that smoking
was not associated with mean gestational age in either obese or normal weight mothers
(Hellerstedt et al., 1997). Another study by Ellard et al., (1996) noted that weight gain
during pregnancy was significantly less in mothers who smoked during pregnancy,
compared to those who did not smoke during pregnancy. Peacock et al., (1991)
examined the confounding affects of alcohol and cafi’eine consumption on the

relationship between maternal smoking and LBW. Alcohol alone (after adjustment for
caffeine and tobacco) reduced birthweight up to 8%, while caffeine alone (after
adjustment for alcohol and tobacco) reduced birthweight up to 6.5%. However, in women
who were heavy smokers (2 13 cigarettes per day) and had high caffeine intake (2 2801

mg per week) showed reductions in birthweight up to 18%.

2.8.3 Smoking cessation

Smoking cessation or reduction in the number of cigarettes smoked during
pregnancy may have positive effects on the birthweight of the offspring (Seeker-Walker
et al., 1998; Mainous and Hueston, 1994). The 1988 National Health Interview Survey
examined 4,876 women and their infants. They found that quitting smoking during the

first trimester of pregnancy reduced the proportion of LBW and preterm birth,

39

 

significantly (Mainous and Hueston, 1994). Hueston et al., (1994) examined the
relationship of smoking and LBW using the 1988 National Health Interview Survey.
They looked at the reduction and cessation of smoking during pregnancy from an
economic perspective and performed a cost-benefit analysis. They determined the cost
related to LBW compared to NEW infants. They calculated the cost of a cost-effective
program based on the success rate of smoking cessation fi'om a smoking cessation
program. They found that the program needed to cost $80 or less in order to be cost-
effective. The CDC estimated the costs of medical care that are attributable to smoking

during pregnancy to be $50 billion in 1993 (MMWR, 1 1/ 17/97).

2.8.4 Long-lasting efi‘ects of smoking

Few research papers have been published in regards to the long-lasting effects of
maternal smoking during pregnancy on their ofl‘spring. Some researchers have suggested
that maternal smoking is related reduced duration of breastfeeding (bHorta et al., 1997;
Fried et al., 1999). However, there is no evidence to suggest that maternal smoking
during pregnancy has an effect on behavioral problems in children at age six years

(Weitzman et al., 1992; Fried et al., 1999; Day et al., 1994).

2.8.5 Validity of self-reported tobacco use

Researchers have investigated the validity of self-reported smoking by through
the use of biomarkers, such as cotinine in the urine (Wang et al., 1997), serum
(Jedrychowski et al., 1998), or saliva (Parazzini et al., 1996), and exhaled carbon

monoxide (Secker-Walker et al., 1997). Cotinine is a metabolite of nicotine. By

40

measuring its levels one can allow a measurement of tobacco exposure. Parazzini et al.,
(1996) found that self-reported smoking during pregnancy had satisfactory validity when

confirmed with saliva cotinine levels in 109 women.

2.9 Alcohol consumption during pregnancy

2.9.1 Introduction

Alcohol consumption during pregnancy appears to have mixed effects on birth
outcomes depending on the quantity of alcohol consumed and the timing of alcohol
consumption during pregnancy. There is a clear relationship between heavy alcohol
consumption and fetal alcohol syndrome (Sokol and Clarren, 1989). Alcoholic beverage
containers are required by law to have a health-warning label with the message that

pregnant women should not drink alcohol (Kaskutas, 1995).

2.9.2 Neonatal mortality and alcohol consumption

An evaluation of Vital Statistics data, 1988-1990, found that neonatal mortality
was 77% higher when the mother consumed alcohol during pregnancy (Hoyert, 1996)
The 1988 National Maternal and Infant Health Survey evaluated the relationship of
alcohol consumption during pregnancy and poor birth outcomes in 9,953 live births,
3,309 fetal deaths, and 5,332 infant deaths through multivariate logistic regression.
Alcohol had a significant relationship with LBW, fetal death and infant death (F aden et

al., 1997).

41

 

 

2.9.3 Quantity of alcohol consumed

Abel and Hannigan (1995) reported a 'J-shaped’ relationship between maternal
alcohol consumption during pregnancy and the birthweight of the infant. A significant
increase in birthweight was seen with low levels of alcohol consumption. Meanwhile,
non-smokers did not show a decrease in birthweight attributable to alcohol. Thus,
smoking during pregnancy was found to be three-times more potent in decreasing
birthweight than drinking. A relationship between heavy drinking (40 cl wine per day)
was associated with poor birth outcomes in a prospective study of 9,000 women. Poor
birth outcomes include elevated stillbirths, lower mean birthweight, increased risk of
small-for-dates infant, and decreased placental weight (Karninski et al., 1978). A
prospective evaluation of 2,714 live births examined the effects of low to moderate
alcohol consumption on IUGR, preterm delivery, and LBW. Mild drinking (0.10 - 0.25
ounces of absolute alcohol per day) during the first month of pregnancy had a protective
efl‘ect on IUGR However, drinking during the seventh month of pregnancy had similar
odds ratios for light and moderate drinking, 2.88 and 2.96, respectively (Lundsberg et al.,
1997)

, A similar relationship was seen between quantity of alcohol consumed and poor
birth outcomes by Marbury et al., (1983). An analysis of 12,440 pregnant women found
that alcohol consumption of more than 14 drinks per week was significantly associated
with LBW, preterm birth, stillbirth and placenta abruptio. Alcohol intake of less than 14
drinks per week was not associated with the poor birth outcomes listed above, except

placenta abruptio.

42

An odds ratio of 0.35 was calculated for alcohol use during pregnancy and
preterm birth, indicating that alcohol consumption may actually lower risk for preterm
birth (Meis et al., 1998). Similar results of another prospective study of 1,513 pregnant
women found that smoking, alcohol and caffeine had no effect on gestational age of
infant (Peacock et al., 1995). Borges et al. (1993) found an odds ratio of 12.1 for women

with Alcohol Dependence Syndrome and risk of LBW and/or preterm delivery.

2.9.4 Timing of alcohol consumption

A 155-g reduction in birthweight was found in women who consumed two drinks
of alcohol per week during the first trimester of pregnancy. The effects of birthweight
reduction were less with alcohol consumption during the second and third trimesters
examined, 75-g and 57-g, respectively. Consumption of less than two alcohol drinks per
week showed no detrimental efl"ects (Shu et al., 1995).

Analysis of 3,447 pregnant women in the Netherlands found alcohol consumption
during the first and second trimesters was unrelated to birthweight after correcting for
gestational age. After examining a subgroup of women who smoked more than one pack
of cigarettes per day and drank less than 120 g of alcohol per week, a 7.2% decrease in
birthweight was seen(Verkerk et al., 1993). A cross-sectional study in New Zealand, a
relative risk of 0.4 was calculated, suggesting that alcohol intake during the third
trimester of pregnancy reduces the risk of preterm birth (Wright et al., 1998).

On the contrary, univariate and multivariate analyses found a decreased mean
birthweight associated with maternal drinking. An odds ratio of 2.35 was calculated for

drinking more than 20 g of absolute alcohol per day and preterm delivery. The National

43

 

Addiction survey calculated even higher odds ratios for alcohol consumption and poor

birth outcomes (Borges et al., 1993).

2.10 Fetal growth

During the first trimester, the fetus’ development is devoted to organogenesis and
little growth occurs. Maternal insults during this period are mainly teratogenic. The
second trimester of pregnancy is the period of most rapid fetal growth. During this
period, the fetal weight increases approximately 12-fold. Growth of the fetus continues
during the third trimester and the fetus’ fat mass is quadrupled (Strauss and Dietz, 1999).
It is not certain the precise gestational age of peak growth velocity. However, fetal
grth slows at 34-36 weeks gestation because of space constraints in the uterus (Falkner
et al., 1994).

Falkner et al. (1994) are developing fetal growth curves in order to diagnose SGA
and IUGR in utero. Sono gram measurements of femur length have good predictive
power for birth length. Abdominal thoracic diameter also shows good predictive power.
The abdominal thoracic diameter is influence by liver size and authors speculated a
relationship between liver size and function and size at birth.

Falkner et al., (1994) stated the most important phase of human growth is from
conception to term. In developed countries, two-thirds of LBW infants are preterm and
one-third are SGA. The opposite occurs in developing countries with more than 75% of
LBW attributed to SGA and only 25% attributed to preterm birth. The shift in
developing countries is due primarily to malnutrition and infection, both of which are

preventable.

44

Intrauterine growth retardation and/or shortened gestation are the primary causes
of LBW. Gestational age is the single most important factor that influences birthweight.
Analysis of US. Natality Data shows that preterm birth accounts for 65% of LBW
(Cogswell and Yip, 1995). Intrauterine growth retardation is associated with placental
dysfunction. The placenta has very complex metabolic and endocrine firnctions. There
are two main types of IUGR: 1) symmetrical IUGR- growth is proportional and the
infants are small; infants show slowed postnatal growth, 2) asymmetric IUGR- the length
and weight of the infant are low and the head circumference is normal; infant shows good

postnatal catch-up grth (Falkner et al., 1994).

2.11 Growth Retardation

2.11.1 Definitions and measurements

There are three general phases of human growth: infancy, childhood and puberty.
The infancy growth phase is largely nutrition-dependent. There is a rapid deceleration of
growth during this period compared to the rapid rate of growth in the fetus. The
childhood phase, with onset at 6-12 months of age, is growth hormone-dependent. A
slower phase of deceleration is noted during childhood. The puberty phase is driven by
sex hormones (Karlberg et al., 1994).

Anthropometric measurements in PedNSS are interpreted using the NCHS/CDC
reference growth charts. There were two original datasets used to compile this reference
curve. Investigators at the Fels Research Institute measured a sample of children from

Yellow Springs, Ohio. These children were used for the less than 24 months of age

45

 

 

portion of the growth curve. For children two years of age and older, a difl’erent
population was used to make the reference curve, NHANES-I (PedNSS User Manual,
1994). The Fels sample used to create the reference curve for children 0-24 months of age
is taller than the NHANES-I population. This causes an abrupt change in the prevalence
of stunting when children age to 2 years. For children less than two years, the prevalence
of stunting may be higher than reality, due to this inconsistency in the reference curve.
The prevalence of stunting varies worldwide from 2-5% in developed countries to
60-70% in developing countries. A study of 70 developing nations found that stunting
was highest in Asia and lowest in Africa (Frongillo et al., 1997). Analysis of PedNSS at
the national level shows the prevalence of stunting in the US. has remained stable from
1980-1991. The PedNSS is sensitive to detect improvements in the health and nutritional
status of the population it monitors. This was noted in the significant decline in the
prevalence of stunting among Asian children less than 2 years of age, 22% to 10% fi'om
1980 to 1991. Only minor variations in growth patterns amongst different race/ ethnicity
groups participating in PedNSS suggest that a single reference chair is valid for all (Yip

et al., 1992).

2.11.2 Determinants of grth retardation

Frongillo et al., (1997) proposed a conceptual model to investigate growth
retardation, which had three components. First, immediate causes of growth retardation
which included adequate dietary intake and health. Second, the underlying causes

affecting a child’s growth were food security, maternal and child care, health services and

46

the environment. Third, the basic causes which included formal and non-formal
institutions, political and ideological factors, economic structure and potential resources.

Allen (1994) proposed that early growth retardation is related to inadequate
maternal nutrition. The fetus is not given adequate nutrition during gestation, due to the
low maternal stores.

Rate of grth has several covariates, including SES, season, disease, feeding
patterns and nutritional status. It is believed that environmental factors such as maternal
illiteracy, poor hygiene and overcrowding, are more likely to cause stunting than ethnic
or genetic background (Karlberg et al., 1994). In addition to nutrition and genetics,
adverse social and family environments have been shown to retard children’s growth
(Skuse et al., 1994). Psychosocial adversity has long-term effects on growth into the
preschool years with poor cognitive and psychomotor development.

Data from the Nutrition Collaborative Support Program found specific nutrients
associated with growth retardation in children (Allen, 1994). The results of this study
found growth faltering most rapid in the first year of life. Dietary analysis showed the
linear growth was not affected by low energy intake and that severe grth retardation
remained even with adequate protein and essential amino acid intake (Allen, 1994). Zinc
appears to be important nutrient for linear growth. Zinc deficiency in children in the
Middle East is associated a high phytate, high fiber, low protein diet. Zinc
supplementation of 5mg of Zinc per day showed a 10% increase in height velocity within
six months of study initiation. Zinc supplementation studies in the US. and Canada have
shown improvements in height with no effect on weight gain among short, well-

nourished children (Allen, 1994). Iron studies have shown mixed results with respect to

47

linear growth. Allen (1994) speculates, based on review of literature that the linear
growth in children responds to iron supplementation if the child was initially anemic.
Other nutrients investigated were cooper, iodine and vitamin A.

It is possible that multiple micronutrient deficiencies are seen in stunted children
(Allen, 1994). This is related to the issue of diet quality, which is important for adequate
growth of children. Allen (1994) proposed that a diet poor in quality would contain few
animal products, hits and vegetables, and would consist of staples such as cereals,
legumes and other plants. This type of diet is low in vitamins and minerals.

Poverty appears to be a risk factor for stunting. Analysis of the National
Longitudinal Survey of Youth, poor children are 50% more likely to be stunted than
children in the highest income group. The prevalence of stunting decreased with
increasing income (Miller and Korenman, 1994). A study of low-income elementary-age
children in one community in eastern Kentucky found that 13% of girls and 4.8% of boys
were stunted. The author felt that despite the fact that 78% of children in the sample
were eligible for free school lunch, the social programs provided were not protecting the
children against stunting (Crooks, 1999).

A thorough analysis of the determinants of stunting at different age subgroups
was completed by Adair and Guilkey (1997). A cohort of 3,080 births was followed for
two years in Cebu, Philippines. The results found a strong association between poverty
and stunting. The weaning period (6-12 months of age) appeared to be a particularly
vulnerable time period for stunting. Stunting is highly persistent in this population, of the

children who were stunted at six months of age, 96% were stunted at two years of age.

48

0.1;. 0 MM. ”V;-
n

20% of stunting in children less than two years old participating in PedNSS was
attributed to birthweight. Significant differences in birthweight between stunted and non-

stunted children continued through the age of three years (Miller and Korenman, 1994).

2.11.3 Effects of growth retardation

Long-term consequences of stunting include cognitive and emotional
developmental delays (Miller and Korenman, 1994). Other long-term consequences
include short stature, reduce capacity to work and elevated risk of poor reproductive
outcomes. Immediate consequences of stunting include an increased susceptibility to
infections and increased risk of mortality (Adair and Guilkey, 1997).

The same cohort previously mentioned from Cebu, Philippines was studied to
determine the long-term effects of stunting. Stunting had persistent effects on the
cognitive abilities of children until age eight and eleven years. Decreased cognitive test

scores were noted in children who were stunted at two years of age.

2.12 Breastfeeding

2.12.1 Benefits of breastfeeding

Health care providers and educators know the health, nutritional and social
benefits of breastfeeding. Breastfed infants contract fewer infections than those who are
given formula (Lucas et al., 1990; Slade and Schwartz, 1987; Mestecky, 1991; Goldman,
1993). Mother’s milk actively helps newborns avoid diseases as it contains molecules

and cells that actively help infants stave ofl’ infection (i.e. antibodies, oligosaccharides,

49

ran—nuns.“ an;

and immune cells such as leukocytes) (Goldman, 1993). The added social benefits of
breastfeeding, such as building the mother-infant relationship are difficult to measure
quantitatively, but are a contributing factor to the overall health and well-being of an
infant. There are also economic benefits of breastfeeding, which have not been

adequately examined in the literature (Riordan, 1997; Montgomery and Splett, 1997;

Tuttle and Dewey, 1996).

2.12.2 Breastfeeding promotion

The need for effective breastfeeding promotion has been identified at least in the
past decade (Interagency Workshop, 1991). Despite the desire for intervention and
numerous attempts, there has been little success in developing a program that is both
cost-effective and produces the desired effects (i.e. increased breastfeeding initiation and
duration). Strategies have included: peer counseling (Long et al., 1995; Webb and
Ellerbee, 1996; Schafer et al., 1998), social support (Arlotti et al., 1998; Raj and Plitchta,
1998) and the use of different forms of media (Carroll, 1994; Kim, 1997). Different time
periods have also been targeted, such as prenatal (Reifsnider and Eckhart, 1997; Pugin et
al., 1996) and postnatal (Lowe, 1998).

The needs of mothers vary, depending on numerous factors (i.e. age, ethnicity,
education level, and social support network). For this reason, it is critical that the
intervention be tailored to the targeted population. Abramson (1992) discussed the
influence of culture on a mother’s attitude toward breastfeeding, the perceived barriers,

and methods to overcome barriers. A cultural assessment model was addressed and

50

stressed that healthcare professionals must be sensitive to the culture and background of

their clients.

2.12.3 Barriers

The Healthy People 2000 goal for was not met at the national level for
breastfeeding initiation nor duration, however there was movement toward the goal in the
right direction, from 52% in 1990 to 62% in 1997 (NCHS, 1999). Caldwell (1999) felt
that efi’orts to meet this goal were initiated too late. For example, the initiation of the
UNICEF’s Baby-Friendly Initiative was launched in 1990 to improve the in-hospital care
of mothers and infants through breastfeeding promotion, but the author feels that it may
take an additional decade before enough hospitals and birthing centers are involved in
order to make a difference in the data. The work protection legislation was not
introduced until 1998. Currently, there is no paid maternity leave guaranteed by

legislation. A national breastfeeding committee was developed in 1998.

51

 

 

Chapter 3 r

METHODS

3.1 Subjects

All subjects included in the study received WIC services from clinics in the State
of Michigan. Each subject provided data required by the program, which were compiled
by the State of Michigan for the PNSS and PedNSS in collaboration with the CDC. Data
collected from the pregnant/postpartum women included a self-reported health history
questionnaire, including dietary intake and health behaviors, and blood, height, and
weight measurements. Similar data were collected after delivery to enroll the infant. The
weight, height, iron status and feeding method are recorded for infants and children at

each visit to a WIC clinic.

3.2 Research design and sampling frame

The present study utilized both cross-sectional and longitudinal study designs for
four objectives. The cross-sectional study encompassed the evaluation of the health
status of mothers and children in reference to nine Healthy People 2000 Objectives in
maternal and child health. The longitudinal study included the identification of the
predictors of birthweight and grth retardation in infants and young children of the
mothers from the cross-sectional study.

Therefore, included in the present study are pregnant and postpartum women in
the PNSS 1995 and their infants and children from the PedNSS 1995,1996, 1997 and

1998 in the State of Michigan. The 1995 PNSS dataset includes information on all

52

mt'm
a - 1

women enrolled in the Michigan WIC program during the 1995 calendar year and their
infants born within the 1995 calendar year. When an infant is expected to be born, but is
not enrolled in WIC, the record includes only the woman’s record during pregnancy and
is submitted as a prenatal record. A complete record includes the woman’s prenatal
information and postpartum record linked with the infant WIC enrollment record. Raw
data from the Michigan WIC program are reported to CDC (Executive Summary
V Michigan WIC, 1995). There is only one record per mother-infant dyad regardless of the
number of visits the mother made to the WIC clinic during the course of her pregnancy.
The PedNSS contains information on participation of each child from birth to the child’s
fifth birthday. The state collected data are sent to CDC on a quarterly basis and dataset
errors are sent by CDC to state health departments for notification, firrther investigation
and correction (www.cdc.gov/nccdhphp/dnpa/PedNSS.htm). There are multiple records
for each child because the law requires the client to visit the WIC clinic at least
biannually to assess eligibility.

The record numbers of unique mother-infant dyads in the raw dataset in PNSS

1995 and PedNSS 1995-1998 are summarized in Table 3.

53

Table 3. The raw and final sample sizes of subjects included in the study.

 

 

 

Dataset Raw Sample Excluded Final Sample
11 n n % total
PNSS 1995 " 41,978 4,390 37,588 89.5

Visits (11) Children (11)

PedNSS 1995 b 649,613 18,519 631,094 300,703 97.1
PedNSS 1996 b 324,213 9147 315,066 214,521 97.2
PedNSS 1997 b 336,905 10,504 326,401 218,307 96.9
PedNSS 1998 b 329,964 26,026 303,938 214,464 92.1

 

‘ Pregnancy Nutrition Surveillance System
b Pediatric Nutrition Surveillance System

3.3 Exclusion criteria/ editing
PNSS 1995

A total of 4,3 90 cases (10.5% of raw dataset) were deleted from the original 1995
PNSS dataset (n=41,97 8) because the record contained no information on infants

(n=4,252) or had a critical error (n=138). Critical error are identified by CDC (Table 4).

Table 4. Variables deleted from records and exclusion criteria for PNS S" 1995.

 

 

Cases Deleted
Variable Exclusion Criteria n % of total
Mother's height (cm) s 101.5 or 2 213.0 627 1.5
Mother's weight (kg) 5 22.6 or 2 226.8 559 1.3
Mother’s BMI (kg/m2) s 9.9 or 2 75. 1,579 3.8
Birthweight (g) s 499 or 2 6000 425 1.0
Gestational age (days) 3 174 or 2 311 25 0.1

 

‘ Pregnancy Nutrition Surveillance System

54

o ain‘t]

 

PedNSS 1995-1998

The CDC cross-checked and edited the PedNSS datasets submitted to them by the
local and state health departments. The CDC is most concerned about the accuracy and
completeness of data to assess growth and nutritional status, such as height, weight,
hemoglobin, and hematocrit. Errors in the record were the result of faulty measuring
equipment, improper measuring technique, erroneous recording or keying of
measurements or improper sampling techniques for biological samples.

In the editing process, for example, the height recorded was compared to sex and
age of the child to determine if the values are consistent according to criteria established
by the CDC. The edit criterion of the CDC were based on standard deviation from the
mean value of the NCHS/CDC reference population for 1) height-for-age, 2) weight-for-
age, 3) weight-for-height, 4) hemoglobin-for-age, or 5) hematocrit-for-age. The standard
deviation for each of the five indices is calculated for each visit. Each of the five
measurements with a standard deviation > 3 .09 is flagged with error codes. The standard
deviation value > 3.09, established by CDC, indicated that the measure had 1 chance in
1,000 of being a real measurement. The error codes may mean normal (0), low value,
multiple indices checked (1), high value, multiple indices checked (2), low value, single
index checked (3), high value, single index checked (4) or unknown (9). These error
codes are regularly reported back to the state (PedNSS User Manual, 1994). Cases with
any of these error codes were deleted from further analysis in the present study (Table 5).

(Table 6shows demographic characteristics of women in the PNSS, 1995.)

55

 

Table 5. Percent of subjects excluded for each error code.

 

 

 

 

 

 

Dataset Error Code Cases Deleted
% of total
1995 PedNSS ‘ Total 2.9
Height 0.9
Weight 1.1
Age 0.6
Hemoglobin 0. 1
Hematocrit 0.2
1996 PedNSS ‘ Total 2.8
Height 0.9
Weight 1.1
Age 0.5
Hemoglobin 0. l
Hematocrit 0.2
1997 PedNSS " Total 3.1
Height 1.0
Weight 1.2
Age 0.6
Hemoglobin 0.1
Hematocrit 0.2
1998 PedNSS " Total 7.9
Height 1.1
Weight 1.3
Age 0.6
Hemoglobin 4.3
Hematocrit 0.6

 

" Pediatric Nutrition Surveillance System

56

Table 6. Characteristics of women participating in Michigan WIC at the

time of enrollment, PNSS ‘ 1995.

 

 

 

 

 

 

 

Characteristic n %
Race/Ethnicity
White 23,775 63.3
Black 11,148 29.7
Hispanic 2,073 5.5
Am. Indian 200 0.5
Asian 392 1.0
Age O’)
< 16 1,011 2.7
16-19 9,144 24.3
20-29 21,538 57.3
30-39 5,619 14.9
2 40 275 0.7
Marital Status
Married 11,581 30.8
Not married 23,241 61.8
Not known 2,766 7.4
Migrant Status
Migrant 475 1.3
Not migrant 33,223 88.4
Not known 3,890 10.3
Education (y)
< 9 4,418 11.8
9-11 10,728 28.6
12 15,484 41.3
13-15 5,842 15.6
2 16 1,022 2.7

 

‘ Pregnancy Nutrition Surveillance System

57

3.4 Creation of a longitudinal dataset: Merging the PNSS and PedNSS datasets.

A key phase of this research project was completed through systematic matching
of unique identification numbers for the mother and her offspring in the edited datasets.
The 1995, 1996, 1997 and 1998 PedNSS datasets contained the same unique infant
identification number as the 1995 PNSS dataset infant identification number that is
associated with the mother’s identification number. A longitudinal dataset was created
following these steps (Figure 1).

l) 1995 PNSS dataset was merged with each of the four PedNSS datasets separately
using SPSS. Infant identification number was used as the key variable for “add
variables” firnction in SPSS. Every visit a child made to a WIC clinic was
captured in the resulting four separate datasets of mother-infant dyads, i.e. 1995
PNSS + 1995 PedNSS; 1995 PNSS + 1996 PedNSS; 1995 PNSS + 1997 PedNSS;
and 1995 PNSS + 1998 PedNSS. Unmatched cases were deleted (n=331).

2) The four datasets of mother-infant dyads were then merged to a single data file of
all mother-infant dyads with the “add variables” firnction in SPSS. This single
dataset contained maternal information in 1995 PNSS merged with each visit of
her offspring in 1995, 1996, 1997 and 1998 PedNSS datasets. There was no
linkage between the four years of children’s data, e. g. the dataset did not contain
information on the child’s entire WIC enrollment period (prenatal record through
four years of age) on a single record.

3) The dataset was converted from SPSS to SAS, in order to array the data into a

longitudinal dataset.

58

4) A stepwise array of the merged dataset created in Step 2, allowed the data to be
constructed into a longitudinal dataset. Five age-group categories (0-5 months, 6-
11 months, 12- 23 months, 24-35 months, and 36-48 months) were devised and
data from the PedNSS datasets were incorporated into these age groups. If a child
had more than one visit during an age group category, the last visit was selected.
The final longitudinal dataset contained all information from the PNSS1995
dataset and growth information at the five age group categories from PedNSS
1995, PedNSS 1996, PedNSS 1997 and PedNSS 1998. A total of 37,257 mother-
infant/ child dyads are included in this dataset. (Table 7)

5) The final longitudinal dataset was converted back to SPSS fi'om SAS for ease of

statistical analysis.

Table 7. The sample sizes of subjects in the merged
longitudinal dataset, PNSS' 1995 and PedNSSb 1995-1998.

 

 

Age-group 11
0-5 months 36,265
6-11 months 23,030
12-23 months 17,196
24-35 months 6,968
36-48 months '

 

‘ Pregnancy Nutrition Surveillance System
b Pediatric Nutrition Surveillance System

59

Figure l. Merging the PNSS and PedNSS datasets to create a longitudinal dataset.

 

 

Step 1

 

 

 

 

PNSS 1995
n=3 7,588
mother-infant
dyads
(Prenatal
through birth
outcomes)

 

 

./ \\

 

 

Step
2

 

 

PedNSS 1995
n= 631, 094

 

 

PedNSS 1996
n= 315, 066

 

 

PedNSS 1997
n= 326, 401

 

 

PedNSS 1998
n= 303, 938

 

 

 

l

l

l

l

 

 

Linel :
Line2:
Line3 :
Line4:
Line5 :
Line6:
Line6:
Line7 :

Dataset Example:
MomA-KidA95Visit 1
MomA-KidA95Visit 2
MomA-KidA96
MomA-KidA97
MomA-KidA98
MomB-KidB96
MomB-KidB97Visit1
MomB-KidB97Visit2

 

 

Steps 3 & 4

 

 

l

 

 

Dataset Example:
Line]: MomA-KidA (O-Smos.) KidA (6-1 lmos.) KidA (12-23mos.) KidA
(24-35mos) KidA (36-48mos)
Line2: MomB-KidB (0-5mos.) KidB (6-11mos.) KidB (12-23mos.) KidB (24-
35mos) KidB (36 -48mos)

Line3: MomC-KidC (0-5mos.) KidC (6-11mos.) KidC (12-23mos.) KidC (24-
35mos) KidC (36 -48mos)

Line4: MomD-KidD (0-5mos.) KidD (6-11mos.) KidD (12-23mos.) KidD
(24-3 5mos) KidD (36-48mos)

 

 

 

Step 5

 

 

l

 

 

Statistical Analysis

 

 

60

 

 

 

3.5 Statistical Analysis

3.5.1 Statistical Software

Statistical Package for the Social Sciences (SPSS) Version 7.5 and Statistical
Application System (SAS) Version 6.12 were used. SPSS was used for statistical
analysis because it is user-fiiendly. SPSS was not capable of the stepwise array function

used in step four of the merging process, therefore SAS was required.

3.5.2 Objective 1: To assess the degrees of the Michigan WIC population achieving
Healthy People 2000 Objectives.

Nine Healthy People 2000 Objectives and their goals were used as a reference to
assess maternal child health indicators for the Michigan WIC program. The cross-
sectional dataset for mothers (PNSS 1995) was examined to assess maternal health status
and lifestyle factors, incidence of low birthweight, and initiation of breastfeeding rates.
PedNSS 1995, 1996, 1997 and 1998 were used separately as cross-sectional datasets to
assess growth retardation in children. Frequency distributions and chi-square tests were
performed to evaluate the significance of differences among sub-population groups for

the nine selected goals (Table 8).

Healthy People 2000 Objectives
1) To reduce the prevalence of prepregnancy overweight to s 25% among low-income

women aged 20 and older; 3 30% among black women aged 20 and older; 3 25% among

61

“ETC-WW

‘12;

Hispanic women aged 20 and older; 5 30% among American Indians and Alaskan
Natives. Overweight defined as BM] 2 27.3 for women 20 years and older.

2) To increase the proportion of mothers who achieve the minimum recommended weight
gain during their pregnancies to 285%.

3) To increase the abstinence from tobacco use by pregnant women to 2 90%.

4) To increase the abstinence from alcohol by pregnant women to 2 95%.

5) To increase to 2 90% the percent of pregnant women who receive prenatal care during
the first trimester of pregnancy.

6) To reduce the prevalence of iron deficiency to .<_ 4% among low-income women of
childbearing age. The reference to assess iron deficiency in the present study were
hemoglobin and hematocrit instead of erythrocyte protoporphyrin, serum ferritin and
transferrin saturation.

7) To reduce the prevalence of LBW (< 2500g) to s 5% of all live births and VLBW (<
1500g) to s 1% of all live births; s 9% and .<_ 2%, respectively, in the black population.

8) To increase breastfeeding initiation rate to 2 75%. Breastfeeding was defined by
exclusive use of human milk or the use of human milk with a supplemental bottle of

formula or cow’s milk. The CDC calculated the rate of breastfeeding initiation based on
infants born alive and alive at the time of the postpartum visit.

9) To reduce the prevalence of growth retardation to s 10% among low-income children
younger than five years of age, low-income black children younger than age 1, low-
income Hispanic children younger than age one year, low-income Hispanic children aged

one year, and low-income Asian and Pacific Islander children aged 2-4 years.

62

3.5.3 Objective 2: To determine whether maternal health status, lifestyle and
sociodemographic characteristics in Michigan WIC women predict the birthweight of
their offspring while accounting for infant characteristics such as gestational age and sex.

Birthweight, as a continuous variable, was predicted by maternal health status,
lifestyle and sociodemographic characteristics, while accounting for infant
characteristics, such as gestational age and sex. Specific maternal predictors were
identified by the following approaches:

1) Pearson’s correlation coefficients (r) were calculated between birthweight (grams)
and prepregnancy BMI, gestational age (days), alcohol use during pregnancy
(drinks per week in third trimester), education (number of grades completed),
previous number pregnancies, previous number of live births, household monthly
income ($), when prenatal care began (month of pregnancy), weight gain during
pregnancy (total pounds gained or lost), tobacco use during pregnancy (number of
cigarettes smoked during third trimester), mother’s age (years), household size
and number of infants resulting from the pregnancy. Spearman’s rank was
calculated between birthweight (grams) and maternal characteristics in nominal or
ordinal scales: migrant status (yes=1, no=0), marital status (yes=1, no=0), sex of
infant (boy=1, girl=0), breastfeeding initiation (yes=1, no=0), and race (white, not
Hispanic=1, all others=0), iron status during pregnancy (anemic=1, normal=0),
Medicaid participant during pregnancy (yes=1, no=0), Food Stamp participant

(yes=1, no=0), and AFDC participant (yes=1, no=0).

63

2)

3)

Stepwise forward multiple linear regression analysis for birthweight (dependent
variable) by the maternal predictors: maternal health status, lifestyle, and
behavioral and sociodemographic characteristics. Independent variables were
included in model if correlated with birthweight at p<0.05 level. Variables were
eliminated stepwise from model depending on probability of the F-value.

The following independent variables were included in the final regression model:
prepregnancy BMI, weight change during pregnancy, tobacco use during

pregnancy, monthly income, race/ ethnicity, and infant sex.

3.5.4 Objective 3: To identify factors differentiating low birthweight and normal

birthweight infants in the Michigan WIC population based on maternal health status,

lifestyle, and behavioral factors during pregnancy.

Maternal lifestyle, behavioral and sociodemographic characteristics that

differentiate NBW and LBW infants by were identified by discriminant analysis and

logistic regression.

1)

2)

The independent variables that were significantly correlated (p<0.05) with
birthweight (Objective 2, Step 1) were included in the stepwise forward
discriminant analysis.

The independent variables were stepwise eliminated based on Wilk’s lambda
values. Wilk’s Lambda is used in an ANOVA (F) test of mean differences in
discriminant analysis. The smaller the erk’s Lambda value, the more the
variable contributes to the discriminant function of the dependent variable,

birthweight. The Lambda values ranges fiom 0 to 1

64

3)

4)

5)

6)

(www2.chassncsu.edu/garson/pa765/discrim.htm). Birthweight was a categorical
variable (LBW=1, NBW=0).

The final discriminant analysis retained the following independent variables:
race/ethnicity (white, not Hispanic=1, all others=0), mother’s age (years),
prepregnancy BMI, marital status (yes=l, no=0), self-reported number of
previous pregnancies, weight gain during pregnancy (pounds), tobacco use during
pregnancy (cigarettes per day during third trimester) and infant sex (boy=1,
girl=0). These variables were then entered simultaneously into the discriminant
function.

\Vrlk’s lambda, standardized canonical discriminant function coefficients and
classification table were calculated to determine the importance of each predictor
and overall success of the discriminant firnction in correctly classifying cases.
Sensitivity, specificity, positive predictive value and negative predictive value
were calculated to determine how well the firnction properly classified cases.

The independent variables included in the initial discriminant analysis (Objective
3, Step 1) were incorporated in the logistic regression model. Stepwise forward
logistic regression was performed in order to achieve the best model for
predicting LBW and NBW. Wald statistic was used to stepwise eliminate the
independent variables from the initial model. The Wald statistic tests the null
hypothesis that the logit coefficient is associated with a given independent
variable (www2.chass.ncsu.edu/garson/pa765/logistic.htm).

The final model included the variables resulting from the stepwise logistic

regression: maternal age in years(< 16, 16-19, 2029 (reference), 30-39, 2 40),

65

prepregnancy BMI (s 17.9, l8-19.7, 19.8-26 (reference), 26-29. 1, > 29.1), weight
gain during pregnancy (less than IOM recommended amount, ideal (reference),
greater than IOM recommended amount), marital status (yes=1, no=0), infant sex
(boy=1, girl=0), number of infants resulting from the pregnancy (singleton
(reference, twins, triplets), tobacco use during pregnancy (cigarettes per day in
third trimester, 0 (reference), 1-9, 10-19, 2 20), maternal iron status (normal=l,
anemic=0), race/ ethnicity (white, not Hispanic=l, all others=0).

7) Odds ratios, 95% confidence intervals (95% CI), unstandardized regression
coefficients and R-values were calculated in order to assess the strength and
statistical significance of the relationships. Sensitivity, specificity, positive
predictive value, and negative predictive value were calculated to determine how
well the logistic regression equation classified cases.

Discriminant analysis and logistic regression use two difl'erent approaches to
examining the same outcome. Discriminant analysis is used to determine the
discriminating factors between two groups, LBW and NBW. Logistic regression is used
to predict a dichotomous outcome (LBW and NBW), by independent variables.

Both logistic regression and discriminant analysis include only cases with valid
values for every variable included in the model. For example, if a case had a missing .
value for one of the independent variables in the model, it would not be included in the
model. This results in different sample sizes for each model depending on the

independent variables included in the model.

66

 

rt

 

3.5.5 Objective 4: To identify the factors associated with growth retardation among
children in the Michigan WIC population at various age-group categories.

1) Correlation coefficients were calculated between infant height-for-age percentile
and maternal characteristics, birth outcome measurements and infant feeding
method. Pearson’s (r) correlation coeflicients was calculated for continuous
variables: height-for—age percentile at age subgroups (0-5 months, 6-11 months,
12- 23 months, 24-35 months, and 36-48 months), mother’s age (years), mother’s
level of education (years), gestational age (days), prepregnancy BMI, number of
previous pregnancies, number of previous live births, household income ($), when
prenatal care began (month of pregnancy), tobacco use during pregnancy
(cigarettes per day in third trimester), alcohol use during pregnancy (drinks per
week in third trimester), and birthweight (g).

Spearman’s rank was calculated for categorical and ordinal variables
including weight gain during pregnancy (below, ideal and above according to
IOM standards), maternal iron status during pregnancy (anemic=l, normal=0),
infant sequence number (1=singleton, 2=twins, 3= triplets, etc.), marital status
(la/es, 0=no), infant sex (boy=1, girl=0), breastfeeding initiation (initiated=1,
never breastfed=0), Food Stamp participant (yes=1, no=0), AFDC participant
(yes=1, no=0), Medicaid participant (yes=1, no=0), and race/ethnicity (white, not
Hispanic=1, all others=0).

2) The logistic regression model included the predictors that were correlated with

height-for-age at p <0.05 level.

67

3) Separate logistic regression models were developed to determine predictors of
stunting at five different age-group categories: 0-5, 6-11, 12-23, 24-35 and 36-48
months of age. Logistic regression was performed in a forward stepwise fashion
with variables eliminated on the basis of the Wald Statistic. Variables in the final
model were entered simultaneously. Stunting (dependent variable) was coded
stunted=1, not stunted =0.

4) Predictors included in the initial model for 0-5 months: maternal age (<16 years,
16-19 years, 20-29 years (reference), 30-39 years, 2 40 years), gestational age
(days), prepregnancy BM] (5 17.9, 18-19.7, 19.8-26 (reference), 26-29.1, >291),
weight gain during pregnancy (less than recommended, ideal (reference), greater
than recommended), maternal iron status during pregnancy (anemic=1,
normal=0), infant sequence number (singleton=1, twins=2, triplets=3), marital
status (yes=1, no=0), number of previous pregnancies, number of previous live
births, household monthly income ($), household size (number of people),
tobacco use during pregnancy (cigarettes per day in third trimester, 0 (reference),
1-9, 10-19, 2 20), alcohol use during pregnancy (drinks per week: 0 (reference),l-
2, 3-10, 2 1]), infant sex (boy=1, girl=0), infant birthweight (LBW=1, NBW=0),
breastfeeding initiation (yes=1, no=0), Food Stamp participant (yes=1, no=0),
Medicaid participant (yes=1, no=0), AFDC participant (yes=1, no=0) and race/
ethnicity (white, not I-Iispanic=1, all others=0).

The following variables were in the final logistic regression model for

predicting stunting at 0-5 months: prepregnancy BMI, tobacco use during

68

5)

6)

pregnancy, birthweight, weight gain during pregnancy, maternal iron status during
pregnancy, and number of infants resulting fi'om the pregnancy.

Predictors included in the initial model to predict stunting at 6-11 months were:
height-for-age percentile at 0-5 months (stunted=1, not stunted=0), weight gain
during pregnancy (less than recommended, ideal (reference), greater than
recommended), maternal iron status during pregnancy (anemic=1, normal=0),
infant sequence number (singleton=1, twins=2, triplets=3), marital status (yes=1,
no=0), number of previous pregnancies, number of previous live births, household
monthly income ($), household size (number of people), when prenatal care
began (month of pregnancy) and tobacco use during pregnancy (cigarettes per day
in third trimester, 0 (reference), 1-9, 10-19, 2 20), infant sex (boy=1, girl=0),
breastfeeding initiation, Food Stamp participant (yes=1, no=0), Medicaid
participant (yes=1, no=0), AFDC participant (yes=1, no=0), and race/ ethnicity
(white, not Hispanic=1, all others=0).

The following variables were included in the final logistic regression model
for predicting stunting at 6-11 months: height-for-age percentile at 0-5 months,
household size, tobacco use during pregnancy, birthweight, weight gain during
pregnancy, and infant sex.

Predictors included in the initial model to predict stunting at 12-23 months were:
height-for-age percentile at 0-5 months (stunted=1, not stunted=0), height-for-age
percentile at 6-11 months (stunted=1, not stunted=0), gestational age (days),
weight gain during pregnancy (less than recommended, ideal (reference), greater

than recommended), maternal iron status during pregnancy (anemic=1,

69

 

7)

normal=0), infant sequence number (singleton=1 (reference), twins=2, triplets=3),
number of previous pregnancies, number of previous live births, household size
(number of people), when prenatal care began (month of pregnancy), tobacco use
during pregnancy (cigarettes per day in third trimester, 0 (reference), 1-9, 10-19, 2
20), birthweight (LBW=1, NBW=0), infant sex (boy=1, girl=0), breastfeeding
initiation (yes=1, no=0), Food Stamps participant (yes=1, no=0), AFDC
participant (yes=1, no=0), Medicaid participant (yes=1, no=0), race/ ethnicity
(white, not Hispanic=1, all others=0) and household monthly income.

The following variables were included in the final logistic regression model
for predicting stunting at 12-23 months: height-for-age percentile at 0-5 months,
height-for-age percentile at 6-11 months, when prenatal care began, household
size, birthweight, and infant sex.

Potential predictors included in the initial model to predict stunting at 24-35
months were: height-for-age percentile at 0-5 months (stunted=1, normal=0),
height-for-age percentile at 6-11 months (stunted=1, normal=0), height-for-age
percentile at 12-23 months (stunted=1, normal=0), maternal age in years (<16, 16-
19, 20-29 (reference), 30-39, 2 40 years), prepregnancy BMI (s 17.9, 18-19.7,
19.8-26 (reference), 26-29. 1, >291), weight gain during pregnancy (less than
recommended, ideal (reference), greater than recommended), maternal iron status
during pregnancy (anemic=1, normal=0), infant sequence number (singleton=1
(reference), twins=2, triplets=3), number of previous pregnancies, number of
previous live births, household size (number of people), tobacco use during

pregnancy (cigarettes per day in third trimester, 0 (reference), 1-9, 10-19, 220),

70

8)

birthweight (LBW=1, NBW=0), breastfeeding initiation (yes=1, no=0), Food
Stamps participant (yes=1, no=0), AFDC Participant (yes=1, no=0), and Medicaid
participant (yes=1, no=0).

The following variables were included in the final logistic regression model
for predicting stunting at 24-35 months of age: height-for-age percentile at 6-11
months, height-for-age percentile at 12-23 months, weight gain during pregnancy,
maternal iron status during pregnancy. V
Potential predictors included in the initial model to predict stunting at 36-48
months were: height-for-age percentile at 0-5 months (stunted=1, normal=0),
height-for-age percentile at 6-11 months (stunted=1, normal=0), height-for-age
percentile at 12-23 months (stunted=1, normal=0), height-for-age percentile at 24-
35 months (stunted=1, normal=0), maternal age in years (<16, 16-19, 20-29
(reference), 30-39, 2 40 years), prepregnancy BM] (3 l7 .9, 18-19.7, 19.8-26
(reference), 26-29. 1, >291), weight gain during pregnancy (less than
recommended, ideal (reference), greater than recommended), maternal iron status
during pregnancy (anemic=1, normal=0), tobacco use during pregnancy
(cigarettes per day in third trimester, 0 (reference), 1-9, 10-19, 220), household
size (number of people), birthweight (LBW=1, NBW=0) and race/ethnicity
(white, not Hispanic=1, all others=0).

The variables included in the final logistic regression model for predicting
stunting at 36-48 months of age were: height-for-age percentile at 6-11 months,
height-for-age percentile at 12-23 months, and height-for-age percentile at 24-35

months.

71

9) Odds ratios, 95% CI, unstandardized regression coefficients and R- values were
calculated in order to assess the strength and statistical significance of the
relationships. Sensitivity, specificity, positive predictive value, and negative
predictive value were calculated to determine how well the logistic regression

equation classified cases.

72

 

Chapter 4

RESULTS

4.1 Objective 1: To assess the degrees of the Michigan WIC population achieving
Healthy People 2000 Objectives.

The Michigan WIC population did not meet in 1995 eight out of the nine Healthy
People 2000 Objectives assessed in this study (Table 8). Increasing self-reported
abstinence from alcohol during pregnancy to 2 5% (Table 13) was the only Healthy
People 2000 Objective goal that was achieved. Special population goal achieved:
incidence of VLBW among blacks (Table 14). The Michigan WIC program has been
making progress towards achievement of the goal to reduce growth retardation from 1995
to 1998 (Table 15).

1) Reduce the prevalence of prepregnancy overweight to s 5%: The overall
prevalence of overweight is 36.1% among women in the Michigan WIC program
in 1995. Black women have the highest prevalence of obesity (41.3%) and Asian
women have a low prevalence of obesity (9.5%), which exceeds the Healthy
People 2000 Objective goal (Table 10).

2) Increase the proportion of women who achieve the minimum recommended
weight gain during pregnancy to 2 85%: 62.8% of women participating in the
Michigan WIC program achieved minimum recommended amount of weight gain
during pregnancy specified by the IOM, this includes both ideal and above ideal
weight gain. 37.2% gained inadequate weight during. The large number of

women gaining less than ideal weight is of concern because of the poor birth

73

 

3)

4)

5)

6)

7)

outcomes associated with inadequate weight gain. Similarly, the gaining
excessive weight during pregnancy is associated with other pregnancy
complications and birth outcomes.

Reduce the prevalence of anemia s 4% among low-income women of
childbearing age: This goal was not achieved by any race/ ethnicity group, neither
during pregnancy or postpartum. The prevalence of anemia was significantly
higher during postpartum than during pregnancy. Black women had the highest
prevalence of anemia during pregnancy (23.9%) and postpartum (48.6%) (Table
10).

Increase the proportion of women who receive prenatal care during the third
trimester to 2 85%: The majority of women (68.3%) began receiving prenatal care
at the time of WIC enrollment in the first trimester. A large proportion of women
enrolled at WIC during postpartum and had no records for the timing of beginning
prenatal care during pregnancy: Hispanic and Asian women, 22.4% and 22.5%,
respectively (Table 11).

Increase abstinence from tobacco use 2 90%: The majority of women (70%) self-
reported not smoking during the third trimester of pregnancy. Very few women
(1.3%) smoked more than one pack of cigarettes per day during the third trimester
of pregnancy (Table 12).

Increase abstinence from alcohol use 2 95%: 97% of the women self-reported no
alcohol use during the third trimester of pregnancy (Table 13).

Reduce incidence of LBW s 5%: The incidence of LBW among infants born to

mother participating in the Michigan WIC program was 7.4% in 1995. The

74

3)

9)

incidence of LBW among infants born to black mothers was 10.1% exceeding the
special population goal of 9%. 1.0% of infants were VLBW, achieving the
special population objective (5 1.0%). The special population goal (3 2%) for
incidence of VLBW among black infants was met, with an incidence of 1.5% in
this population in 1995 (Table 14).

Increase breastfeeding initiation 2 75%: Only 41% of mothers participating in the
Michigan WIC program in 1995 reported breastfeeding their infant at least once.
Reduce prevalence of growth retardation s 10%: About 9.7% and 9.8% of
children less than five years old were growth stunted in 1997 and 1998,
respectively. Growth stunting is more prevalent among infants 0-5 months old
compared to other age groups (Table 15). American Indian children have the
lowest prevalence of growth retardation when compared to other race/ ethnicity

groups (Table 16).

75

 

Table 8. Public health indicators of Healthy People 2000 Objectives of Michigan WIC
population in 1995.

 

 

Healthy People 2000 Objective Healthy

People

2000
Goal MI 1995
Number Description % %
2.3 Prevalence of overweight ‘ s 25 36.1
14.6 Achievement of minimum recommended 2 85 62.8

weight gain during pregnancy b

3.7, 14.10 Tobacco abstinence during pregnancy 2 90 70.0
14.10 Alcohol abstinence during pregnancy 2 95 97.0
14.11 Early prenatal care ° 2 90 68.3
2.10 Maternal anemia - during pregnancy d s 4 13.8
2.10 - postpartum d s 4 39.4
14.5 Low birthweight (<2500g) s 5 7.4
2.11, 14.9 Breastfeeding initiation ‘ 2 75 41.0
2.4 Growth retardation in children 1‘ s 10 10.3

 

'BMI > 27 .3 for women aged 20 years and older.

b Institute of Medicine Standards: 28-40 pounds for underweight pre-pregnancy (BMI s
19.7); 25-35 pounds for normal weight pre-pregnancy (BMI 19.8-29.0); 15-25 pounds for
overweight pre-pregnancy (BMI 2 29.1).

c Prenatal care obtained, beginning the first trimester of pregnancy.

d Centers for Disease Control Standards for anemia: The first, second, and third trimester
for <11, <10.5, <11 g/100ml for hemoglobin or < 33%, < 32%, < 33% for hematocrit,
respectively.

°Postpartum anemia: hemoglobin < 12.0 g/100ml or hematocrit <3 6%.

f Infant breastfed at least once in early postpartum period.

3 < 5th percentile height-for-age of NCHS reference.

76

Table 9. Prevalence of prepregnancy overweight ’ in Michigan WIC
women aged 20 and older in 1995 in comparison with the Healthy
People 2000 Objective goal.

 

 

 

Overweight ‘ Goalb
Race/ Ethnicity All (n) n % %
Total 25,637 9,256 36.1 S 25
White 16,628 5,684 34.2
Black 7,286 3,007 41.3 S 30
Hispanic 1,275 488 38.3 s 25
American Indian 131 47 35.9 s 30
Asian 317 30 9.5

 

‘BMI > 27 .3 in women aged 20 years and older.
l’Healthy People 2000 Objective goal: to reduce the prevalence of

prepregnancy overweight to S 25% among low-income women aged 20
and older.

77

 

Table 10. Prevalence of anemia " in pregnant and postpartum
women in Michigan WIC program in 1995 vs. Healthy People
2000 Objective goal.

 

 

 

Race/ Ethnicity All (11) % GOALb
Prenatal

Total 29,822 13.8 s 4
White 19,378 9.5

Black 8,409 23.9

Hispanic 1,570 12.4

American Indian 158 11.4

Asian 307 13.7

p ° 0.000

Postpartum

Total 30,660 39.4 s 4
White 20,670 35.6

Black 7,719 48.6

Hispanic 1,778 44.0

American Indian 179 36.9

Asian 314 38.5

p ° 0.000

 

‘ Centers for Disease Control Standards for anemia: The first,
second, third trimester and postpartum <11, <10.5, <11, <12
g/100ml for hemoglobin and < 33%, < 32%, < 33%, 36% for
hematocrit, respectively.

b Healthy People 2000 Objective goal: to reduce the
prevalence of iron deficiency to s 4% among low-income
women of childbearing age.

° Statistical significance of prevalence of anemia among race/
ethnicity subgroups.

78

 

Table 11. Trimester of pregnancy when prenatal care began at the time of WIC

enrollment among Michigan WIC women in 1995 vs. Healthy People 2000 Objective

 

 

 

gal.
Trimester of Pregnancy
1st 2nd 3rd
Race/ Ethnicity Total % GOAL ’ % %
(II)

Total 37,526 68.3 2 90 13.2 1.4
White 23,737 70.3 10.9 1.1
Black 11,128 65.6 18.0 1.9
Hispanic 2,071 62.2 13.1 2.3
American Indian 199 67.3 10.6 3.5
Asian 391 58.8 16.9 1.8
p" 0.000

 

‘Healthy People 2000 Objective goal: to increase to 2 90% the percent of pregnant
women who receive prenatal care during the first trimester of pregnancy.

b Statistical significance of timing of entrance to prenatal care among race/ ethnicity
subgroups.

Table 12. Tobacco use during third trimester of pregnancy by
Michigan WIC women in 1995 vs. Healthy People 2000 Objective.

 

 

Cigarettes per day 11 % GOAL ‘
0 19,123 70.0 2 90
1-5 2,166 8.0
6-10 3,356 12.0
11-20 2,324 8.5
21-40 345 1.3
2 41 12 0.0

 

' Healthy People 2000 Objective goal: to increase the abstinence
from tobacco use by pregnant women to 2 90%.

79

Table 13. Alcohol use during third trimester of pregnancy by Michigan WIC Women in 1995 vs.

Healthy People 2000 Objective goal.

 

 

Drinks per Week n % GOAL a
NONE 25,435 97 2 95
1-2 160 0.6
3-10 117 0.4
2 11 37 0.1
Unknown Quantity 471 1.8

 

‘ Healthy People 2000 Objective goal: to increase the abstinence fi'om alcohol by pregnant

women to 2 95%.

Table 14. Incidence of low birthweight (LBW) and very low birthweight (VLBW) among infants
born to Michigan WIC women in 1995 vs. Healthy People 2000 Objective.

 

 

 

 

VLBW ' LBW "

Race/ Ethnicity Total n % GOAL ° 11 % GOAL °

(11) (°/o) ("/o)
All ° 33,684 349 1.0 s 1 2,513 7.4 s 5
White, not Hispanic 22,071 185 0.8 1,421 6.4
Black, not Hispanic 9,170 141 1.5 g 2 927 10.1 s 9
Hispanic 1,904 14 0.7 126 6.6
American Indian 187 3 1.6 12 6.4
Asian 352 6 1.7 27 7.7
p ° 0.000
’Defined as < 2500 grams.

bDefined as < 1500 grams.

°Healthy People 2000 Objective goal: to reduce the prevalence of LBW to s 5% of all live births
and VLBW to s 1% of all live births; s 9% and s 2% in the black population.
d Statistical significance of incidence of LBW and VLBW among race/ ethnicity subgroups

80

Table 15. Prevalence of growth retardation in Michigan WIC
infants and children by age-group in 1995-1998 vs. Healthy

 

 

People 2000 Objective.

Age-group All Stunted ’ GOAL b
n % %

1995

Total 300,703 12.0 s 10

0-5 mos 120,147 16.7

6-11 mos 28,881 11.7

12-23 mos 48,165 11.1

24-35 mos 41,536 5.5

36-48 mos 61,974 8.2

1996

Total 214,521 10.6 s 10

0-5 mos 61,262 16.4

6-11 mos 23,411 10.8

12-23 mos 42,925 10.6

24-35 mos 34,661 4.9

36—48 mos 52,262 7.6

1997

Total 218,307 10.4 < 10

0-5 mos 64,476 16.3

6-11 mos 24,238 10.2

12-23 mos 41,848 9.8

24-35 mos 34,370 4.7

36-48 mos 53,375 7.3

1998

Total 214,464 10.2 < 10

0-5 mos 64,370 16.0

6-11 mos 24,445 10.4

12-23 mos 41,134 9.6

24-35 mos 32,834 4.2

36-48 mos 51,681 7.0

 

' < 5’11 percentile height for age of NCHS reference.

bHealthy People 2000 Objective goal: to reduce the prevalence
of growth retardation to s 10% among low-income children
younger than 5 years of age.

81

Table 16. Prevalence of growth retardation among Michigan WIC infants
and children by race/ ethnicity 1995-1998 vs. Healthy People 2000

 

 

 

 

 

Race/Ethnicity All 0-5 6-11 12-23 24-35 36-48 Goal
mos. mos. mos. mos. mos.
n % % % % % %
1995
All 300,703 16.7 11.7 11.1 5.5 8.2 $10
White 120,147 14.2 10.5 9.7 5.1 8.6
Black 28,881 21.8 14.3 13.0 5.5 6.7
Hispanic 48,165 14.1 11.4 12.3 7.0 10.7
American Indian 41,536 11.2 9.1 9.1 4.1 5.0
Asian 61,974 13.4 11.6 14.0 9.7 13.6
1996
All 214,521 16.4 10.8 10.6 4.9 7.6 310
White 61,262 13.9 9.8 9.4 4.7 8.3
Black 23,411 21.3 12.9 12.2 4.9 5.9
Hispanic 42,925 14.5 10.9 12.0 6.4 9.6
American Indian 34,661 9.2 8.6 11.2 4.7 6.0
Asian 52,262 12.7 10.6 15.0 5.7 11.4
1997
All 218,307 16.3 10.2 9.8 4.7 7.3 s 10
White 64,476 14.3 9.6 8.8 4.4 7.7
Black 24,238 20.7 11.5 11.2 4.8 5.9
Hispanic 41,848 13.7 10.3 12.1 6.1 10.2
American Indian 34,370 12.2 9.2 8.7 3.8 5.7
Asian 53,375 14.9 9.6 11.0 5.1 10.9
1998
All 214,464 16.0 10.4 9.6 4.2 7.0 s 10
White 64,370 13.8 9.6 8.6 4.0 7.5
Black 24,445 20.8 11.8 10.8 4.1 5.4
Hispanic 41,134 13.5 10.9 12.3 5.9 9.6
American Indian 32,834 10.8 9.8 6.3 3.0 5.0
Asian 51,681 14.1 10.5 11.5 7.8 9.3

 

" Defined as <5th percentile height-for-age of NCHS reference.

 

b Healthy People 2000 Objective goal: to reduce the prevalence of growth
retardation to _<_ 10% among low-income children younger than 5 years of

age.

82

4.2 Objective 2: To determine whether maternal health status, lifestyle and
sociodemographic characteristics in Michigan WIC women predict the birthweight of
their offspring while accounting for infant characteristics such as gestational age and sex.

Correlation coefficients of predictor variables with birthweight were calculated
(Table 17). Maternal and infant predictors of infant birthweight identified were 1)
prepregnancy BMI, 2) weight gained or lost during pregnancy (pounds), 3) tobacco use
during pregnancy (cigarettes per day during third trimester), 4) monthly income ($), 5)
race/ethnicity (white, not Hispanic=l, all others=0), and 6) infant sex (boy=1, girl=0)
(Table 11). The three most significant positive predictors for infant birthweight are
prepregnancy BMI, weight gain during pregnancy, and race/ ethnicity of mother being
white, not Hispanic, with standardized beta coefficients 0.169, 0.156 and 0.144,
respectively (Table 11). Three out of the six predictors in this model are modifiable:
prepregnancy BMI, weight gain during pregnancy and tobacco use during pregnancy.
This suggests that improvements in birthweight can be achieved through intervention
prior to and during pregnancy.

In this linear regression model, the six independent variables accounted for 9.5%
of the variance in infant birthweight. These predictors that contributed significantly to
birthweight are very important to decreasing the number of LBW and VLBW infants in

the US.

83

Table 17. Correlation coefficients for independent variables with
birthweight

 

 

Independent Variable Birthweight

Pearson’s r
Mother's age (years) 0.01 *
Gestational age (days) 0.06*
Prepregnancy BMI (kg/m2) 0.13“
Alcohol use (drinks/week third trimester) -0.02*
Education (years) 0.03*
Previous pregnancies (n) 0.00
Previous live births (n) 0.01
Monthly income (S) 008*
Household size (n) 004*
Weight gain/loss during pregnancy (lbs.) 0.13*
Tobacco use (cigarettes/day third trimester) -0. 12*

Spearman’s rank

 

Race/ ethnicity (white, not Hispanic=l, others=0) -0. 12*
Migrant status (yes=1, no=0) 00]“
Marital status (yes=1, no=0) 0.10*
Infant sex (male=1, female=0) 0.11*
Food stamp participant (yes=1, no=0) -0.08*
Medicaid participant (yes=1, no=0) -0.04*
AFDC participant (yes=1, no=0) -0.08*
Breastfeeding initiation (yes=1, no=0) 0.10*
Iron status (anemic=1, normal=0) -0.20*
Number of infants (singleton=1, twins=2, etc.) -0.05*
Trimester prenatal care began -0.23*

 

* significant at p <0.05 level

84

Table 18. Multiple linear regression model to predict birthweight by maternal
characteristics.

 

 

 

Standardized Coefficients t-statistic b
Beta“

(Constant) 144.66 *
Pregnancy weight gain/loss (lbs.) 0.159 25.68 *
Pre-pregnancy BMI (kg/m2) 0.169 27.85 *
Smoking (cigarettes/day in 3rd trimester) -0. 139 -22.80 *
Race (White, not Hispanic = 1, others=0) 0.144 22.89 *
Infant sex (male = 1, female = 0) 0.103 17.34 *
Monthly Income ($) 0.035 5.73 *

 

R2 = 0.095

 

* significant at p<0.05 level

‘Beta (Standardized Coefficients) reflects the relative importance of each independent
variable in predicting birthweight
b t-test for the significance of each b coefficient

85

4.3 Objective 3: To identify factors differentiating low birthweight and normal
birthweight infants in the Michigan WIC population based on maternal health and

lifestyle factors during pregnancy.

4.3.1 Discriminant Analysis

Table 17 shows correlation coeflicients between maternal predictors and
birthweight. The final discriminant function included the following independent
variables: 1) race/ethnicity, 2) mother’s age, 3) prepregnancy BMI, 4) marital status, 5)
previous pregnancies, 6) weight gain during pregnancy, 7) smoking during third trimester
and 8) infant sex (Table 19 shows the descriptive statistics for the predictor variables).
All of the variables had Wilk’s Lambda significant (p < 0.05) by the F test. Mothers of
LBW infants were significantly older, thinner (lower BMI), greater number of previous
pregnancies, gained fewer pounds during pregnancy, and smoked more cigarettes than
mothers of NEW infants. The three most important independent variables based on
standardized canonical discriminant firnction coeflicients were weight gain during
pregnancy (0.569), race/ ethnicity (0.466) and prepregnancy BMI (0.442). The specificity

of this discriminant firnction is 99.9% (Table 21).

4.3.2 Logistic Regression

The variables in the final model were: 1) maternal age, 2) prepregnancy BMI, 3)
weight gain during pregnancy, 4) marital status, 5) maternal iron status during pregnancy,
6) race/ ethnicity, 7) sex of the infant, 8) number of infants resulting from the pregnancy,

and 9) tobacco use during pregnancy (Table 20 Logistic regression statistics). High

86

maternal age (2 40 years) have an increased risk (OR=1.46) of delivering a low
birthweight women compared to women 20-29 years. Low prepregnancy BMI (< 19.7)
have increased risk of LBW while women with high BMI (2 29.1) have a decreased risk
of LBW compared to normal weight women. Similarly, women with inadequate weight
gain during pregnancy are at increased risk and women with excessive weight gain are at
decreased risk of delivering LBW infant, OR=1.76, 0.62, respectively. Twins are more
likely to be LBW than singletons (OR=20.5). The odds ratio is extremely high (OR >100)
for triplets with a very wide 95% CI (0 - >1000), this is due to a small number of triplets
in the dataset. The specificity of the model was 99.6% (Table 21 shows the classification

results).

87

Table 19. Predictors for low birthweight identified by discriminant analysis.

 

 

 

Predictors LBWa NBWb Wilk's Standardized
Lambdac Canonical

Discriminant

Functions‘I

mean:SD mean:SD

Race (dummy)° 0.59 i 0.5 0.71 i 0.5 0.995 * 0.47
Age (years) 24.18 i 6.2 23.48 i 5.5 0999* -034
BMI(kg/m2) 24.62 i 6.6 25.73 :t 6.6 0998* 0.44
Marital Status (dummy)f 0.28 i 0.4 0.35 i 0.5 0998* 0.26
Previous Pregnanciesg 1.72 i 1.9 1.53 i 1.7 0.999* 0.05
Weight Gain (pounds)h 23.86 i 16.0 29.20 i 17.0 0992* 0.57
Smoking (cig./day)i 4.59 i 7.3 3.31 i 6.5 0997* -039
Infant Sex (dummy)i 0.45 i 0.5 0.51 i 0.5 0999* 0.22

 

* significant at p< 0.05 level

‘ Low birthweight (LBW) defined as < 2500 grams.

b Normal birthweight (NBW) defined as _>_ 2500 grams.

° Wilk’s Lambda is used in an AN OVA (F ) test of mean difference in discriminant
analysis. Values from 0-1, with smaller values indicating the independent variable
contributes more to the discriminant function.

4 Standardized Canonical Discriminant Functions are used to compare the relative
importance of the independent variables.

° White, not Hispanic = 1, to indicate reference category; all remaining race/ ethnic

groups = 0
married =1 and not married = 0.

‘3 Self reported number of previous pregnancies, regardless of outcome.
" Self-reported number of pounds gained or lost by mother during gestation.
' Self-reported number of cigarettes smoked per day during the third trimester of

pregnancy.
’ boy =1, girl = 0.

88

Table 20. Odds ratios and Confidence Intervals for Low Birthweight (< 2500g) by

 

 

Maternal Characteristics
R ‘ Odds Ratio " 95% Cl C
Maternal Age (y) 0.04
<16 0.00 1.10 0.82, 1.48
16-19 0.00 1.08 0.95, 1.22
20-29 1.0
30-39 0.04 1.46 1.27, 1.68
240 0.00 1.36 0.78, 2.36
Pre-pregnancy BMI 0.08
Very Underweight (s 17.9) 0.05 1.83 1.50, 2.23
Underweight (18- 19.7) 0.02 1.26 1.09, 1.47
Normal (19.8- 26.0) 1.0 '
Overweight (26.1-29.0) 0.00 0.94 0.80, 1.12
Very Overweight (2 29.1) -0.05 0.66 0.57, 0.76
Weight Gain During Pregnancy d 0.13
Below Ideal 0.08 1.76 1.56, 1.97
Ideal 1.0
Above Ideal -0.05 0.62 0.53, 0.72
Smoking (cig./day) 0.08
0 1.0
1-9 0.06 1.83 1.55, 2.15
10-19 0.06 1.82 1.55, 2.14
2 20 0.02 1.20 1.06, 1.36

 

’ R- Partial contribution- assesses the relative importance of the independent variable
b Odds ratio of the independents with the dependent.
° 95% Confidence Interval for the Odds Ratio
d Institute of Medicine Standards: 28-40 pounds for underweight pre-pregnancy (BMI
s 19.7); 25-35 pounds for normal weight pre-pregnancy (BMI 19.8-29.0); 15-25

pounds for overweight pre-pregnancy (BMI 2 29.1).

89

n" “it.“

Table 20. Odds ratios and Confidence Intervals for Low Birthweight (< 2500g) by

 

 

Maternal Characteristics
R * Odds Ratio " 95% 0°

Race/ Ethnicity

White, not Hispanic 1.0

Others 0.07 1.58 1.42, 1.76
Marital Status

Married 1.0

Not Married 0.04 1.33 1.18, 1.50
Iron Status

Normal 1.0

Anenric 0.00 0.90 0.78, 1.04
Infant Sex

Boy 1.0

Girl 0.04 1.29 1.17, 1.43
Number of Infants 0.27

Singleton 1.0

Twin 0.27 20.5 16.9, 24.9

Triplet 0.00 >100 0, >1000

 

‘ R- Partial contribution- assesses the relative importance of the independent variable

b Odds ratio of the independents with the dependent.
° 95% Confidence Interval for the Odds Ratio

d Centers for Disease Control Standards for anemia: The first, second, third trimester
and postpartum <11, <10.5, <11, <12 g/100ml for hemoglobin and < 33%, < 32%, <

33%, 36% for hematocrit, respectively.

90

Table 21. Specificity, sensitivity, negative predictive value (PV-)
and positive predictive value (PV+) of discriminant function and
logistic regression model in Michigan WIC population, 1995.“

 

Discriminant Function

 

Actual
LBW NBW
Predicted (risk) Eon-risk)
LBW (risk) 4 6

NBW (non-risk) 2,069 22,606
Total 2,073 22,612

 

Sensitivity b (4/ (2,069+4)) *100=o.2%
Specificity ° (22,606/(22,606+6))* 100=99.9%

 

 

PV- " (22,606/(22,606+2,069)) *100= 91.6%
PV+ ° (4/(4+6)) * 100 = 40%
Logistic Regression Model
Actual
LBW NBW
Predicted (risk) (non-risk)
LBW (risk) 156 98

NBW (non-risk) 1,692 22,017
Total 1,848 22,115

 

Sensitivity " (1,56/ (156+1,692)) *100=8.4%
Specificity ° (2,2017/(22,017+98))* 100=99.6%
PV- ‘ (22,017/(22,017+1,692)) *100= 92.9%
PV+ ° (156/Q 56+98)) * 100 = 61.4%

 

' Calculations are related to the absence of LBW (>2500g
birthweight) or the presence of LBW (<2500g).

b Sensitivity is the fraction of LBW infants the discriminant function
or logistic regression model predicts correctly.

° Specificity is fraction of infants NBW that the discriminant firnction
or logistic regression model predicts correctly.

d PV- is the fi'action of negative results according to discriminant
function or logistic regression model that are true negatives.

° PV+ is the fraction of positive results according to the discriminant
function or logistic regression model that are true positives.

91

4.4 Objective 4: To identify the factors associated with growth retardation among

children in the Michigan WIC population at various age-group categories.

4.4.1 Children 0-5 Months

Tables 22-24 Show correlation coefficients of potential predictors and height-for-
age percentile at 0-5 months of age. The following variables were included in the final
logistic regression model to predict stunting at 0-5 months: prepregnancy BMI, tobacco
use during pregnancy, birthweight, weight gain during pregnancy, iron status, and
number of infants resulting from the pregnancy (Table 25 logistic regression model
statistics). The three most important predictors by R values were: LBW (0.39), weight
gain during pregnancy (0.06), and number of infants resulting fiom the pregnancy (0.04).
Cigarette smoking followed a dose-dependent response with greater amount of cigarettes
smoked per day during the third trimester, the higher odds ratio in relation to stunting at
0-5 months of age (1.43, 1.49, 1.66 for 0-9 cigarettes per day, 10-19, and 2 20,
respectively). Above ideal weight gain (OR=0.73) had a protective efi’ect on Stunting at
0-5 months of age when compared to less than ideal (OR= 1.22). LBW infants were
31.48 times more likely to be stunted at 0-5 months than NBW infants. Meanwhile,
twins and triplets were at a greater risk of stunting than singleton births. Table 30 shows
the classification results of predicting stunting at 0-5 months of age by logistic regression.
A high specificity was calculated (98.1%) while a low sensitivity (42.5%) was found

(Table 31).

92

4.4.2 Children 6-11 Months

Tables 22-24 Show correlation coeflicients of potential predictors of stunting at 6-
11 months of age. The following variables were included in the final logistic regression
model to predict stunting at 6-11 months: stunting at 0-5 months of age, household size,
tobacco use during pregnancy, birthweight, weight gain during pregnancy, and infant sex
(Table 26 logistic regression model statistics). The three most important predictors by R-
values were: stunting at 0-5 months ofage (0.18), LBW (0.13), and female sex (-0.05).
Similar to findings at 0-5 months of age, cigarette smoking followed a dose-dependent
response with greater amount of cigarettes smoked per day during the third trimester,
higher odds ratio in relation to stunting at 6-11 months of age (1.19, 1.34, 1.31 for 1-9
cigarettes per day, 10-19, and 2 20, respectively.) Above ideal weight gain (OR=0.82)
had a protective effect on stunting at 6-11 months of age. LBW were 2.91 times more
likely to be stunted at 6-11 months than NBW infants. Males were at a greater risk of
being stunted at 6-11 months than females (OR=0.76). Table 30 shows the classification
results of predicting stunting at 6-11 months of age by logistic regression. A high

specificity was calculated (99.7%) and a low sensitivity (2.7%) was found (Table 31).

4.4.3 Children 12-23 Months

Tables 22-24 shows correlation coefficients of potential predictors of stunting at
12-23 months of age. The following variables were included in the final logistic
regression model to predict stunting at 12-23 months: stunting at 0-5 months of age,
stunting at 6-11 months of age, when prenatal care began, household Size, birthweight,

and infant sex (Table 27 logistic regression model statistics). The three most important

93

predictors by R- values were: stunting at 6-11 months of age (0.37), stunting at 0-5
months of age (0.12) and female sex (0.05). Cigarette smoking by the mother during
pregnancy did not have an effect on the linear growth retardation of the child at this age-
group. Stunting at previous age groups was a significant predictor of stunting at 12-23
months. Prior stunting was a significant risk factor (OR=2.58, 12.16 for stunting at 0-5
and 6-11 months of age, respectively). Males were at a greater risk of being stunted at
12-23 months than females (OR=0.71). Table 30 Shows the classification results of
predicting stunting at 12-23 months of age by logistic regression. A high specificity was

calculated (97.8%) and a low sensitivity (4.0%) was found (Table 31).

4.4.4 Children 24-35 Months

Tables 22-24 shows correlation coefficients of potential predictors of stunting at
24-35 months of age. The following variables were included in the final logistic
regression model to predict Stunting at 24-35 months: stunting at 6-11 months of age,
stunting at 12-23 months of age, weight gain during pregnancy, maternal iron status
during pregnancy (Table 28 logistic regression model statistics). The most important
predictor by R value were: stunting at 12-23 months of age (0.34). Above ideal weight
gain (OR=0.65) had a protective effect on stunting at 24-35 months of age . Maternal iron
status at the initial visit had an effect on stunting at 24-35 months of age. Mother’s who
were anemic at the initial visit, their offspring were 1.7 times more likely to be stunted at
24-35 months of age than toddlers whose mothers with normal iron status. Birthweight
did not have an effect on stunting at this age group. Stunting at previous age groups had

the highest odds ratios, stunting at 6-11 months (OR=3.53) and stunting at 12-23 months

94

(OR= 14.98) (Table 30 Shows the classification results of predicting stunting at 24-35
months of age by logistic regression). A very high specificity was calculated (99.9%) and

a low sensitivity (5.4%) was found (Table 31).

4.4.5 Predicting Growth Retardation Among Children 36-48 Months

Tables 22-24 correlation coefficients of potential predictors of stunting at 36-48
months of age. The following variables were included in the final logistic regression
model to predict stunting at 36-48 months: stunting at 6-11 months of age, stunting at 12-
23 months of age, and stunting at 24-35 months of age (Table 29 logistic regression
model statistics). The most important predictor by R value was: stunting at 24-35 months
of age (0.33). The maternal variables did not contribute to the prediction of stunting in
this population. (Table 30 Shows the classification results of predicting stunting at 36-59
months of age by logistic regression. A high specificity was calculated (99.3%) and a low

sensitivity (39.5%) was found (Table 31).

95

Table 22. Pearson’s (r) correlation coeflicients of predictor variables and height-for-
age percentile of Michigan WIC infants and children, 1995-1998.

Height-for-age Percentile

 

0-5 6-1 1 12-23 23-35 35-48
mos. mos. mos. mos. mos.
Height-for-age percentile
0-5 mos. 0.44 0.39 * 0.31 0.29 *
6-11 mos. 0.71 * 0.61 0.59 *
12-23 mos. 0.72 0.70 *
24-35 mos. 0.81 *
Mother's age (years) -0.02 * 0.01 0.04 0.04 0.05 *
Gestational Age (days) 0.04 * 0.03 0.01 * 0.01 0.01
Prepregnancy BMI(kg/m2) 0.07 * 0.01 0.02 0.02 0.05 *
Previous pregnancies (number) -0.04 * -0.05 * -0.06 * -0.03 -0.02
Previous live births (number) -0.02 * -0.01 * -0.05 * -0.03 -0.01
Household income ($) 0.04 * 0.02 * 0.01 * 0.02 -0.00
Medical care began (trimester) 0.01 -0.02 * -0.02 * -0.01 -0.01
Household size (number of people) 0.01 * -0.04 * -0.05 * -0.05 -0.03 *
Smoking 3lrd trimester (cig. /day) -0.12 * -0.10 * -0.08 * -0.07 -0.09 *
Drinking 3ml trimester (drinks/ week) -002 * -001 0.00 0.00 0.01

 

* significant at p<0.05 level

96

Table 23. Spearman’s (rank) correlation coefficients of predictor variables and height-for-
age Percentile of Michigan WIC infants and children, 1995-1998.

Height-for-age Percentile
0-5 mos. 6-11 mos. 12-23 mos. 23-35 mos. 35-48 mos.

 

Weight gain during pregnancya 0.12 * 0.10 * 0.08 * 0.07 * 0.06 *
Iron status (initial visit) " -005 * -002 * -003 * -003 * -003 *
Infant sequence numberc -0.17 * -0.09 * -0.07 * -0.03 * -0.01
Marital statusd 0.07 * 0.02 * 0.00 -001 -002
Infant sex" -001 * 0.04 * 0.05 * -001 0.02
Breastfeeding initiationf 0.05 * 0.04 * 0.04 * 0.02 * 0.01
Food Stamp participationg -0.06 * -0.05 * -0.05 * -0.03 * -0.00
Medicaid participation‘ -003 * -003 * -0.02 * -0.03 * -0.02
AFDC participationg -0.07 * -0.05 * -0.05 * -0.04 * -0.01
Race/Ethnicityj -010 * -003 * -004 * 0.00 0.06 *

 

* significant at p<0.05 level

‘ Institute of Medicine Standards: 28-40 pounds for underweight pre—pregnancy (BMI<
19.7); 25-35 pounds for normal weight pre-pregnancy (BMI 19.8-29.0); 15-25 pounds for
overweight pre-pregnancy (BMI 2 29.1); < recommended, ideal, > recommended.

b Centers for Disease Control Standards for anemia: The first, second, third trimester and
posptartum for <11, <10.5, <11, <12 g/100ml for hemoglobin and < 33%, < 32%, < 33%,
36% for hematocrit, respectively, (anemic=1, normal=0)

° singleton=1, twins=2, triplets=3

d married=1, not married=0

° boy=1, girl=0

f infant breastfed at least once during early postpartum period=1, never breastfed=0.
Fparticipant=1, not participant=0

’ white, not Hispanic=1, all others=0

Table 24. Pearson’s (r) correlation coefficients of birthweight and height-for-age percentile
of Michigan WIC infants and children, 1995-1998.

 

Height-for-age Percentile
0-5 mos. 6-11 mos. 12-23 mos. 23-35 mos. 35-48 mos.

 

Birthweight 0.61 * 0.40 * 0.34 * 0.28 * 0.24 *

 

* significant at p<0.05 level

97

Table 25. Predictors for stunting ‘ at 0-5 months identified by logistic regression.

 

 

Predictors Rb OR ° 95% c11
Birthweight (g)

Normal (5 2500) 1.0

Low ( <2500) 0.39 31.48 27.5, 36.1
Prepregnancy BMI

Very underweight (s 17.9) 0.02 1.39 1.12, 1.71

Underweight (18.0- 19.7) 0.00 1.06 0.91, 1.24

Normal (19.8-26.0) 0.02 1.00

Overweight (261-290) 0.00 0.99 0.85, 1.16

Very overweight (2 29.1) 0.00 0.91 0.80, 1.03
Smoking during pregnancy (cig/day)

0 0.06 1.00

1-9 0.04 1.43 1.24, 1.66

10-19 0.04 1.49 1.30, 1.70

2 20 0.05 1.66 1.41, 1.95
Weight Gain ‘

Below Ideal 0.03 1.22 1.09, 1.36

Ideal 0.06 1.00

Above Ideal -0.03 0.73 0.64, 0.83

Maternal Iron Status 8

Normal 1.0

Anemic 0.01 1.18 1.03, 1.34
Number of Infants

Singleton 0.04 1 .00

Twins 0.04 2.25 1.71, 2.97

Triplets 0.00 74.71 0.00, > 100

 

" < 5th percentile height-for-age of NCHS reference.

b R- partial contribution- assesses the relative importance of the independent
variable;

° Odds ratios of the independents with the dependent

d 95% Confidence Interval for the odds ratio

fInstitute of Medicine Standards: 28-40 pounds for underweight pre-pregnancy
(BMI< 19.7); 25-35 pounds for normal weight pre-pregnancy (BMI 19.8-29.0);
15-25 pounds for overweight pre-pregnancy (BMI > 29.1).

3 Centers for Disease Control Standards for anemia: The first, second, third
trimester and postpartum for <11, <10.5, <11, <12 g/ 100ml for hemoglobin and <
33%, < 32%, < 33%, 36% for hematocrit, respectively.

98

Table 26. Predictors for stunting ‘ at 6-11 months identified by
logistic regression.

 

 

Predictors R b OR ° 95% CI 6
Birthweight (g)

Normal (2 2500) 1.0

Low (< 2500) 0.13 2.91 2.47, 3.42
Ht-for-age %ile at 0-5 months

Normal 1.0

Stunted " 0.18 3.76 3.25, 4.33
Household Size (persons) 0.02

1-3 1.0

4-6 0.02 1.16 1.01, 1.33

2 7 0.01 1.40 0.94, 2.10
Smoking during pregnancy (cig/day)

0 0.03 1.0

1-9 0.01 1.19 0.99, 1.43

10-19 0.03 1.34 1.13,1.57

2 20 0.02 1.31 1.07, 1.60
Weight gain ° 0.03

Below ideal 0.01 1.12 0.98, 1.27

Ideal 1.0

Above ideal -0.02 0.82 0.71, 0.96
Infant sex

Boy 1.0

Girl 0.05 0.76 0.68, 0.85

 

‘ < 5ih percentile height-for-age of NCHS reference population.

b R- partial contribution- assesses the relative importance of the
independent variable;

° Odds ratios of the independents with the dependent

d 95% Confidence Interval for the odds ratio

° Institute of Medicine Standards: 28-40 pounds for underweight

pre-pregnancy (BMI s 19.7); 25-35 pounds for normal weight
pre-pregnancy (BMI 19.8-29.0); 15-25 pounds for overweight
pre-pregnancy (Blvfl 2 29.1).

99

Table 27. Predictors for stunting ‘ at 12-23 months identified

by logistic regression.

 

 

Predictors Rb OR ° 95% CI 6
Birthweight (g)

Normal (2 2500) 1.0

Low (< 2500) 0.01 1.19 0.96, 1.47
Ht-for-age %ile at 0-5 months

Normal 1.0

Stunted ' 0.12 2.58 2.16, 3.08
Ht-for-age %ile at 6-11 months

Normal 1.0

Stunted ' 0.37 12.16 10.49, 14.09
Prenatal care began (trim) 0.02

First 1.0

Second 0.00 1.06 0.88, 1.27

Third 0.02 1.71 1.11, 2.63
Household Size (persons) 0.01

1-3 1.0

4-6 0.01 1.14 0.97, 1.33

2 7 0.01 1.51 0.97, 2.34
Infant Sex

Boy 1.0

Girl -0.05 0.71 0.62, 0.82

 

‘ < 5‘h percentile height-for-age of NCHS reference

population.

- partial contribution- assesses the relative importance of

the independent variable;

° Odds ratios of the independents with the dependent
d 95% Confidence Interval for the odds ratio

100

 

Table 28. Predictors for stunting ‘ at 24-35 months identified by

logistic regression.

 

 

Predictors R b OR c 95% CI ‘T
Ht-for-age %ile at 6-11 months

Normal 1.0

Stunted ‘ 0.16 3.53 2.62, 4.77
Ht-for-age %ile at 12-23 months

Normal 1.0

Stunted ‘ 0.34 14.98 11.03, 20.35
Weight gain during pregnancy °

Ideal . 0.06 1.00

Below ideal 0.03 1.33 0.99, 1.76

Above ideal -0.03 0.65 0.44, 0.97
Maternal Iron Status

Normal 1.0

Anemic ‘ 0.05 1.68 1.18, 2.39

 

‘ < 5fh percentile height-for-age of NCHS reference population.
b R- partial contribution- assesses the relative importance of the

independent variable;

° Odds ratios of the independents with the dependent

d 95% Confidence Interval for the odds ratio
° Institute of Medicine Standards: 28-40 pounds for underweight
pre-pregnancy (BMI s 19.7); 25-35 pounds for normal weight
pre-pregnancy (BMI 19.8-29.0); 15-25 pounds for overweight

re-pregnancy (BMI 2 29.1).

Centers for Disease Control Standards for anemia: The first,
second, third trimester and postpartum for <11, <10.5, <11, <12
g/ 100ml for hemoglobin and < 33%, < 32%, < 33%, <36% for

hematocrit, respectively.

101

Table 29. Predictors of stunting ‘ at 36-48 months
identified by logistic regression

 

 

Predictors R b OR ° 95% CI d
Ht-for-age %ile at 6-11 months

Normal 1.0

Stunted ‘ 0.07 1.91 1.26, 2.91

Ht-for-age %ile at 12-23 months

Normal
Stunted‘ 0.25 7.75 5.24,11.44

Ht-for-age %ile at 24-35 months

 

Normal 1.0

Stunted ‘ 0.33 22.28 14.26, 34.81
‘ < 5’11 percentile height-for-age of NCHS reference
population.

- partial contribution- assesses the relative importance
of the independent variable
° Odds ratios of the independents with the dependent
d 95% Confidence Interval for the odds ratio

102

 

1‘

_—
s

Table 30. Classification tables for logistic regression models predicting
stunting ‘ in the Michigan WIC population, 1995-1998.‘

 

 

Actual
Age of Risk Normal Total
children (stunted)
0-5 months Risk (stunted) 1219 326 1845
Predicted Normal 1649 16653 1 83 02

Total 2868 16979

 

 

 

 

Actual
6-11 months Risk (stunted) 38 48 86
Predicted Normal 1433 143 87 15820
Total 1471 14435
Actual
12-23 months Risk (stunted) 35 253 288
Predicted Normal 828 11340 12168
Total 863 11593
Actual
24-35 months Risk (stunted) l6 8 24
Predicted Normal 278 7696 7974
Total 294 7704
Actual
36-48 months Risk (stunted) 87 25 112
Predicted Normal 133 3761 3894
Total 220 3786

 

a Stunting < 5’11 percentile height-for-age of NCHS reference population.
b Normal 2 5th percentile height-for-age of NCHS reference population.

103

Table 31. Specificity, sensitivity, negative predictive value (PV-) and
positive predictive value (PV+) of logistic regression models for
predicting stunting in the Michigan WIC population, 1995-1998.‘

Age Group Sensitivityb Specificityc PV—d PV+°

0-5 months 42.5 98.1 91.0 78.9
6-11 months 2.7 99.7 90.9 44.2
12-23 months 4.0 97.8 93.2 12.2
24-35 months 5.4 99.9 96.6 66.7
36-48 months 39.5 99.3 96.6 77.7

 

‘ Calculations are related to the absence of stunting (2 5'11 percentile
height for age on NCHS reference charts) or the presence of stunting
(<5'h percentile height for age).

b Sensitivity is the fraction of stunted infants the logistic regression
model predicts correctly

° Specificity is fraction of infants not stunted that the logistic regression
model predicts correctly

d PV- is the fraction of negative results according to logistic regression
model that are true negatives.

" PV+ is the fraction of positive results according to the logistic
regression model that are true positives.

104

Chapter 5

DISCUSSION

5.1 Overall
The most unique contribution of this research to the existing body of

knowledge was the development of the longitudinal dataset. The creation of a single
dataset which tracks Michigan WIC participants from in utero up to the age of four years
was innovative and essential to examine the proposed objectives. As a result of the
information contained in the newly developed dataset, previously difiicult to answer
research questions were addressed. The data from the present study contributed
significantly to the body of knowledge in the area of maternal and child health in a high
risk population, especially with respect to growth retardation among children.

Numerous predictors of the outcomes of interest, LBW and stunting, were
identified through sound statistical methodology. The predictors identified work in
context with one another and a single best predictor was not identified. Therefore, the
health of mother and child must be examined through holistic approaches. Modifiable
risk factors similar to those known for the general population such as weight gain and
tobacco use during pregnancy are needed also a priority for WIC mother in Michigan.
Through changes in risk factors identified, improvements in outcome are possible and
worthy of investigation.

Low birthweight was successfully predicted by maternal health, lifestyle and
sociodemographic characteristics. Maternal factors served as significant predictors of

growth retardation in offspring through the age of two years. Stunting is persistent

105

among Michigan WIC children where stunting at younger age-groups is a significant
predictors of stunting in older children, aged three and four years. The results of this
study are consistent with previously published research in low-income, high-risk
populations (Peny et al., 1995; Cogswell and Yip, 1995; Adair and Guilkey, 1997).
Consideration of measurements, standards, and cut-offs were important when
interpreting results and drawing conclusions. Numerous definitions of overweight and
obesity are available for researchers to employ in their studies. Healthy People 2000
Objectives defines overweight as BMI > 27.3 in women aged 20 and older. This
definition was used in Objective One to produce results comparable to baseline and
follow-up measurements used in monitoring the goal at the national level. The CDC used
a different definition of overweight and because data examined in this study were
property of the CDC, their cut-off values were used in the present study. The iron
measurements available in this study (hemoglobin and hematocrit) are not used to
diagnose iron deficiency, which requires serum ferritin, erythrocyte protoporphyrin,
transferring saturation. Therefore, the Healthy People 2000 Objective goal was modified
to assess anemia rather than iron deficiency. The difference in measurements is
important because hemoglobin and hematocrit are late indicators of iron status, the actual
prevalence of iron deficiency may be higher than the prevalence of anemia determined in

the present study.

5.2 Objective 1
The Michigan WIC population achieved only one of the nine Healthy People

2000 Objectives evaluated. The Michigan WIC population was far from meeting most of

106

 

the goals of the Healthy People 2000 Objectives. The objectives that were most deviant
from the goals are minimum weight gain (62.8% vs. 85% Healthy People 2000 Objective
goal). Ironically, the low percent of women who gained the minimal recommended
weight coexisted with the high prevalence of prepregnancy overweight (3 6% vs. 20%
Healthy People 2000 Objective goal) although the minimum weight gain during
pregnancy was adjusted from prepregnancy weight status. Michigan WIC program’s met
the goal related to alcohol abstinence (97% vs. 95% Healthy People 2000 Objective
goal), but needs to be carefully interpreted in relation to the validity of responses.

The Michigan WIC program was compared to the overall US. population with
respect to the Healthy People 2000 Objectives examined in this study (Table 32). The
prevalence of overweight among Michigan WIC women was similar to the US.
prevalence for overweight, 36% and 37%, respectively. The rate of tobacco incidence
among Michigan WIC women was similar to the US. general population, 70% and 71%,
respectively. The rate of breastfeeding initiation among Michigan WIC mothers was
similar to the US. general population, 41% and 40%, respectively. The Michigan WIC
population was abstaining from alcohol use during pregnancy at a greater rate than the

US. general population, 97% and 85%, respectively.

107

5;”- m

t and-

Table 32. Michigan WIC program vs. US. population in progress toward
achievement of Healthy People 2000 Objectives.

 

 

 

 

Healthy People 2000 Objective MI WIC US Healthy
1995 1994 People 2000
Goal
% % %
Prevalence of overweight ° 36 37 s 25
Minimum weight gain during pregnancy d 37 62.8 ‘ 2 85
Tobacco abstinence during pregnancy 70 71 2 90
Alcohol abstinence during pregnancy 97 81 b 2 95
Early prenatal care ° 68 81 2 90
Maternal anemia during pregnancy f 14 8 s 4
Maternal anemia postpartum g 39 8 s 4
Low birthweight (<2500g) 7 7 g 5
Breastfeeding initiation h _ 41 40 2 75
Growth retardation in children ‘ 10 8 2 10
1: 1988 data
1993 data

:Overweight defined as BMI > 27. 3 in women aged 20 years and older.
dMinimum weight gain defined by Institute of Medicine standards: 28-40 pounds for

underweight pre-pregnancy (BMI s 19.7); 25- 35 pounds for normal weight pre-

pregnancy (BMI 19.8-29.0); 15-25 pounds for overweight pre-pregnancy (BMI 2
29. 1).

° Medical care during the first trimester of

Centers for Disease Control Standards for anemia: The first, second, and third
trimester for <11, <10.5, <11g/100ml for hemoglobin and < 33%, < 32%, < 33% for
hematocrit, respectively.
3 Centers for Disease Control Standards for postpartum anerrria: <12 g/100ml for
hemoglobin or <3 6% for hematocrit

hInfant breastfed at least one time during the early postpartum period
'Stunting < 5th percentile height- for-age of NCHS reference population

5.3 Objective 2

The hypothesis set forth for this objective proved true, as certain maternal health
Status and lifestyle characteristics predicted the birthweight of their offspring. Two most
important predictors of birthweight, prepregnancy BMI and weight gain during

pregnancy, were modifiable and can be improved through intervention. The

108

 

predictability of prepregnancy BMI and weight gain during pregnancy for birthweight
however, was confined with race/ ethnicity and other predictors for the birthweight. The
third most important predictor variable was race/ ethnicity, which suggested that there are
differences in birthweight that are attributable to the race/ ethnicity of the mother (CDC/
NCHS/NCVS, 2000).

Although it may be diflicult to make changes in the prepregnancy weight of
women in the WIC program because the majority do not enter into WIC care until late in
pregnancy, advising women regarding the recommended amount of weight gain during
pregnancy is feasible. Women are weighed as a routine part of prenatal care and one
study found that advice given by health care providers regarding weight gain was
Significantly correlated with actual weight gain (Cogswell et al., 1999).

The beta value for smoking was a negative, confirming the negative impact of
tobacco use on birthweight reported in numerous other studies (Cnattingus and Haglund,

1997; Hellerstedt et al., 1997; Seeker-Walker et al., 1997, ‘Horta et al., 1997).

5.4 Objective 3

This objective incorporated two powerful statistical methods, discriminant
analysis and logistic regression. The results were consistent by both methods.

A low specificity (0.2%) (fraction of LBW infants the model predicts correctly)
was calculated for the classification of subjects by the discriminant function, the
researchers suggest that this may be due, in part, to the relatively low prevalence of LBW
(7.5%) in this population compared to NBW. The low prevalence of LBW in this sample

could also be attributed to the high negative predictive value (the fraction of negative

109

results according to the firnction that are true negatives) and high sensitivity (fraction of
NBW infants the model predicts correctly), where the discriminant function classified the
majority of infants as NBW.

The logistic regression model identified many of the same maternal characteristics
as the discriminant firnction, including maternal age, prepregnancy BMI, weight gain
during pregnancy, marital status, previous pregnancies, tobacco use during pregnancy,
and infant sex. The results of the present study were similar to previous research that
showed increased risk of LBW associated with: adolescence (< 16 years), advanced
maternal age (2 30 years), underweight prepregnancy (BMI < 17 .9) (Perry et al., 1995),
inadequate weight gain (Cogswell and Yip, 1995), female infant, multiple birth, and
smoking (Cnattingus and Haglund, 1997). Variables that were inverse predictors for the
risk of having a LBW infant were very overweight prepregnancy (BMI > 29.1) and
excessive weight gain during pregnancy (Perry et al., 1995; Johnson et al., 1992).

The majority of these risk factors were modifiable. Maternal age, prepregnancy
BMI, weight gain, and smoking could be modified through effective farme planning
methods and preconception counseling to reduce or increase weight status prior to
pregnancy, respectively. The relationship between quantity of cigarettes smoked and
odds ratio is the opposite of expected dose-dependent response direction found in other
studies (‘Horta et al., 1997, Cnattingus and Haglund, 1997). However, tobacco use at any
level was a significant predictor of LBW. In this study, women who smoked 2 20
cigarettes per day had nearly and equivalent odds ratio to non-smokers (1.15 (95% CI
1.03-1.28) vs. 1.0 in non-smokers). Smoking cessation programs among pregnant

women have shown themselves to be beneficial and cost-effective (Hueston et al., 1994).

110

‘f‘__‘*

5.5 Objective 4

The majority of the literature review related to stunting was fi'om developing
nations where the prevalence of grth retardation is much higher than in the US.
Although the populations are not necessarily generalizable to each other, they provided
insight into the patterns, causes and interventions associated with growth retardation.

Maternal characteristics were overwhelming predictors of stunting at 0-5 months.
The predictors of LBW (objective 3) were also the predictors of stunting at age 0-5
months, reflecting birthweight and insufficient time to demonstrate catch-up growth.
This finding stressed the importance of maternal health prior to and during pregnancy on
the health of her infant. Very underweight prepregnancy (BMI < 17.9), tobacco use
during pregnancy, inadequate weight gain, and maternal anemia were the maternal
characteristics identified with increased risk of stunting for 0-5 months old infants.

LBW was a critical predictor of stunting in 0-5 months group (OR 31.48).
Confirming birthweight is an important predictor of postnatal growth (Binkin et al.,
1988). Infant birthweight continued to be an important predictor of stunting through two
years of age. LBW infants were at an increased risk of Stunting compared to NEW
infants. It was noted that as children aged, the maternal characteristics were no longer
significant predictors of stunting. Stunting appeared to be persistent, with previous
history of stunting being a significant predictor of stunting in older age groups. LBW
was not a predictor of stunting at 36-48 months, suggesting that children overcome the

effects of birthweight at this time in this target population.

111

Due to eligibility requires and priority status, the WIC program retains only very
high-risk children in the older age groups. These children are at a greater risk for health
and nutritional problems, including stunting (\Vrlcox and Marks, 1994). The priority
status is reflected in the rapid decline in the number of participants in PedNSS as their
age increases. Because of this issue Yip et al., (1992) state that cross-sectional data of

children enrolled in WIC cannot be compared fairly across all ages.

112

"1‘4

': g.’ ." .
.u

"0
I I,
Am.‘-. -

its:

Chapter 6

STRENGTHS AND LIMITATIONS

6.1 General aspects related to the datasets

The strengths and limitations in this study are primarily related to the datasets
themselves. It is important to address the strengths to justify this study and the
importance of its objectives. The limitations are important to understanding the scope
and generalizability of study and to identify areas of research needed to confirm the

findings of this study.

6.2 Strengths of the datasets

The Pregnancy Nutrition Surveillance System is a unique dataset in that it is the
only system in the US. that collects numerous indicators on a continuous basis for low-
income pregnant women. The large numbers of records collected allows researchers to
perform detailed analyses and comparisons. PNSS is also a low-cost surveillance system
because it is based on data that states routinely collect anyway (Perry et al., 1995).

This method of evaluating the Michigan WIC program services allows public
health officials to track progress over time. One advantage of comparing this population
to the Healthy People 2000 Objectives is that it serves as a standard to which the program
should be striving to achieve. The researchers believe that the analysis completed in this
project demonstrated in a quantitative manner, areas of success and areas that need

improvement.

113

The longitudinal dataset created in this study provides the tool to examine
pregnancy and its potential long-term effects on a child. Cross-sectional datasets do not

allow researchers to study this relationship.

6.3 Weaknesses of the datasets

The greatest limitation for nutrition researchers using the PNSS and PedNSS is
the lack of detailed dietary information. There is no record of food intake or usual
dietary patterns. In addition, there is no medical history information, which would
facilitate greater use of the datasets and more in-depth analysis. These factors may have
had important implications.

Analyzing PNSS at the national level does have its limitations, including changes
in the number of states reporting, counties and clinics participating and program
eligibility. The greatest limitation, as seen in the use of PNSS in this project is the
variation in data quality, where the accuracy of reporting is difficult to define (Perry et
al., 1995). PNSS is not a randomized sample of the general population, rather it is a
convenience sample and the results are only generalizable to the PNSS population (Perry
et al., 1995).

PedNSS is a nutrition surveillance system and is not designed to identify the
specific causes in stunting in each child, however it works effectively at the population
level. A greater than expected prevalence of stunting (>5% of children at < 5’h percentile
height-for-age) indicates that some of the children in the population are stunted due to

health and/ or nutritional reasons. Only long-term and significant changes in

114

'mm

environment, health, and nutritional factors will affect the height distribution of a
population (Wilcox and Marks, 1994).

One record is created for each visit a child makes to a WIC clinic, which results in
multiple records per child. This is a potential problem due to the fact that the nutritional
and health status of the child is counted more than once. The average is 2.1 visits per

year (Yip et al., 1992).

115

1'»- afila'i-J.’ImA—:Ii

Chapter 7

RECOMMENDATIONS FOR FUTURE STUDIES

4.7 Low birthweight

In the present study, the maternal characteristics associated with LBW in the
target population were identified. Identification is only the first Step in the overall goal of
this project, to reduce poor birth outcomes. Future studies are needed to provide WIC
program directors with areas of critically needed intervention. A literature review of
interventions that are most successful and cost-effective in the low-income population
would be helpful for documenting the efficacy for focused intervention efforts. Carefully
planned interventions and detailed documentation of results will be key in reducing the
prevalence of LBW in the target population.

In order to confirm the results of the present study, it would be beneficial to
collect a more detailed “picture” of the pregnancy. This would include a diet record (e.g.
24 hour recall and/or food frequency questionnaire), medical history (e. g. history of
previous LBW infant, gestational diabetes, preeclampsia, and complications during
delivery), the mother’s birthweight, confirmation of tobacco use by biomarkers, exercise
during pregnancy, and additional sociodemographic information (e. g. mother’s

employment, planned vs. unplanned pregnancy).

7.2 Growth retardation

In the present study, the predictors of Stunting in the Michigan WIC population

were identified. Due to the lack of other datasets to confirm the results of this study in

116

the low-income population, the researchers propose another approach. A simulation
study will allow the researchers to confirm the maternal predictors identified in the
present study are actually predicting linear grth retardation. Maternal characteristics
outside of the optimal or normal values will be changed to normal and the logistic
regression model will be re-run with the modified dataset. For example, in the modified
dataset all mothers will have normal prepregnancy weight, ideal weight gain, normal iron
status and all women will abstain from tobacco use during the pregnancy. This will allow
researchers to determine if the risk factors associated identified in the present study are
truly contributing to the prediction of stunting.

Pediatric obesity is a growing problem in the US. (Mei et al., 1998; Ogden et al.,
1997). Research has shown that children who obese often become obese adults.
Identification of predictors of obesity among infants and children through research
methods and modeling procedures employed in the present study will provide insight into

the risk factors associated with obesity among Michigan WIC infants and young children.

117

 

APPENDICES

118

$.- ‘QA " n A u

APPENDIX A

Definitions and Standards of Variables

Mother’s age (years): Computed by the CDC from the mother’s date of birth and the
initial visit, coded in continuous and interval scales (< 16, 16-19, 20-29, 30-39, 2 40).
ch/ Ethnicfl: White, not Hispanic; Black, not Hispanic; Hispanic/ Spanish origin;
American Indian; Asian or Pacific Islander categories. Race of the infant/ child was
categorized according to race of the mother.

Education (years): Number of years of school the woman has completed coded in
continuous and interval scales (< 9, 9-11, 12, 13-15, 216).

Marital Status: Married or not married.

Level of prenatal care: Month of pregnancy the woman began prenatal medical care, not

 

visits to a W1C clinic.

Iron status: 1) Recorded at the initial and postpartum visits to the nearest tenth of a gram
per 100 milliliters for hemoglobin and to the nearest tenth of a percent for hematocrit.
Either hemoglobin or hematocrit (but not both) is needed at each visit. 2) The CDC
created a low iron status (anemic) variable, after applying the criteria to determine
anemia based on the woman’s hemoglobin and hematocrit measurements. The Healthy
People 2000 Objective (Objective 1) defines iron deficiency as having abnormal results
for two of the following tests: mean corpuscular volume, erythrocyte protoporhyrin, and
transferrin saturation, none of which measurements are available in PNSS and PedNSS.

(Table 33 .CDC cut-off values for hemoglobin and hematocrit.)

119

it

tea -—s'n" mu“

Table 33. Hemoglobin and hematocrit cut-off values for anemia in pregnant and
postpartum women.

 

 

Trimester Hemoglobin (g/100ml) Hematocrit (%)
Prenatal

First 1 1 33

Second 10.5 32

Third I 1 33
Post-partum 12 36

 

Reference: Recommendations to Prevent and Control Iron Deficiency. MMWR 47 (RR-
3); 1-36. Publication date: 4/3/98

Mother’s heigh_t: Measured without shoes, and is recorded to the nearest tenth of a
centimeter.

Mother’s weigm: Measured with minimal clothing, without shoes and recorded to the
nearest tenth of a kilogram. The pre-pregnancy weight was either self-reported or
estimated from a measurement during early first trimester of pregnancy.

Mother’s BMI: Calculated (kg /m2) from the height and weight at prepregnancy, initial
and postpartum visits. Criteria according to the CDC (1995): very underweight (BMI
10.0 — 17.9); underweight (BMI 18.0 - 19.7); normal weight (BMI 19.8 - 26.0);
overweight (BMI 26.1 — 29.0) and very overweight (BMI 29.1-74.9). The Healthy People
2000 Objective goal defined overweight (BMI > 27 .3) for women aged 20 years and
older.

Weight gain or loss during pregnancy: Self-reported and recorded to the nearest pounds
gained or lost during the current pregnancy. The CDC adopted the Institute of
Medicine’s (1990) standards of minimum recommended weight gain based on the

prepregnancy weight status: 28-40 pounds for underweight women (BMI g 19.70); 25-35

120

pounds for normal and overweight women, (BMI 19.8 — 29.0); 15-25 pounds for very
overweight women (Blvfl 2 29.1).

Tobacco use: Self-reported average number of cigarettes smoked during: 1) three
months prior to pregnancy, 2) third trimester of pregnancy, and 3) early post-partum
period. The present study measured smoking during pregnancy by self-reported number
of cigarettes smoked in third trimester of pregnancy.

Alcohol consumption: Self-reported alcohol use prior to, during and after pregnancy is
measured by frequency (the average number of days a week the woman drank beer, wine
or liquor) and the quantity (the average number of drinks of beer, wine or liquor the
woman drank each day that she did drink). The CDC recorded the number of drinks per
week. The present Study measured alcohol use during pregnancy by the self-reported
number of drinks per week in the third trimester.

Breastfeeding at postpartum visit: Infant breastfed at the time the woman attends
postpartum visit.

fiant ever breastfed: This variable is used only for the women who responded “no” to
the variable regarding whether the infant was breastfed at the time of the postpartum
visit. The variable is recorded as the age of the infant when breastfeeding was
completely stopped.

Other milk: Age infant began any use of formula or milk other than breastmilk.

Migrant status: Self-reported migrant status or not.

Previous pregnancies: Times the woman has been pregnant before (regardless of the

 

birth outcome), not counting the current pregnancy.

12]

“he “maxim

ugh-2:1??? 0'

Previous live births: Times the woman has given birth, not counting the current
pregnancy. Multiple births are counted as one.

Infant numeric ID: Unique identification code for each infant assigned from clinic
records.

Infants sex: Boy or girl.

Number of infants: Number of infants resulting from the current pregnancy.

Infants’ status at birth: Born dead or alive. Stillbirths and miscarriages are coded as dead
at birth.

lnfants’ status at postpartum visit: Alive or dead at the time of the postpartum visit. If
dead, the age at death is recorded.

Birthweight: Recorded to the nearest whole gram; obtained from the mother’s self report,
hospital records, birth or death certificate. The Michigan 1995 PNSS dataset does not
specify the source of this information.

Gestational age: CDC calculated from a self-reported date of last menstrual period and
the infant’s date of birth.

Household income: Self-reported monthly income during the current pregnancy.
Household size: Number of people living in the household during the current pregnancy
Food stamp participation: Participated in the USDA Food Stamp program during and
after the pregnancy.

Medicaid participation: Participated in Medicaid during and after the pregnancy.

AFDC participation: Received payments from AFDC or welfare during and after the

pregnancy.

122

Height-for-age percentile: Height-for-age is a growth index derived from the reference

curve. It can be expressed as a percentile value or a z-score (standard deviation value).

123

WIT-‘5‘“ fir—3'

APPENDIX B

_o>o_ moqu 3 “gamma—ma *

 

 

2.0. m3 sod mod 8.0 86 8.0 8.? m3 cod So @Ewafiea
mod- 86 8.0- cod Ed :5. mod 8.0- cod .3 asexmavmoaoam
mod- 85 cod- 8.? 3o 8.? 25. 3o 86. Seismic?
no :3 one cod- :3 cod :3. Rd 385 22.88:
3o 2.0 2.0 8.? cod cod 3o avoeooaizaoz
So So 8d mod 8.0- one Eases ozzsosoca
8o 36 So So- .3 Emoeaeawoaaaosoa
8d cod So 2.0 Aaaobaoaaosom
8.0 cod cod 983383 33282
Reno:
006 Ed Em 55.23505
56. A933 owe 383830
@893 own {2:62
:30 Eon o>= .woa
wEonm 85 8:0: 0885 58m 505 53823. .28? ~35 <0 ow< a 9:830;

 

ESE—ova 50363 356508 cows—8.80

124

_o>o_ 3.on 3 Ema-«Ema ...

 

 

1.8.? some- to; 1.3.0- .86. ...:.o .53 1.8.6 use- Amtawmofiea
1.8.? 48¢ 86 48¢ 8.0 .5? Se an; 81.28:2885:2828Bess:
91.8an
1.8.? So- ...8.o- So- cod 1.86 18.0- So- .muase.Teocoaaéaaeocobaeaz
amoe- toS- 8.6 :3 1.8.0 .33. suoa._nab aofiafiwaeoeaoem
cone ammo 8.0 .63- ...ooe- and 818.785 Basia on”?
33 Se smoo- Lee- 1.86 Aouoe._ua©§ao§a cacao:
So some- ...8.o- some are: snoobcaaeomeaa aeae coca
So Lod- ood Grease 2125:3235
seed .6;- auoa ._nmobnasa auto:
*3 .o Sue: 4.1-mob mafia ESE:
Sumac—=0
4.1-25%;.“ 8: SEE @6650 Bond
mafia 35¢:— wE—ooom 953m
so: .02 -385 on"? 28:52 econ xom .852 ESE: con-m x5. @563on

 

.moEmtg 58:66:. 5253 356508 coca—850

125

BIBLIOGRAPHY

126

Bibliography

Abel EL, Hannigan JH. ‘J-Shaped’ relationship between drinking during pregnancy and
birthweight: reanalysis of prospective epidemiological data. Alcohol. 1995;30:345-355.

Abrams B. Preventing low birth weight: does WIC work? A review of the evaluations of
the Special Supplemental Food Program for Women Infants and Children. Ann NY Acad
Sci. 1993;678:306-316.

Abramson R. Cultural sensitivity in the promotion of breastfeeding. NAA COG Clin
Issues. 1992:717-722.

Adair LS, Guilkey DK. Age-specific determinants of stunting in Filipino children J
Nutr. 1997; 127:314-320.

Ahluwalia LB, Hogan VK, Grummer-Strawn L, Colville W Peterson A. The effects of
WIC participation on small-for-gestational age births: Michigan, 1992. Am J Public
Health. 1998;88:1374-1377.

Alexander GR, Cornely DA. Prenatal care utilization: its measurement and relationship
to pregnancy outcome. Am J Prev Med. 1987 ;3 2243-253.

Alexander GIL Howell E. Preventing preterm birth and increasing access to prenatal
care: two important but distinct national goals. Am J Prev Med. 1997;13:290-291.

Allen LH, Lung’aho MS, Shaheen M, Harrison GG, Neumann C, Kirksey A. Maternal
body mass index and pregnancy outcome in the Nutrition Collaborative Research Support
Program. EuroJ Clin Nutr. 1994;48(Supp13): S68-S77.

Allen LH. Nutritional influences on linear growth: a general review. Euro J Clin Nutr.
1994;48(Suppl 1):S75-S89.

Arlotti JP, Cotrell BH, Lee SH, Curtin JJ. Breastfeeding among low-income women with
and without peer support. J Community Health Nurs. 1998;15:163-178.

Avruch S, Cackley AP. Savings achieved by giving WIC benefits to women prenatally.
Public Health Rep. 1995;] 10:27.

Baldwin LM, Larson EH, Connell FA, Nordlund D, Cain KC, Cawthon ML, Byrns P,
Rosenblatt RA. The effects of expanding Medicaid prenatal services on birth outcomes.
Am J Public Health. 1998;88:1623-1629.

Binkin NJ, Yip R, Trowbridge FL. The relationship between birthweight and subsequent
childhood growth. Pediatrics. 1988;82:828-834.

127

Borel M], Smith SM, Derr J, Beard JL. Day-to-day variations in iron status indices in
healthy men and women. Am J Clin Nutr. 1991;54:729-735.

Borges G, Lopez-Cervantes M, Medina-Mora ME, Tapia-Conyer R, Garrido F. Alcohol
consumption, low birthweight, and preterm delivery 1n the National Addiction Survey
(Mexico). IntJ Addict 1993 ,228 355- 368.

Boutry M, Needlman R Use of diet history in the screening of iron deficiency.
Pediatrics. 1996;98:1 138-1142

Bracero LA, Byrne DW. Optimal maternal weight gain during singleton pregnancy.
Gynecol Obstet Invest. 1998; 4629-16.

Buescher PA, Larson LC, Nelson MD, Lenihan AJ. Prenatal WIC participation can
reduce low birthweight and newborn medical costs: a cost benefit analysis of WIC
participation in North Carolina. J Am Dietet Assoc. 1993;93:163-166.

Byrd TL, Mullen PD, Selwyn BJ, Lorimar R. Initiation of prenatal care by low-income
Hispanic women in Houston. Public Health Rep. 1996;] 11:536-540.

Caldwell K. Reaching the goals of “Healthy People 2000” regarding breastfeeding. Clin
Perinatal. 1999;26:527-537.

Carriaga MT, Skikne BS, Finley B, Cutler B, Cook JD. Serum transferrin receptor for
the detection of iron deficiency in pregnancy. Am J Clin Nutr. 1991 ;54: 1077-1081.

Carroll TP. Substantially increasing breastfeeding: an accomplishment of the Alabama
WIC program. J Hum Lact. 1994;10:129-130.

Caulfield LE, Witter FR, Stoltzfiis RJ. Determinants of gestational weight gain outside

of recommended range among black and white women. Obstet Gynecol. 1996;87:760-
766.

Centers for Disease Control and Prevention (CDC), National Center for Health Statistics
(NCHS), National Vital Statistics System (NV SS). Births: final data for 1998. National
Vital Statistics Reports. 2000;48:1-99.

Chandra RK and Saraya AK. Impaired immunocompetence associated with iron
deficiency. J Pediatr. 1975;86:899-902.

Cnattingus S, Haglund B. Decreasing smoking prevalence during pregnancy in Sweden:
the effect on small-for-gestational-age births. Am J Public Health. 1997;87:410-413.

Cogswell ME, Yip R. The influence of fetal and maternal factors on the distribution of
birthweight. Semin Perinatal. 1995;19:222-240.

128

Cogswell ME, Scanlon KS, Fein SB, Schieve LA. Medically advised, mother’s personal
target and actual weight gain during pregnancy. Obstet Gynecol. 1999;94:616-622.

Copper RL, DuBard MB, Goldenberg RL. The relationship of maternal attitude toward
weight gain during pregnancy and low birthweight. Obstet Gyncel. 1995; 85:590-595.

Crane SS, Wojowycz MA, Dye TD, Aubry RH, Artal R Association between pre-
pregnancy obesity and the risk cesarean delivery. Obstet Gynecol. 1997;89:213-216.

Crooks DL. Child growth and nutritional status in a high poverty community in eastern
Kentucky. Am J Phys Anthropol. 1999; 109:129-142.

Dallman PR. Iron deficiency and the immune response. Am J Clin Nutr. 1987;46:329-
334.

Dallman PR, Looker AC, Johnson CL, Carroll M. Influence of age on laboratory criteria

for the diagnosis of iron deficiency anemia and iron deficiency in infants and children.
In: Hallberg L, Asp NG, Eds. Iron nutrition in health and disease. London, UK. John
Libby and Co. 1996:65-74.

Day NL, Richardson GA, Geva D, Robles N. Alcohol, marijuana and tobacco: effects of
prenatal exposure on offspring growth and morphology at age six. Alcohol Clin Exp Res.
1994;18:786-794.

Devaney BL, Ellwood MR, Love JM. Programs that mitigate the effects of poverty on
children. Future Child. 1997;7z88-112.

Diaz 8, Herreros C, Aravena R, Casado ME, Reyes MV Schiappacasse V. Breast-
feeding duration and growth of fully breast-fed infants in a poor urban Chilean
population. Am J Clin Nutr. 1995;62:371-376.

Edwards LE, Hellerstedt WL, Alton IR, Story M, Himes JH.. Pregnancy complications

and birth outcomes in obese and normal-weight women: effect of gestational weight
change. Obstet Gynecol. 1996;87:389-394.

Ellard GA, Johnstone FD, Prescott RJ, Ji-Xian W, Jian-Hua M. Smoking during
pregnancy: the dose dependence of birthweight deficits. Brit J Obstet Gynaecol.
1996;103 :806-813.

Executive Summary Michigan WIC Program. Michigan Department of Community
Health, 1995.

Executive Summary Michigan WIC Program. Michigan Department of Community
Health, 1996.

129

Faden VB, Graubard BI, Dufour M. The relationship of drinking and birth outcome in a
US. national sample of expectant mothers. Paediatri Perinat Epidemiol. 1997;11:167-
1 80.

Falkner F, Holzgreve W, Schloo RH. Prenatal influences on postnatal growth: overview
of pointers for needed research. Euro J Clin Nutr. 1994; 48 (Suppl 1): S15-S24.

Fiscella K. Does prenatal care improve birth outcomes? A critical review. Obstet
Gynecol. 1995;85:468-479.

Fried PA, Watkinson B, Gray R. Grth from birth to early adolescence in offspring
prenatally exposed to cigarettes and marijuana. Neurotoxicol Teratol. 1999;21:513-525.

Frongillo EA, deOnis M, Hanson KMP. Socioeconomic and demographic factors are
associated with worldwide patterns of stunting and wasting of children. J Nutr. 1997;
127: 2302-2309.

Goldman AS. The immune system of human milk: antimicrobial, antiflammatory, and
irnmunomodulating properties. Pediatr Infect Dis J. 1993;12:64-71.

Handler A, Rosenberg D. Improving pregnancy outcomes: public versus private care for
urban, low-income women. Birth. 1992;19: 123-130.

Hellerstedt WL, Himes JH, Story M, Alton IR, Edwards LE The effects of cigarette
smoking and gestational weight change on birth outcomes in obese and normal weight
women. Am J Public Health. 1997;87:591-596.

Hickey CA, Civer SP, McNeal SF, Hoffman HJ, Goldenberg RL. Prenatal weight gain
patterns and spontaneous preterm birth among nonobese black and white women. Obstet
Gynecol. 1995; 85: 909-914.

Higgins AC, Pencharz PB, Strawbridge IE, Maughan GB, Moxley JE. Maternal
haemoglobin changes and their relationship to infant birthweight in mothers receiving a
program of nutritional assessment and rehabilitation. Nutr Res. 1982;2:641-649.

Homan RK, Korenbrot CC. Explaining variation in birth outcomes of Medicaid-eligible
women with variation in the adequacy of prenatal support services. Med Care.
1998;36:190-201.

‘Horta BL, Victora CG, et al. Low birthweight, preterm births and intrauterine grth
retardation in relation to maternal smoking. Paediatri Perinat Epidemiol. 1997;11:140-
151.

bHorta BL, Victora CG, Menezes AN, Barros FC. Environmental tobacco smoke and
breastfeeding duration. Am J Epidemiol. 1997;146:128-133.

130

Hoyert DL. Medical and life-style risk factors affecting fetal mortality, 1989-90. Vital-
Health-Stat-Z 0-Data-Natl- Vital-Stat-Syst. 1996; 3 1:1-32.

Hueston WJ, Mainous AGD, Farrell JB. A cost-benefit analysis of smoking cessation
programs during the first trimester of pregnancy for the prevention of low birthweight. J
Fam Pract. 1994;39:353-357.

Institute of Medicine. Nutrition during pregnancy: part, weight gain, part II, nutrient
supplements.Washington DC: National Academy Press, 1990.

Interagency Workshop. 1991 Statement on the immediate need of breastfeeding
promotion. J Hum Lact. 1991;7 :22.

J edrychowski W, Whyatt RM, et al. Exposure misclassification error in Studies on
prenatal effects of tobacco smoking in pregnancy and the birth weight of children. J
Expo Anal Environ Epidemiol. 1998;82347-357.

Johnson JW C, Longmate JA, Frentzen B. Excessive maternal weight and pregnancy
outcome. AmJ Obstet Gynecol. 1992; 167: 353-372.

Johnson JW C, Yancey MK. A critique of the new recommendations for weight gain
during pregnancy. Am J Obstet Gynecol. 1996; 174:254-258.

Kaminski M, Rumeau C, et al. Alcohol consumption in pregnant women and the
outcome of pregnancy. Alcohol Clinical Exp Res. 1978;2:155-163.

Karlberg J, J alil F, Lam B, Low L, Yeung CY. Linear growth retardation in relation to
the three phases of growth Euro J Clin Nutr. 1994; 48(Suppl 1): $25-$44.

Kaskutas A. Interpretation of risk: the use of scientific information in the development
of the alcohol warning label policy. Int J Addict. 1995;30:1519-1540.

Keppel KG, Taffel SM. Pregnancy-related weight gain and retention: Implications of
1990 Institute of Medicine guidelines. Am J Public Health. 1993; 83:1100-1103.

Kim H. Support of breastfeeding through telephone counseling in Korea. J Hum Lact.
1997;13:29-32.

Kirchengast S, Hartmann B. Maternal prepregnancy weight status and pregnancy weight
gain as major determinants for newborn weight and size. Annals Hum Biol. 1998; 25: 17-
28.

Kitay DZ, Harbort RA. Iron and folic acid deficiency in pregnancy. Clin Perinatal.
1975;22255-273.

131

Klebanoff MA, Shiono PH, Shelby JV, Trachtenberg AI, Graubard BI. Anemia and
spontaneous preterm birth. Am J Obstet Gynecol. 1991;164:59-63.

Kogan MD, Martin JA, Alexander GR, Kotelchuck M, Ventura SJ, Frigoletto FD. The
changing pattern of prenatal care utilization in the United States, 1981-1995, using
different prenatal care indices. JAMA 1998;279:1623-1628.

Kotelchuck M. The adequacy of prenatal care utilization index: its US distribution and
Association with Low Birthweight. Am J Public Health. 1994;84: 1486-1489.

Lehmann F, Gray-Donald K, Mongeon M, Di Tommaso S. Iron deficiency anemia in one
year-old children of disadvantaged families in Montreal. Can Med Assoc J.
1992;146:1571-1577.

Lewis CT, Matthew T], Heuser RL. Prenatal care in the United States, 1980-1994. Vital
Health Stat 21. 1996;54:1-17.

Long DG, Funk-Archuleta MA, Geiger CJ, Mozar AJ, Heins JN. Peer counselor program
increases breastfeeding rates in Utah Native American WIC population. J Hum Lact.
1995;11:279-284.

Long SH, Marquis MS. The effect of Florida’s Medicaid eligibility expansion for
pregnant women. Am J Public Health. 1998;88:371-376.

Looker AC, Dallman PR, Carroll MD, Gunter EW, Johnson EL. Prevalence of iron
deficiency in the United States JAMA 1997;277:973-976.

Lowe N. Breastfeeding information and support services offered by Melbourne hospitals
in antenatal classes. Breastfeed Rev. 1998:6223-28.

Lozofi‘ B, Jiminez E, Wolf AW. Long term developmental outcome of infants with iron
deficiency. New EnglJ Med. 1991;325:687-694.

Lu ZM, Goldenberg RL, Cliver SP, Cutter G, Blankson M. The relationship between
maternal hematocrit and pregnancy outcome. Obstet Gynecol. 1991;77:190-194.

Lucas A, Morley R, Cole T, Lister G, Leesan-Payne C. Breastmilk and subsequent
intelligence quotient in children born preterm. Lancet. 1990;339:261-2644.

Lundsberg LS, Bracken MB, et al. Low to moderate gestational alcohol use and
intrauterine growth retardation, low birthweight, and preterm delivery. Ann Epidemiol.
1997;7z498-508.

Mainous AGD, Hueston W]. The effect of smoking cessation during pregnancy on
preterm delivery and low birthweight. J Fam Pract. 1994; 38:262-266.

132

Marbury MC, Linn S, Monson R, Schoenbaum S, Stubblefield PG, Ryan K]. The
association of alcohol consumption with outcome of pregnancy. Am J Public Health.
1983;73:1165-1168.

Mei Z, Scanlon KS, Grummer-Strawn LM, Freedman DS, Yip R, Trowbridge FL.
Increasing prevalence of overweight among US. low-income preschool children: The

Centers for Disease Control and Prevention Pediatric Nutrition Surveillance, 1983 to
1995. Pediatr. 1998; 101(1):E12.

Meis PJ, Goldenberg RL, Mercer BM, Iams JD, Moawad AH, Miodovnik M, Menard
MK, Caritis SN, Thurnau GK Bottoms SF, Das A, Roberts JM, McNellis D. The
preterm prediction study: risk factors for indicated preterm birth. Am J Obstet Gynecol.
1998;178:562-567.

Melnikow J, Alemagno S. Adequacy of prenatal care among inner-city women. J Fam
Pract. 1993;37:575-582.

Mendez MA, Adair LS. Severity and timing of stunting in the first two years of life
affect performance on cognitive tests in late childhood. J Nutr. 1999;129:1555-1562.

Mestecky J. Immunology of milk and the neonate. Plenum Press. 1991.

Miller IE, Korenman S. Poverty and children’s nutritional status in the United States.
Am J Epidemiol. 1994; 1402233-243.

Miller V, Swaney G, Deinard AS. Impact of the WIC program on the iron status of
infants. Pediatrics. 1985;75:100-105.

Misra DP, Guyer B. Benefits and limitations of prenatal care. From counting visits to
measuring content. JAll/IA. 1998;279:1661-1662.

MMWR. Medical-care expenditures attributable to cigarette smoking during
pregnancy—United States, 1995. Morb Mortal Wkly Rep. 11/171997;46:1048-1050.

MMWR. Recommendations to prevent and control iron deficiency in the United States.
Morb Mortal Wkly Rep 04/03/1998 47 (RR-3); 1-36.

Montgomery DL, Splett PL. Economic benefit of breastfeeding infants enrolled in W1C.
J Am Dietet Assoc. 1997;97:379-385.

Moss N, Carver K. The effect of WIC and Medicaid on infant mortality in the United
States. Am J Public Health. 1998;88:353-361.

National Academy of Sciences. Preventing low birthweight. National Academy Press.
Washington DC: 1985.

133

National Center for Health Statistics. Healthy People 2000 Review, 1998-1999.
Hyattsville, Maryland: Public Health Service 1999.

Nordentoft M, Lou HC, Hansen D, Nim J, Pyrds O, Rubin P, Hemmingsen R.
Intrauterine grth retardation and premature delivery: the influence of maternal
smoking and psychosocial factors. Am J Pub Health. 1996;86:347-3 54.

Ogden CL, Troiano RP, Briefel RR, Kuczmarski RJ, Flegal KM, Johnson CL.
Prevalence of overweight among preschool children in the United States, 1971-1994.
Pediatr. 1997,99:E1.

Owen AL, Owen GM. Twenty years on WIC : A review of some effects of the program.
J Am Dietet Assoc. 1997, 97: 777-82.

Ozki FA. Iron deficiency in infancy and childhood. N Engl J Med. 1993;329:190-193.

Parazzini F, Davoli E, et al. Validity of self-reported smoking habits in pregnancy: a
saliva cotinine analysis. Acta Obstet Gynecol Scand. 1996;75:352-354.

Parker JD, Abrams B. Prenatal weight gain advice: an examination of the recent prenatal
weight gain recommendations of the Institute of Medicine. Obstet Gynecol. 1992, 79:
664-669.

Peacock JL, Bland JM, Anderson HR. Effects on birthweight of alcohol and caffeine
consumption in smoking women. J Epidemiol Comm Health. 1991;45:159-163.

Peacock JL, Bland JM, Anderson HR Preterm delivery: effects of socioeconomic
factors, psychological stress, smoking, alcohol, and cafieme. Brit Med J. 1995;311:531-
535

PedNSS User Manual. United States Department of Health and Human Services, Public
Health Service. 1994.

Perry GS, Yip R, Zyrkowski C. Nutritional risk factors among low-income pregnant US.

women: The Centers for Disease Control and Prevention (CDC) Pregnancy Nutrition
Surveillance System. Semin Perinatol. 1995;19:211-221.

Pizarro F, Yip R, Dallman PR, Olivares M, Hetrampf E, Walter T. Iron status with
different infant feeding methods: relevance to screening and prevention of iron
deficiency. J Pediatr. 1991;118:687-692.

Public Health Service. Healthy People 2000: national health promotion and disease

prevention objectives. Washington DC: Government Printing Office; 1991. Report No.:

(PHS) 91-50212.

134

 

We fl'fi'fi

Pugin E, Valdes V, Labbok MH, Perez A, Aravena R. Does prenatal breastfeeding skills
group education increase effectiveness of a comprehensive breastfeeding promotion
program? J Hum Lact. 1996;12:15-19. ‘

Puolakka J, J anne O, mGo R Serum ferritin in the diagnosis of anemia during
pregnancy. Acta Obstet Gynecol Scand Suppl. 1980;95:57-63.

Raj VK, Plitchta SB. The role of social support in breastfeeding promotion: a literature
review. J Hum Lact. 1998;14:41-45.

Ray WA, Mitchel EF, Piper JM. Effect of Medicaid expansions on preterm birth. Am J
Prev Med. 1997, 13 :292-297.

Reifsnider E, Eckhart D. Prenatal breastfeeding education: its effect on breastfeeding
among WIC participants. J Hum Lact. 1997;13:121-125.

Riordan JM. The cost of not breastfeeding: a commentary. J Hum Lact 1997, 13:93-97.

Rosenberg KD. Letter of the editor regarding benefits and limitations of prenatal care.
JAA/IA. 1998;280.

Rowland, ML. Self-reported weight and height. Am J Clin Nutr. 1990;52:1125-1133.

Rusia U, Madan N, Agarwal N, Sikka M, Sood SK. Effect of maternal iron deficiency
anaemia on foetal outcome. Indian J Pathol Microbiol. 1995;38:273-279.

Russell CM, Williamson DF, Byers T. Can the year 2000 objective for reducing
overweight in the United States be reached?: a simulation study of the required changes
in body weight. Int J Obes Relat Metab Disord. 1995,19: 149-153.

Schafer E, Vogel MK, Viegas S, Hausafus C. Volunteer peer counselors increase
breastfeeding duration among rural low-income women. Birth. 1998,25: 101-6.

Schieve LA, Perry GS, Cogswell ME, Scanlon KS, Rosenberg D, Carmichael S, Ferre C.

Validity of self-reported delivery weight: an analysis of the 1988 National Maternal and
Infant Health Survey. Am J Epidemiol. 1999;150:947-956.

Scholl TO, Heidiger ML, Fischer R1, Shearer JW. Anemia vs iron deficiency: increased
risk of preterm delivery in a prospective study. Am J Clin Nutr. 1992;55:985-958.

Scholl TO, Heidiger ML. Anemia and iron deficiency anemia: compilation of data on
pregnancy outcome. Am J Clin Nutr. 1994;59:4925-501 S

Schramm WF. Weighing costs and benefits of adequate prenatal care for 12,023 births in
Missouri’s Medicaid Program, 1988. Public Health Rep. 1992;107:647-652.

135

Secker-Walker RH, Vacek PM, Flynn BS, Mead PB. Smoking in pregnancy, exhaled
carbon monoxide, and birthweight. Obstet Gynecol. 1997;89:648-653.

Secker-Walker RH, Vacek PM, Flynn BS, Mead PB. Estimated gains in birthweight
associated with reductions in smoking during pregnancy. J Reprod Med. 1998,43 :967-
974.

Selvin S, Abrams B. Analysing the relationship between maternal weight gain and
birthweight: exploration of four statistical issues. Paediatr Perinatal Epidemiol. 1996,
10: 220-234.

Shu X0, Hatch MC, Mills J Clemens J, Susser M.. Maternal smoking, alcohol drinking,
caffeine consumption and fetal growth: results from a prospective study. Epidemiol.
1995,62115-120.

Siega-Riz AM, Adair LS, Hobel CJ. Maternal underweight and inadequate rate of ewight

gain during the third trimester of pregnancy increases risk of preterm delivery. J Nutr.
1996; 126: 146-153.

Siega-Riz AM, Hobel CJ. Predictors of poor maternal weight gain from baseline
anthropometric, psychosocial, and demographic information in a Hispanic population. J
Am Dietet Assoc. 1997, 97: 1264-1268.

Skuse D, Reilly S, Wolke D. Psychosocial adversity and growth during infancy. Euro J
Clin Nutr. 1994;48(Supp1 1): $113-$130.

Slade HB, Schwartz SA. Mucosal immunity: the immunology of breastmilk. J Allergy
Clin Immun. 1987, 80:348-356.

Sokol RJ, Clarren SK. Guidelines use of alcohol terminology describing the impact of
prenatal alcohol on offspring. Alcohol Clin Exp Res. 1989;13:597-598.

Stein AD, Ravelli ACJ, Lumey LH. Famine, third-trimester pregnancy weight gain, and
intrauterine growth: The Dutch famine birth cohort study. Hum Biol. 1995; 67: 135-150.

Strauss RS, Dietz WH. Low maternal weight gain in the second or third trimester
increases the risk of intrauterine growth retardation. J Nutr. 1999, 129: 988-993.

Summers L, Price RA. Preconception care. An opportunity to maximize health in
pregnancy. J Nurse Midwif. 1993;38:188-194.

Tuttle CR, Dewey KG. Potential cost savings for Medi-Cal, Food Stamps, and WIC

programs associated with increasing breastfeeding among low-income Hmong women in
California. J Am Dietet Assoc. 1996;96:885-890.

136

Vazquez-Scone P, Windom R, Pearson HA. Disappearance of iron deficiency anemia in
a high risk population given supplemental iron. N Engl J Med. 1985;313:1239-1240.

Verkerk PH, van Noord-Zaadstra BM, Florey CD, et al. The effect of moderate maternal
alcohol consumption on birthweight and gestational age in low risk population. Early
Hum Dev. 1993;32:121-129.

Walter T, DeAndraca I, Chaduel P, Perales CG. Iron deficiency anemia: adverse effects
on infants. Pediatrics. 1989;84:7-17.

Wang X, Tager IB, VanVunakis H, Speizer FE, Hanrahan JP. Maternal smoking during
pregnancy, urine cotinine concentrations, and birth outcomes. A prospective cohort
study. IntJEpidemiol. 1997;26:978-988.

Webb MO, Ellerbee SM. Breastfeeding in Arkansas: the role of the Arkansas
Department of Health. J Ark Med Soc. 1996,93: 185-187.

Weitzman MS, Gortmaker S, Sobol A. Maternal smoking and behavior problems of
children. Pediatrics. 1992;90:342-349.

Wen SW, Goldenberg RL, Cutter GR, Hoffman J, Cliver SP, Davis R0, DuBard MB.
Smoking, maternal age, fetal growth, and gestational age at delivery. Am J Obstet
Gynecol. 1990;162:53-58.

Windham GC, Eaton A, et al. Evidence for an association between environmental
tobacco smoke exposure and birthweight: a meta-analysis and new data. Paediatr
Perinat Epidemiol. 1999,13 :35-57.

Wilcox LS and J S Marks (eds). 1994. From data to action. CDC ’s public health
surveillance for women, infants, and children. US Department of Health and Human
Services, Public Health Service, Centers for Disease Control and Prevention.

Witter FR, Caulfield LE, Stoltzfus RJ. Influence of maternal anthropometric status and
birth weight on the risk of cesarean delivery. Obstet Gynecol. 1995, 85: 947-951.

Wright SP, Mitchell EA, Thompson JM, et al. Risk factors for preterm birth: a New
Zealand Study. New Zealand J Med. 1998;111:14-16.

Yip R, Binkin NJ , Fleshood L, Trowbridge FL. Declining prevalence of anemia among
low-income children in the United States. JM. 1987 ;258: 1619-1623.

Yip R, Parvanta I, Scanlon K, Borland EW, Russell CM, Trowbridge FL. Pediatric
Nutrition Surveillance System - United States, 1980-1991.11/fll/1WR. 44(55-7): 1-24.1992

Zaren B, Lindmark G, Gebre-Medhin M. Maternal smoking and body composition of the
newborn. Acta Paediatr. 1996;85:213-219.

137

World Wide Web Resources

Centers for Disease Control. National Center for Chronic Disease Prevention and Health
Promotion. The Pediatric Nutrition Surveillance System. [Online] Available
www.cdc.gov/nccdphp/dnpa/PedNSS.htrn , February 28, 2000.

Centers for Disease Control. National Center for Chronic Disease Prevention and Health
Promotion. The Pregnancy Nutrition Surveillance System. [Online] Available
www.cdc.gov/nccdphp/dnpa/PNSS.htm, February 27, 2000.

Discriminant Analysis [Online] Available
www2.chass.ncsu.edngarson/pa765/discrimhtm, March 7, 2000.

Logistic Regression. [Online] Available
www2.chassncsu.edu/garson/pa765/logistic.htm, March 7, 2000.

Michigan Department of Community Health. 1998 Michigan Resident Birth File.
[Online] Available
www.mdmh.state.mi.us, May 19, 2000.

National Center for Health Statistics. Healthy People 2000. [Online] Available
www.health.gov/healthypeople/default.htm, February 28, 2000.

United States Department of Agriculture, Women, Infants, and Children. [Online]

Available
www.fns.usda.gov/wic, March 29, 2000.

138

“lllllll'lllllllf