Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
The causal-effect of childhood obesity on asthma in young and adolescent children
(USC Thesis Other)
The causal-effect of childhood obesity on asthma in young and adolescent children
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
THE CAUSAL-EFFECT OF CHILDHOOD OBESITY
ON ASTHMA IN YOUNG AND ADOLESCENT CHILDREN
by
Janice Kayla Chung
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PHARMACEUTICAL ECONOMICS AND POLICY)
August 2013
Copyright 2013 Janice Kayla Chung
ii
Dedication
To my parents for their unwavering support and encouragement over the years, you have
been an inspiration to me throughout my life. A very special thank you goes out to Tom
for your emotional support and your patience through the months of writing. You have
realized my potential and encouraged me to chase my dreams.
iii
Acknowledgements
I would like to express my deepest gratitude to my dissertation committee:
committee chair, Dr. Michael Nichol, Dr. Neeraj Sood, and Dr. Jeffrey McCombs for
their support, guidance, inspiration and upmost patience. Special appreciation is
extended to Dr. Stanley Azen and Dr. Geoffrey Joyce for their expert and valuable
guidance and encouragement. Without the guidance, expertise and persistent
encouragement from the faculty members at the Department of Pharmaceutical
Economics and Policy, this dissertation would not have been possible. I thank the
Department of Education for making the data available and extending the permission to
disclose the findings as my dissertation.
iv
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables vi
List of Figures viii
Abstract ix
1 Introduction 1
2 Background 9
2.1 Childhood Obesity 9
2.1.1 Prevalence of Childhood Obesity 9
2.1.2 Potential Risk Factors for Childhood Obesity 11
2.1.3 Comorbidities Associated with Childhood Obesity 15
2.1.4 Social and Economic Effects of Childhood Obesity 17
2.1.5 Externalities Associated with Childhood Obesity 18
2.1.6 Anti-Childhood Obesity Policy 20
2.1.7 Classification of Childhood Obesity 21
2.2 Asthma 23
2.2.1 Prevalence of Asthma 23
2.2.2 Potential Risk Factors for Childhood Asthma 24
2.3 Studies on Childhood Obesity and Asthma 28
3 Study Design and Methods 32
3.1 Study Design and Data 32
3.1.1 Observational Retrospective Data Analysis 32
3.1.2 Data Source 32
3.1.3 Study Population 36
3.2 Study Variables 37
3.2.1 Dependent Variable 37
3.2.2 Independent Variable 37
3.3 Cross-Sectional Contemporaneous Association between Childhood
Obesity and Asthma 40
3.3.1 Empirical Model of Childhood Obesity and Asthma 40
3.3.2 Marginal Effect of Childhood Obesity and Asthma 41
3.4 Cross-Sectional Lagged Association between Childhood Obesity
and Asthma 42
v
3.4.1 Empirical Model of Lagged Childhood Obesity and Asthma 42
3.4.2 Marginal Effect of Lagged Childhood Obesity on Asthma 43
3.5 Causal Effect of Childhood Obesity on Asthma 43
3.5.1 Empirical Model of Childhood Obesity on Asthma 43
3.5.2 Potential Endogeneity Bias 44
3.5.3 Instrumental Variable Models 46
3.5.4 Obesity Elasticity on Asthma and Attributable Fraction 55
4 Results 57
4.1 Trends in Childhood Obesity and Asthma 58
4.1.1 Changes in Mean BMI and Prevalence of Childhood
Obesity and Asthma 58
4.2 Baseline Descriptive Statistics 64
4.3 Cross-Sectional Contemporaneous Association between Childhood
Obesity and Asthma 67
4.3.1 Probit Model 67
4.3.2 Probit Models for Subgroup Analyses 70
4.4 Cross-Sectional Lagged Association between Childhood Obesity
and Asthma 75
4.4.1 Probit Model 75
4.5 Causal Effect of Childhood Obesity and Asthma 78
4.5.1 Validity Tests for Instrumental Variables 78
4.5.2 Instrumental Variable Bivariate Probit Model 85
4.5.3 Obesity Elasticity on Asthma and Attributable Fraction 92
5 Discussions and Conclusions 93
5.1 Summary of Findings 93
5.2 Study Limitations 96
5.3 Policy Implications 98
5.4 Conclusion 101
Bibliography 103
vi
List of Tables
Table 1: Change in Prevalence of Asthma in 3
rd
Grade Non-Asthmatic
Children 62
Table 2: Change in Prevalence of Asthma in 3
rd
and 5
th
Grade Non-
Asthmatic Children 62
Table 3: Baseline Demographic Characteristics (3
rd
Grade) 65
Table 4: Weight-Adjusted Probit Estimates for Contemporaneous
Effects of Children 69
Table 5: Weight-Adjusted Probit Estimates for Contemporaneous
Effects of BMI on Developing Asthma among Obese and
Non-Obese (Normal Weight) Children 72
Table 6: Weight-Adjusted Probit Estimates for Contemporaneous
Effects of BMI on Developing Asthma among Obese Children 73
Table 7: Weight-Adjusted Probit Estimates for Contemporaneous
Effects of BMI on Developing Asthma among Non-Obese
(Normal Weight) 74
Table 8: Weight-Adjusted Probit Estimates for Lagged Obesity
Effect on Developing Asthma 77
Table 9: Tested IVs and Results for Relevancy and Exocgeneity 80
Table 10: Results from Different Combination of Instruments using
IV Probit Model 81
Table 11: Observable Characteristics across IV1 (Price Index for Fresh
Healthy Food) for 3
rd
Grade Children 83
Table 12: Relationship between Air Pollutants and IV1 (Price Index
for Fresh Healthy Food) 84
Table 13: Results from Instrumental Variable Model for 3
rd
Grade
Children 86
vii
Table 14: Results from Instrumental Variable Model for 5
th
Grade
Children 87
Table 15: Results from Instrumental Variable Model for 8
th
Grade
Children 88
Table 16: Marginal Effects from Instrumental Variable Model 89
Table 17: Marginal Effects from Instrumental Variable Model
using Different Obesity BMI Threshold Level 91
viii
List of Figures
Figure 1: Cohort Diagram 58
Figure 2: Increase in Body Mass Index (BMI) Over Time (All Children) 59
Figure 3: Prevalence of Overweight and Obesity Over Time (All Children) 60
Figure 4: Prevalence of Asthma Over Time (All Children) 60
Figure 5: Change in Prevalence of Asthma in 3
rd
Grade Non-Asthmatic
Children (Only Children with all 3 years of Complete
Asthma Data) 63
Figure 6: Change in Prevalence of Asthma by Weight Group
(All Children) 63
ix
Abstract
OBJECTIVES: (1) To examine the trends of childhood obesity and changes in
body mass index (BMI) over time, (2) To analyze the contemporaneous/lagged
association between obesity and asthma, and (3) To analyze the contemporaneous/lagged
causal effect of obesity on asthma by implementing instrumental variable method to
control for endogeneity bias.
Methods: A Cohort of young and adolescent children was identified from the
Early Childhood Longitudinal Study Kindergarten Class of 1998-99 (ECLS-K) data from
the U.S. Department of Education. A secondary data source on food prices has been used
to identify valid instruments for obesity to reflect the price variations across different
food groups within specific geographical unit or region. Changes in BMI obtained from
the actual measured height and weight and self-reported asthma status were analyzed
from a longitudinal data from 3
rd
to 8
th
grade for an associative as well as a causal effect
using bivariate probit models.
RESULTS: Two price indices for healthy and unhealthy food were identified as
valid IVs. The marginal effects for childhood obesity were consistent through all grade
levels and different BMI thresholds. The results indicated that childhood obesity
increased the probability of being diagnosed with asthma and the harmful effects of
childhood obesity on asthma increased as children got older. Obese children in 3
rd
grade
had a higher probability of being asthmatic where childhood obesity increased the
probability of asthma by 0.10 percentage points while holding other covariates constant
x
and using a standard 95% BMI threshold was used to define childhood obesity. Similarly,
being obese in 5
th
and 8
th
grade increased the probability by 0.14 and 0.19 percentage
points. There was an increase of 0.02 in marginal effect for obesity with using a 97
th
percentile BMI threshold for 3
rd
grade children while holding other covariates constant at
their mean. In 5
th
and 8
th
grade children, the marginal effects of obesity on asthma were
increased by 4% and 3%, respectively by increasing the BMI threshold from 95
th
percentile to 97
th
percentile. Attributable fraction of obesity on asthma which is the same
as the absolute impact of obesity on asthma in our study showed that 13.0% of the
increase in asthma between 3
rd
and 5
th
grade were explained by the increase in obesity
while 18.4% of the increase in asthma between 5
th
and 8
th
grade were explained by the
increase in obesity.
CONCLUSION: This paper demonstrated that childhood obesity was
endogenously determined with asthma. Gradual increases in mean BMI and prevalence
of childhood obesity along with prevalence of asthma were observed in this study.
Children who were exposed to obesity at any point in time during their kindergarten years
were at higher risk for developing asthma compared to those children who were never
obese during the same time. This may warrant not only a contemporaneous effect of
childhood obesity on asthma but also a lagged effect of obesity status on developing
asthma later in life. This study focused on analyzing the causality of childhood obesity
on asthma while controlling for not only observable but also unobservable factors using
the instrumental variables approach. Through this approach, the study determined that
childhood obesity significantly increases the risk for developing asthma especially in
xi
boys and African American children. We suspected that the instrumental variables
approach was a superior method to investigate a causal relationship between childhood
obesity and asthma over panel analyses due to infrequent changes in one’s weight or
obesity and asthma status.
1
Chapter 1 Introduction
1.1 Introduction
Childhood obesity has become a serious epidemic and continues to be one of the
fastest growing public health concerns in the United States. The prevalence of overall
obesity has increased drastically across all age groups, races and gender but showing
especially alarming rates for children (Hedley, Ogden et al. 2004, Ogden, Carroll et al.
2006, Flegal, Carroll et al. 2010). The prevalence of obesity has doubled in adults aged
20 years or older between 1980 and 2002 while the obesity rate has tripled for children
and adolescents aged 6 to 19 years. Obesity has been well-documented as an association
and a leading cause of numerous chronic conditions including cardiovascular disease,
type 2 diabetes mellitus (DM), hypertension, stroke, dyslipidemia, osteoarthritis and
some cancers (FX 1991, Pi-Sunyer 1991, Must, Spadano et al. 1999, Mokdad, Ford et al.
2003, Han, Lawlor et al. 2010). Obesity in children can also lead to increased risk for
conditions including cardiovascular, pulmonary, orthopedic, gastrointestinal, endocrine
and psychological conditions and may also decrease life-expectancy as an adult (Reilly,
Methven et al. 2003, Bibbins-Domingo, Coxson et al. 2007).
The current definition and cutoffs for obesity in the United States have been
developed from the 2000 age and sex-specific growth chart from Center for Disease
Control and Prevention (CDC). Obesity is defined as elevated body adiposity
composition and body mass index (BMI) is one of the most widely used tools to measure
fat composition given one’s height and weight. BMI is expressed as weight in kilograms
2
divided by height in meters squared (kg/m²). Based on the recommendations of expert
committees, children with BMI values at or above the 95
th
percentile and between 85% to
95% of the 2000 age and sex-specific growth chart from CDC are categorized as obese
and overweight respectively. Although BMI is known to be a less sensitive measure
compared to the other measures such as skinfold thickness, BMI has been the leading
indicator to classify weight groups in obesity studies related to both children and adult
populations (Dietz and Robinson 1998).
There has been a lot of interest in developing effective treatments for obesity,
with the overwhelming majority of the research focused on adult population. The most
frequently used and acceptable treatment and management approaches to children and
adolescent obesity include life style change with dietary modifications, increased
physical activity levels and reduced sedentary behavior (Epstein, Myers et al. 1998,
Barlow and Dietz 2002). Despite the high levels of relapses, this modest and
conservative approach towards the fight against obesity has been shown to be highly
effective over more drastic and infrequently used methods such as surgical interventions,
prescription or over-the-counter weight loss drugs (Luttikhuis, Baur et al. 2009). Early
identification by clinician has also been positively associated with treating childhood
obesity in addition to life style changes, however many children with obesity are not
correctly identified per CDC definition of obesity due to lack of attention towards obesity
(Barlow and Dietz 2002). Motivation, time available for counseling, self-efficacy, access
to competent providers and lack of reimbursement to the provider play important and
interrelated roles as barriers in treatment of childhood obesity.
3
The dramatic increase in children and adolescent obesity is an alarming concern
in both clinical and policy sense due to both short-term and long-term negative health
consequences associated with obesity. The effect of health consequences of obesity has
been well documented and indicates strong association with several acute and chronic
diseases such as hypertension, dyslipidemia, insulin resistance, type 2 diabetes, coronary
heart disease, stroke, gall bladder disease, osteoarthritis, cancer, sleep apnea and
respiratory problems (Reilly, Methven et al. 2003, Bibbins-Domingo, Coxson et al. 2007).
Additionally, obesity is also known to be a risk factor for sustaining an extremity fracture
requiring operative intervention in pediatric trauma patients and it is also associated with
higher risk for certain complications including deep venous thrombosis and decubitus
ulcers (Rana, Michalsky et al. 2009). Obesity during adolescence has also shown
important social and economic consequences where people with obesity were more likely
to have fewer years of school, less likely to be married, have lower household incomes
with higher rates of household poverty (Gortmaker, Must et al. 1993). Additionally,
significant wage discrimination against obesity has also been observed for obese adult
working population when compared to normal weight group (Bhattacharya and Bundorf
2009).
Recent studies have documented the impact of obesity on annual medical
expenditures among adults. Obese adults aged 18 to 65 years of age had 36% higher
average annual medical expenditures compared with those of normal weight (Roland
2002). Furthermore, aggregate obesity-related medical expenditures accounted for 5.3%
of adult medical expenditures in the United States from 1996 to 1999 of which 50% of
4
these expenditures are financed by Medicare and Medicaid (Finkelstein, Fiebelkorn et al.
2003). Annual U.S. obesity-attributable medical expenditures are reported at $75 billion
in 2003 and California state-level estimates at $7.7 billion (Finkelstein, Fiebelkorn et al.
2004). Several studies have shown similar patterns of economic burden of obesity in
children and adolescents. The proportion of discharges with obesity-related diseases has
increased dramatically which led to a significant growth in economic costs. Obesity-
attributable annual hospital costs have tripled from $35 million during 1979-1981 to $127
million during 1997-1999 in children and adolescent populations (Wang and Dietz 2002).
Without dramatic changes in successfully screening, preventing and treating obesity, the
health care costs attributable to obesity is projected to be greater than $860 billion by
2030, accounting for approximately 20% of projected total U.S. health-care costs (Wang,
Beydoun et al. 2008).
The increasing trend of childhood obesity is not only a concern clinically due to
both the potential short and long-term obesity-related clinical complications but also
economically. The economic burden of childhood obesity has a negative impact on both
the individuals and the society including the government and tax payers. Although the
prevalence of childhood obesity along with obesity-related co-morbidities has increased
substantially, the current health care system lacks the ability to comprehensively treat
these children. The scope of the treatment is based on managing the symptoms rather
than treating the disease and there is no standard of care treatment available for these
children. The system also lacks in correctly identifying children who are at risk to refer
them to certain special form of care to prevent rapid progression of disease. Insurance
5
companies have little incentives to invest in prevention of obesity-related comorbidities
that may not be apparent for many years to come. Lack of evidence-based prevention
and treatment for obesity and the state of the current health care system combined with
economic downturn, physicians and healthcare system have limited motivation to
improve screening and specialty care infrastructures to handle childhood obese cases.
Therefore understanding the implications of early childhood obesity through studying the
interrelationship between childhood obesity and co-morbidities is a priority in measuring
the scope and the severity of the problem before developing solutions. Such an
understanding is crucial to raise public awareness about the severity of childhood obesity
as well as to increase awareness of needs for public policy to successfully manage
childhood obesity problems.
The physiological interactions between obesity and asthma have been investigated
in adults by numerous studies which demonstrated that pathogenic features such as lung
function and chronic inflammation may be associated with obesity (Giamila 2005,
Cottrell, Neal et al. 2010). In obese individuals, the lungs were under-expanded and the
sizes of breaths were smaller. They also had greater possibility of low-grade chronic
inflammation that could lead to narrow airways and had higher blood levels of certain
hormones from fat tissue which can lead to even more inflammation. In a study by
Cottrell et al., abnormal lipid and glucose metabolism beyond BMI was significantly
associated with asthma (Cottrell, Neal et al. 2010). However, currently no research is
known to have been conducted to demonstrate the causal effect of obesity in predicting
developing asthma in children. The study by Kajbaf addressed a strong association
6
between asthma symptoms and both overweight and obesity in both sexes among school-
age children (Kajbaf, Asar et al. 2011). Other studies have reported similar conclusions
on the association between asthma and obesity or BMI in children (Zhang, Lai et al. 2010,
Murray, Canoy et al. 2011). Systematic review conducted by Camilo et al. based on
databases such as PubMed has shown various studies on growing prevalence of asthma
and obesity but no causal effect relationships have been established between them
(Camilo, Ribeiro et al. 2010). Results from the study by Wang and Dietz examining the
trend of obesity-associated diseases in youths and related economic costs have indicated
asthma to be the most prevalent principal diagnosis listed for hospital discharges where
obesity has been listed as a secondary diagnosis (Wang and Dietz 2002). Michelson et al.
conducted a study using the National health and Nutrition Examination Survey to
examine the association between obesity and asthma severity in children and adolescents
and found significant association with obesity and asthma severity (Michelson, Williams
et al. 2009).
A major challenge with investigating the causal relationship between child obesity
and asthma is endogeneity. This may be a result of three reasons including simultaneity,
omitted variable bias and measurement error. First, there may be simultaneity where
child obesity may affect asthma, it is also possible that asthma may affect child obesity.
Second, there may be omitted variable bias where unobserved factors which affect both
obesity and asthma were excluded from the regression model. Third, there may be
measurement error if obesity measures were based on self-reported height and weight or
measured by using uncalibrated machines. In the presence of endogeneity, regression
7
models are biased and yield unreliable estimates of the causal effect of obesity on asthma.
It is possible to remove endogeneity bias implementing an instrumental variables (IV)
approach which uses instruments that are correlated with obesity but uncorrelated with
unobservable factors affecting asthma to isolate the variation in obesity that is
independent of asthma. The endogeneity bias is removed by examining the impact of the
exogenous variation in child obesity on asthma but to date, no studies have used this
approach to determine the relationship between child obesity and asthma.
The purpose of this research is to examine the various aspects of childhood
obesity, utilizing the data from the Early Childhood Longitudinal Study-Kindergarten
Class (ECLS-K), a nationally representative dataset for the U.S. that followed a cohort of
kindergarten children from the 1998-1999 school year.
The specific aims for the study are to:
1) Examine the trends of childhood obesity and changes in BMI over time
2) Analyze the contemporaneous/lagged association between obesity and asthma
3) Analyze the contemporaneous/lagged causal effect of obesity on asthma by
implementing IV method
Because of the public health importance and the increasing trends in both child obesity
and asthma, the causal effect relationship between child obesity and asthma should be
determined. Understanding the true relationship between the two is an important step in
identifying priorities in child obesity research and to prioritize in implementing policies
to manage child obesity epidemic. The study findings can be used as an aid for
policymakers in developing more effective screening, prevention and treatment programs
8
along with better health care infrastructure for children early on rather than dealing with
much greater costs of treating obesity-attributable diseases throughout their adulthood.
9
Chapter 2 Background
2.1 Childhood Obesity
2.1.1 Prevalence of Childhood Obesity
World Health Organization (WHO) defines obesity as abnormal or excessive fat
accumulation that presents a risk to health. The prevalence of obesity is considered an
important public health issue in the United States. Healthy People 2010 which provides a
framework for prevention by promoting a set of health objectives for the nation to
achieve over the first decade of the new century has identified obesity as 1 of the 10
leading health indicators. The National Health and Nutrition Examination Survey
(NHANES) has been the leading data source for monitoring the national prevalence of
obesity for children and adults since 1960 in the United States. Adults are defined as
persons aged 20 years or older and children as persons aged 2 through 19 years. Adult
population are categorized into weight class (normal, overweight, obesity, extreme
obesity) according to their BMI of under 25.0, 25.0 to 29.9, 30.0 or more, and 40.0 or
more, respectively. In general, the 2000 centers for Disease Control and Prevention
(CDC) Growth Charts for the United States are used in order to define weight classes for
children.
The NHANES database offer alarming statistics showing substantial increases in
the prevalence of overweight and obesity from pre-school children to adolescence. These
increases have been noted in all racial and ethnic groups while some groups are affected
10
more compared to others. Among preschool children aged 2 to 5, obesity has increased
from 5.0% to 10.4% between 1976-1980 and 1999-2000 and from 6.5% to 19.6% among
those aged 6 to 11. Among adolescents aged 12 to 19, obesity has increased from 5.0%
to 18.1% during the same period (Flegal, Carroll et al. 2010). The data showed
significant racial and ethnic disparities in obesity prevalence among U.S. children and
adolescents where the prevalence of obesity was significantly higher among Mexican-
American adolescent boys (26.8%) than among non-Hispanic white adolescent boys
(16.7%) in 2007-2008. Among the girls during the same time period, non-Hispanic black
adolescents (29.2%) were significantly more likely to be obese compared with non-
Hispanic white adolescents (14.5%).
Studies by CDC demonstrate state-level geographic variability of obesity across
the nation, showing the highest obesity prevalence in California, South Dakota and Texas
(CDC 2009). In 2008, 1 of 7 low-income preschool-aged children was obese and the
prevalence of obesity for children of 2 to 4 year olds in low-income household has
increased from 12.4% in 1998 to 14.6% in 2008. Among these study population, obesity
prevalence was highest among American Indian or Alaska Native (21.2%) and Hispanic
children (18.5%) and lowest among white (12.6%) and Asian or Pacific Islander (12.3%)
in 2008. CDC has listed obesity as one of the “Winnable Battles” which are CDC’s few
targeted areas for public health priorities to address the leading causes of death and
disability. Obesity has been chosen based on the magnitude of the immediate and future
health problems and a possibility for significant improvements in health outcomes
through effective strategies.
11
In 1999-2002, the percentage of adults at a healthy weight (33.0%) was
approximately half of the Healthy People 2010 target level of 60% and the prevalence of
obesity among adults (30.4%) was double the target prevalence (15%). Similarly, the
data during the same period showed the prevalence rate for children’s obesity being far
from reaching its goal of Healthy People 2010 target level.
2.1.2 Potential Risk Factors for Childhood Obesity
There are several potential causes of childhood obesity including a variety of
environmental factors such as healthy food choices and physical activity in different
environment settings from children’s home, school as well as the community. Drinks
that are high in refined carbohydrates with added sugar and fat or energy-dense
foods/snacks have become more readily available for children on and outside of school
campuses. Children have constant access to sugar drinks and unhealthy foods at school
and outside of school from vending machines, convenient stores and fast food restaurants.
The children’s food environment state indicator report from year 2011 by the CDC noted
carbonated drinks as the largest source of added sugar intake and an important
contributor of calories in the diets of children in the U.S (Center for Disease Control,
2011). And high consumption of sugar drinks has already repeatedly been shown as a
significant association with childhood obesity (Vartanian, Schwartz et al. 2007).
Nationwide advertising of these sugary drinks and less healthy foods that are high in total
calories, sugars, salt and fat also play a main role in determining children’s ability or lack
thereof to make healthy food choices.
12
Although the total calories and sugar intake by children have steadily increased
over the period, the amount of daily physical activity has decreased drastically due to
more sedentary lifestyle from highly developed technologies including video and
computer games and smart phones. Physical activity guidelines advisory committee
report by the U.S. Department of Health and Human Services indicated that most young
children and adolescents fell short of the physical activity guidelines recommendation of
at least 60 minutes of aerobic physical activity per day (U.S. Department of Health and
Human Services, 2008). Below 20% of students in grades 9 through 12 have met such
recommendation in 2007 and approximately 30% of students attended daily physical
education classes at school. Rideout et al. studied the changes in media use patterns for
children and young adults from 8 to 18 years of age. Different mediums of media use
including watching television, playing video games, spending time on computers and
smartphones were analyzed across different age groups as well as gender and races. The
report showed an average of 7.5 hours per day using entertainment media where more
than half of these 7.5 hours were spent on watching television. Several studies analyzed
the effects of television viewing on childhood obesity and found significant positive
association between the television viewing and childhood obesity where children
substantially increased energy intake through snacking and eating meals in front of the
television while spending considerably less of their time on physical activities (Robinson
2001, Connor 2006, Zimmerman and Bell 2010).
Other potential risk factors associated with childhood obesity include biological,
genetic predisposition to infant feeding practices and certain socioeconomic
13
characteristics or household environment, such as maternal age at childbirth, maternal
employment status, family type or parent’s marriage status, poverty and race. A study by
Gibbs and Forste examined the relationship between the breast feeding practices, the
early introduction of solid foods and other maternal characteristics such as BMI, age at
birth and smoking with early childhood obesity, using a nationally representative
longitudinal survey of early childhood. The results showed that all of the maternal
characteristics including the BMI, age at birth and smoking were positively associated
with childhood obesity. In addition, those children who were introduced to solid foods
before they became 4 months-old were at significantly greater risk for childhood obesity.
Children who had only been fed formula were at the highest risk for childhood obesity
when compared to children who were fed only breast milk or in combination of breast
milk and formula.
Several researches have found maternal employment to be associated with an
increased risk of childhood obesity. Anderson et al. used matched mother-child data
from the National Longitudinal Survey of Youth to determine a causal relationship
between maternal employment status and childhood weight problems while controlling
for observable and unobservable differences (Anderson, Butcher et al. 2003). The results
indicated that a child was more likely to develop weight problem if his/her mother was
employed and the risks for childhood obesity increased as the hours employed per week
over the child’s life increased. The causal effect of maternal employment on childhood
weight problems was greatest among higher socioeconomic status mothers. Cawley and
Liu found similar results of positive correlation between maternal employment status and
14
childhood obesity using the American Time Use Survey (Cawley and Liu 2012). The
authors reported that employed mother’s spend significantly less time preparing for meals,
eating with their children and playing with their children while they were more likely to
purchase fast-food or prepared foods. Results from a similar study in the U.K. setting
suggested that maternal employment had greater detrimental effect on early childhood
weight problems than any other socioeconomic status such as poverty or lack of money
while there was no evident significant relationship between employment status of the
partner and childhood weight problems. These study findings provided evidence that
recent change in patterns of family life due to maternal employment leads to lack of
young children’s access to healthy foods and physical activity.
Aside from the maternal employment, other home/family environment such as
parental status has also been evaluated by various researchers as one of many
contributing factors to childhood obesity. Using the National Health and Nutrition
Examination Survey, Huffman et al. evaluated the association between parental status
and children’s health outcomes including eating habits, body weight and blood
cholesterol (Huffman, Kanikireddy et al. 2010). Children from single-parent households
were at greater risk of being obese than children of dual-parent households and also had
greater total calorie and saturated fatty acid intakes. Additional studies implied a strong
relationship between family structure of single-parent families vs. the two-parent families
and weight problems in children (Gibson, Byrne et al. 2007, Gable, S et al. 2000).
15
2.1.3 Comorbidities Associated with Childhood Obesity
There are two levels of clinical consequences of child obesity: one that is more
short term that has an immediate effect during childhood and the other is more long term
which predisposes them to serious ailments in their later life as adults. Short-term
consequences of obesity are well documented in both chronic and acute medical settings
which include conditions such as cardiovascular diseases, type 2 diabetes mellitus and
psychological problems.
One of the most well established medical consequences of childhood obesity is
the effect on cardiovascular disease. Studies consistently showed associations between
obesity and most of the major cardiovascular risk factors including high blood pressure,
dyslipidaemia, abnormalities in left ventricular mass and function, abnormalities in
endothelial function as well as insulin resistance (Freedman, Dietz et al. 1999, Tounian,
Aggoun et al. 2001, Sorof and Daniels 2002, Reilly, Methven et al. 2003). Freedman et
al. using the study population of 5 to 10 year olds from U.S. reported significant odds
ratios for raised diastolic blood pressure (OR=2.4), raised systolic blood pressure
(OR=4.5), raised LDL cholesterol (OR=3.0), low HDL cholesterol (OR=3.4), raised
triglycerides (OR=7.1), and high fasting insulin concentration (OR=12.1) for obese
children. It is fairly well validated by numerous studies that childhood obesity has
adverse effects on cardiovascular system which are similar to those well known among
adult population (Must, Spadano et al. 1999).
16
Various studies demonstrated the evidence of a positive association between
asthma and obesity in children. Figueroa-Munoz et al. reported significant association
between asthma and BMI for children aged 4 to 11 in the U.K (Figueroa-Muñoz, Chinn
et al. 2001). The association was stronger in girls than in boys in the inner city sample
regardless of ethnicity. One longitudinal study demonstrated reported that becoming
obese significantly increased risk of developing asthma symptoms in girls who were non-
asthmatic at baseline (Castro-Rodríguez, Holberg et al. 2001).
The long-term medical consequences of obesity have also been summarized in
detail by number of studies. A projection of incidence rate for future adult coronary heart
disease (CHD) has been studied using the historical trends and prevalence of adolescent
obesity in a simulation model (Bibbins-Domingo, Coxson et al. 2007). The authors
project that adolescent overweight is to increase the prevalence of obese 35-years-olds in
2020 to a range of 30 to 37% in men and 34 to 44% in women. Hence increases the
incidence of coronary heart disease as well as the total number of CHD-related deaths in
young adulthood. Reilly and Kelly also provided evidence demonstrating the adverse
consequences of childhood and adolescence obesity on premature mortality and physical
morbidity in adulthood (Reilly and Kelly 2010). The study reported significant
association between obesity and premature mortality as well as an increased risk for later
cardiometabolic morbidities including diabetes, hypertension, ischemic heart disease, and
stroke in adult life. In addition, child obesity was positively associated with later
disability pension, asthma, polycystic ovary syndrome symptoms and certain cancers
including prostate, colorectal and breast cancers. Earlier study done by Reilly
17
demonstrated persistence of child and adolescence obesity into adulthood predisposing
one to a greater risk for obesity-attributable diseases (Reilly, Methven et al. 2003).
2.1.4 Social and Economic Effects of Child Obesity
Childhood obesity has also been known to have social and economic effects on an
individual throughout his or her adulthood (Gortmaker, Must et al. 1993, Reilly, Methven
et al. 2003). For example, a study by Gortmaker et al. demonstrated that the adolescence
obesity had more severe impact on social and economic consequences than those of many
other chronic physical conditions. Men who had been obese were 11% less likely to be
married (p=0.005) and both men and women who were obese during their adolescence
had significantly fewer years of school as well as lower household income ($6,710 less
per year; p<0.001). Childhood obesity was also associated with higher rates of household
poverty by 10% (p<0.001) than non-obese children, independent of their base-line
socioeconomic status and aptitude test scores.
A study by Bhattacharya and Bundorf analyzed the health care costs of obesity
among working adult population utilizing data from the National Longitudinal Survey of
Youth and the Medical Expenditure Panel Survey (Bhattacharya and Bundorf 2009). The
authors concluded that the incremental healthcare costs associated with obesity are
transferred on to obese workers with employer-sponsored health insurance in the form of
lower cash wages. The wage differential was greater for obese female workers where
they were paid $1.66 less hourly wage compared to normal weight female workers while
hourly wage penalty for obese male workers was $1.21 (both p=0.001). However, when
18
the analysis only included individuals with continuous coverage of employer-sponsored
health insurance, the wage offset for obesity disappeared for male workers. The authors
explained the wage differential between obese female and male workers with continuous
coverage of employer-sponsored health insurance as significantly higher total medical
expenditures incurred by obese female workers compared to males. Other studies by
Cawley and Averett and Korenman found similar patterns of wage discrimination against
obese individuals (Cawley 2000; Averett and Korenman 1996). However, the authors
have indicated the reduced productivity from obesity-related illness and pure employer
discrimination as the potential factors in the wage differential between obese and normal
weight groups.
2.1.5 Externalities Associated with Childhood Obesity
It is evident that adulthood and childhood obesity not only affects one’s health but
also deteriorates social, psychological and financial well-beings of individuals. But one
of the main societal concerns for obesity remains unanswered: who pays the healthcare
costs associated with obesity? As Bhattacharya and Bundorf have shown in their study,
the healthcare costs for obese working population with employer-sponsored health
insurance are transferred to obese individuals in a form of lower wage. However, the
magnitudes of healthcare utilizations and costs associated with obesity for uninsured and
children still remain unanswered.
Children and uninsured individuals who are obese may follow different patterns
of healthcare utilizations and expenditures. For example, one may speculate that
19
uninsured obese individuals may choose not to seek any preventive or healthcare services
until obesity associated conditions have already progressed, hence the higher utilizations
of emergency department or inpatient services are often observed among this population.
On the other hand, children are able to receive broader healthcare services because they
are often generously covered by federal or state funded health insurance system such as
Children’s Health Insurance Program (CHIP). Both uninsured individuals and children
who qualify for government or state funded healthcare system transfer the healthcare
costs associated with obesity to tax payers in some moderation, thus creates externality
associated with obesity.
Annual obesity-attributable U.S. medical expenses were estimated at $75 billion
for 2003 and taxpayers funded about half of this amount through Medicare and Medicaid
(National Conference of State Legislatures 2008). The health care cost externalities have
been more likely to be generated from adult weight problems in the past but with the
stark increase in obesity-related health problems in children, these health care cost
externalities associated with childhood obesity are rapidly growing. Therefore, it is of a
great importance to study and understand the short-term and long-term implications
related to childhood obesity to develop efficient tools and policys. The study findings
can aid policy makers as well as clinicians to facilitate healthy eating and a physically
active lifestyle to maintain a healthy weight and to reduce obesity-related chronic
diseases.
20
2.1.6 Anti-Childhood Obesity Policy
More dynamic anti-obesity policies for children have been considered and enacted
by state legislatures within the past decade. In 2005, approximately 17 states including
California have enacted the School Nutrition Legislation to establish nutritional standards,
including food and beverages sold in vending machines (National Conference of State
Legislatures). Other policies considered or enacted include nutrition education, increase
in physical activity, recess or physical education and printing nutrition information on
school menu or labeling. However, these policies have proven to fall short in fight
against childhood obesity. Recent changes in increased awareness of childhood obesity
with a help of the media and prominent political figures and celebrities, the society has
taken a huge leap towards recognizing and taking actions against the problem of
childhood obesity.
The Affordable Care Act also introduced initiatives to address childhood obesity
by providing $25 million in funding the Childhood Obesity Demonstration Project to
develop comprehensive and systematic model for reducing childhood obesity. The
initiatives also call for providing guidance to states and health care providers on
preventive and obesity related services and require each state to design a public
awareness campaign on these services. The future of the anti-obesity policy for children
is unknown due to a recent economic downfall and expected budget cuts in the near
future under the new government administration but the efforts to fight against childhood
obesity need to continue.
21
Obesity has adverse short-term and long-term health consequences that not only
have a direct effect on the individual but also on society through health care and other
costs. The causes of childhood obesity are complex and multidimensional which can
depend on any factors including energy intake and expenditure as well as genetic,
metabolic disorders and global environmental factors. No one unique public policy will
be effective alone to respond to the growing challenges of childhood obesity but a
multifaceted public health policy approach is required to efficiently reduce the current
obesity epidemic. Public policies facilitating healthy diet and exercise while strategically
utilizing media as a way to disseminate information about exercise, health and nutrition
to children should be considered. The root of the childhood obesity epidemic problems
can only be solved with collaborations between children, parents, policy makers,
clinicians, schools, and researchers and I greatly hope that my research can contribute in
shedding some light into the truth about healthcare utilizations and expenditures
associated with childhood obesity.
2.1.7 Classification of Childhood Obesity
According to the CDC, BMI is a simple and one of the most frequently used
measures for adiposity to classify obesity among adults as well as in children. BMI is
expressed as weight in kilograms divided by height in meters squared (kg/m²) and WHO
has defined the BMI cutoffs for obesity classification although some nations such as
Japan and China have modified the definition of obesity. According to the WHO, adults
with a BMI of 30 or greater are considered obese whereas definition of child obesity is
22
evaluated differently with definitions and cutoffs for obesity that have been developed
from the 2000 sex-specific CDC growth charts for the U.S. BMI of male and female
children from ages 2 through 20 are shown in percentile are used to categorize
underweight, healthy weight, overweight, obese and severely obese. Cutoff points for
overweight is BMI for age and gender at or above 85
th
percentile and obese at or above
95
th
percentile. BMI for age and gender above 99
th
percentile is categorized as severely
obese.
Waist circumference and waist-hip ratio are considered another simple method to
measure obesity. They are more specific measures abdominal adipose tissue and fat
distribution at the waist defined as the largest abdominal circumference (Cox and
Whichelow 1996). They have been shown to be a strong predictor of intra-abdominal fat
than the BMI and show a stronger association with cardiovascular risk than BMI in cross
sectional studies. However, it has not been customary to measure waist circumferences
in clinical settings whereas height and weight are measured in clinical settings on a
regular basis. In addition, there is a lack of clear definition for child obese categorization
using waist circumference or waist-hip ratio.
Body fat percentage or content is another indicator to define one’s adiposity.
Body fat percentage is a total body fat expressed as a percentage of total body weight.
Several techniques including near-infrared interactance, dual energy x-ray absorptiometry
and skinfold methods can be used to estimate one’s body fat percentage. However,
similar to waist circumference and waist-hip ratio, there is no generally accepted
23
definition of obesity based on body fat percentage measure. Study by Strain showed
essentially the same correlation between BMI and body fat content, demonstrating that
BMI gives just as an accurate measurement of obesity as the more technically complex
and expensive measurement of body fat content (Strain and Zumoff 1992). Gray and
Fujioka also demonstrated nearly identical relationship between BMI and body fat
measured by underwater weighting (Gray and Fujioka 1991).
42
Although BMI can be calculated quickly and without expensive equipment, BMI
still has its limitations. BMI categories do not take into consideration for varying factors
such as higher muscularity and frame size such as abnormally short or tallness. It also
fails to account for varying proportions of bone and fat (Strain and Zumoff 1992).
Despite all these limitations, BMI continues to be used as an accepted standard measure
to define obesity in children, given the ease of measurement, low cost and high reliability.
2.2 Asthma
2.2.1 Prevalence of Asthma
Asthma is characterized by episodic airflow of obstruction, increased airway
responsiveness, and airway inflammation (American Lung Association, 2011). The
prevalence of asthma among school-age children has been rising in many regions of the
developed world and poses as a growing threat to the health of children (Kajbaf, Asar et
al. 2011). According to the American Lung Association, asthma is considered one of the
most common chronic disorders in childhood which currently affects an estimated 7.1
24
million children under 18 years and is the third leading cause of hospitalization among
children under the age of 15 (American Lung Association, 2011). More than 10% of
youth ages 5 to 17 have been professional diagnosed with asthma and in 2005 the
prevalence rate of child asthma was estimated at 142.7 per 1000 with the higher
prevalence rates for females and African Americans. It is reported that asthma is one of
the leading causes of school absenteeism where in 2008, asthma resulted in an estimated
14.4 million lost school days in children per American Lung Association. Despite the
advances in therapy, asthma prevalence, morbidity and mortality have all been increasing
over the years with few subgroups being highly recognized with more prevalent with
asthma. Specific causes for this epidemic are unknown, although changes in frequency
and severity of early-life infections, diet, and exposure to allergens and air pollutants
have all been linked with asthma.
2.2.2 Potential Risk Factors for Childhood Asthma
Despite the similarities between adult and childhood asthma, the pediatric and
adolescent population presents certain specific characteristics and risk factors associated
with asthma. The subjects of intense research efforts for asthma include potential
etiologies and risk factors associated with childhood asthma including gestational weight,
premature birth, maternal age at child birth, ambient air quality, and etc. Aligne et al.
studied the associative effects of numerous demographic variables with asthma using the
parent-reported surveys from the 1988 National Health Interview Survey (Aligne,
Auinger et al. 2000). The study concluded that age, race, geographical region and
poverty had no significant associative effects on asthma in children from 5 to 17 years of
25
age but other controlling demographic variable such as living in an urban setting showed
an increased risk for asthma. The differences in asthma prevalence between different
race population groups were due to different exposure to environmental factors where
African American may have been disproportionately living in impoverished inner-city
areas where asthma is known to be at its worst. Other variables including whether child
was a male or the mother being younger than 17 years of age at child’s birth were both
positively associated with asthma after controlling for other risk factors. In a study by
von Mitius et al., diagnosis of asthma was more likely for preadolescent children with
lower birth weight or premature birth especially in girls (von Mutius, Nicolai et al. 1993).
In addition, significant decrements were demonstrated for different measurement of lung
function in children with premature births.
Numerous studies have related low birth weight and lower gestational ages to
subsequent asthma in children and adolescents. Lu et at. tested the hypothesis that
children with a history of low birth weight having a greater risk of developing asthma by
using a cohort of 75,871 junior high school students in Taiwan where birth weight and
estimated gestational age were obtained from the birth registry (Lu, Hsieh et al. 2012).
The study results showed that asthma was more prevalent in children with birth weights
below 3,000g especially for the children with lower birth weight and have rapidly gained
weight and considered overweight or obese during early childhood. Another study by
Jeong et al. examined the association of the current BMI at age three and birth weight
with developing chronic respiratory illness including asthma in childhood among
hospital based birth cohorts (Jeong, Jung-Choi et al. 2010). Birth related data were
26
obtained from the birth registries and the presence of respiratory symptoms was evaluated
by using a validated questionnaire for wheezing and asthma for children. The results
indicated that children in the lowest birth weight category at three years of age were at an
increased risk for asthma and in addition, children who were initially in the lowest birth
weight category and gained weight rapidly and now in the highest weight category by the
age of three had higher risk of asthma compared to those who remained in the same
weight category.
The relationships of childhood new-onset asthma with ambient air quality and
traffic-related pollution near homes and schools were evaluated by numerous studies.
The Southern California Children’s Health Study assessed the association between
asthma and traffic-related pollution exposure based on regional ambient ozone, nitrogen
dioxide and particulate matter in 13 communities (McConnell, Islam et al. 2010). During
the 3 years of follow-up of a cohort of kindergarten and 1
st
grade children, both ambient
nitrogen dioxide and ozone were positive contributing risk factors to the development of
asthma. Similar studies by Gehring et al. and Brauer et al. and Gehring et al. assessed the
development of asthma during the first 2 years and 4 years of life among birth cohorts in
Germany and Canada, respectively (Gehring, Cyrys et al. 2002, Brauer, Hoek et al.
2007) . And the authors for both studies concluded that outdoor concentration of traffic-
related air pollutants, measured by nitrogen dioxide was positively associated with
asthma and respiratory infections during both of the first 2 and 4 years of life.
27
Several studies investigated the effects of home environment, mainly the early
exposure to environmental tobacco smoke on the occurrence of chronic respiratory
diseases including childhood asthma (Infante-Rivard 1993, Dong, Cao et al. 2007, Lin,
Gomez et al. 2008). Dong et al. evaluated the effect of first 2 years of environmental
tobacco smoke and current environmental tobacco smoke on respiratory symptoms and
diagnosis of respiratory diseases among kindergarten-aged children using a self-
administered questionnaire by parents of children. The results showed that boys were
more susceptible to environmental tobacco smoke and more likely to suffer from
breathing-related problems and to be diagnosed with asthma with early environmental
tobacco smoke than girls after controlling for other covariates in the analysis. Lin et al.
found similar results for significant environmental risk factors including the presence of
smokers in the household to be positively associated with being diagnosed with asthma
among children 1 to 17 years of age living in Buffalo, New York. A study by Infante-
Rivard excluded the presence of smokers in the home as significant risk factors for first-
time diagnosis of asthma among 3 and 4 year old children, whereas other environmental
risk factors such as humidifier in the child’s room, the presence of an electric heating
system in the home to significantly associated with asthma. A history of pneumonia, the
absence of breast feeding and a family history of asthma were also significantly
associated with asthma. The results from these studies confirm the possibility of role of
susceptibility factors in asthma which show that both indoor and outdoor environmental
factors may contribute to the incidence of asthma.
28
2.3 Studies on Childhood Obesity and Asthma
Both asthma and obesity often begin in childhood and share some common risk
factors. The increased prevalence of these two diseases in the past decade has raised
public health concerns for these two conditions and naturally raised speculation that the
development obesity and asthma might be related. Although causality has never been
determined, a strong association between obesity and asthma in children has been
demonstrated in numerous studies.
Ahmad et al. explored the association between obesity and asthma in U.S.
children and adolescents. Subgroup analysis was performed by different age groups and
the adjusted odds ratio for asthma with obesity was significantly bigger than 1 for
children in the 7-12 and 13-17 year-old age-groups indicating strong association between
two conditions (Ahmad, Biswas et al. 2009). One study of young children under the age
of 9, by Murray et al. investigated the relationship of BMI with allergic disease including
asthma within a birth cohort (Murray et al. 2011). Increasing BMI significantly increased
the risk of physician-diangnosed asthma at age 8 in girl, but not in boys. Another study
of schoolchildren by Kusunoki et al. found a significant association between obesity and
higher prevalence of asthma in girls (p=0.009) (Kusunoki, Morimoto et al. 2008).
Mai et al. evaluated the impact of childhood overweight on the development of
asthma in children born with very low birth weight (Mai, Gäddlin et al. 2005). A birth
cohort of very low birth weight children was followed from birth and investigated on
diagnosis of asthma at 12 year of age and the study results concluded that being
29
overweight at 12 years of age was significantly associated with an increased risk of
current asthma in the very low birth weight children (OR: 5.5). Systematic review by
Camilo et al., demonstrated growing prevalence of child obesity and asthma as well as
strong association using evidence from cross-sectional, case-controls and prospective
cohort studies on obesity and asthma (Camilo, Ribeiro et al. 2010). Another systematic
review conducted by Matricardi et al. showed similar findings of consistently supporting
a link between obesity and asthma diagnosis in children and the association became
stronger in girls than in boys after puberty (Matricardi, Grüber et al. 2007).
Cassol et al. evaluated the relationship between the prevalence of asthma and
obesity in adolescents living in southern region of Brazil in a cross-sectional study
(Cassol, Rizzato et al. 2006). The study showed that there is a positive association
between obesity and prevalence of asthma symptoms as well as asthma severity. The
authors also state that the association was more prominent in females than males. Similar
results were found in a 20-year prospective community study of young adults (Hasler,
Gergen et al. 2006). In this study, obesity was significantly associated with asthma in a
cross-sectional analysis with an odds ratio of 3.9. Multivariate longitudinal analyses
revealed that two conditions were associated among women after controlling for
confounding factors.
Gilliand et al. examined longitudinal data from the Children’s Health Study
residing in the Southern California region to analyze the relationship between childhood
obesity and new-onset asthma among school-age children (Gilliland, Berhane et al. 2003).
Among the children who were asthma-free at enrollment, significantly greater risk of
30
new-onset asthma was observed among obese and overweight children during the five
year follow-up period. Obese and overweight boys showed greater susceptibility to
asthma compared to girls and the effect of being overweight and obese was greater in
nonallergic children than in allergic children.
In a study by Cottrell et al., diagnosis of asthma were more likely than non-
asthmatics to have higher triglyceride levels after controlling for gender differences and
smoke exposure (Cottrell, Neal et al. 2010). Obesity is known to be commonly
accompanied by elevated triglycerides and the study findings uniquely describe a
statistically significant association between metabolic variables related to obesity with
asthma while controlling for smoke exposure which is a frequently publicized risk factor
of asthma. Another study evaluated the role of resistin, which is produced by adipose
tissues on childhood asthma (Kim, Shin et al. 2008). Multiple regression analysis
revealed that resistin was a significant predictive factor for asthma. Both studies by
Cottrell et al. and Kim et al. show strong relationship between clinical indicators for
obesity with diagnosis of asthma. Another anthropometric measures of obesity has been
examined to assess the relationship between obesity and asthma (Musaad, Patterson et al.
2009). BMI and central obesity which was measured using waist circumference were
highly correlated and both measures were significantly associated with the presence of
asthma as well as asthma severity in children. The study results confirmed the previous
findings of studies in adults.
Saha et al. studied urban children aged between 5 and 18 years to identify the
neighborhood socioeconomic and housing factors at the census-block level and the
31
individual level sociodemographic factors that are associated with the risk of asthma
(Saha, Riner et al. 2005). The authors found that the prevalence rate of ever having
asthma was the highest (27.4%) in black boys and none of the census-block
characteristics, such as median age of housing, family income, families with linguistic
isolation, were significant in predicting asthma. Conversely, the authors found age, race,
sex and BMI to be significant predictors for childhood asthma and the association was
stronger in females than males.
To explore the association between obesity and asthma severity in children and
adolescents, Michelson et al. utilized the 2 rounds of National Health and Nutrition
Examination Survey (Michelson, Williams et al. 2009). In multivariable models,
elevated BMI were significantly associated with worse asthma severity. Although the
study did not focus on the relationship between obesity and asthma, the study findings
provide support for the hypothesis that obesity-induced inflammation may contribute to
greater asthma severity.
In summary, wide range of studies with BMI as independent variables and
clinical indicators for obesity, favor the current hypothesis of strong association between
child obesity and asthma although the findings show different strength of the association
by gender and age. Therefore, subsample analysis is strongly warranted to study the true
relationship between the two conditions.
32
Chapter 3 Study Design and Methods
3.1 Study Design and Data
3.1.1 Observational Retrospective Data Analysis
Surveys taken from nationally representative cohorts have been used in many
epidemiological studies to determine trends of specific diseases and to evaluate different
health, mental, and social manifestations associated with specific diseases. Observational
retrospective data analysis approach using the cohort data reflects the true characteristics
of study population from the real world setting with varying degrees of environmental
exposures which may have a significant impact on the outcome of interest. Although
some limitations apply to observational studies, appropriate adjustments on the data
analysis can be made to address the issues and limitations. Observational studies are
considered to be highly useful in studying the detrimental health effects of a disease due
to ethical and practical reasons.
3.1.2 Data Source
The primary data source for this study is the Early Childhood Longitudinal Study
Kindergarten Class of 1998-99 (ECLS-K) data from the U.S. Department of Education.
Co-sponsored by the U.S. Department of Education and Institute of Education Sciences
National Center for Education Statistics (IES), the ECLS-K collected information from a
nationally representative sample of children from kindergarten who were followed
33
through the 8
th
grade. The purpose of the ECLS-K was to provide comprehensive and
reliable data on children’s development and experiences in the elementary and middle
school grades. The multifaceted data collected across the years give opportunities to
researchers and policymakers to study how wide range of child, home, classroom, school,
and community factors at certain points in children’s lives relate to cognitive, social,
emotional, and physical development. Trained evaluators assessed children in their
schools and collected information from parents, teachers and school administrators via
telephone or questionnaires. Information was collected in the fall and the spring of
kindergarten (1998-99), the fall and spring of 1
st
grade (1999-2000), the spring of 3
rd
grade (2002), the spring of 5
th
grade (2004), and the spring of 8
th
grade (2007). Most of
the data is easily accessible via the Department of Education’s website with an exception
of restricted data involving individuals and geographical identifiers must be obtained
through a restricted data license. Unlike other publicly available database, researchers
are able to obtain crucial information such as geographical information including
residence zip codes or school zip codes through licensed restricted data. Unlike studies
that rely on self-reported measured of height and weight which are crucial key variables
to calculate BMI, ECLS-K trained evaluators directly measured children’s height and
weight at each data collection. The primary data source for analysis of assessing the
causal-effect of child obesity on developing asthma will be from years 2002 to 2007 as
information on asthma became available starting in the 3
rd
grade.
The secondary data source for this study to be used to identify valid instruments
for obesity is the Quarterly Food-at-Home Price Database (QFAHPD). As instruments
34
for obesity, food price data reflects the price variations across different food groups
within specific geographical unit or region which provide key information on how readily
available each food groups are within the geographical region at what selling prices. The
first requirement of an instrument is that it is highly correlated with childhood obesity
conditional on the other variables that affect childhood asthma. Evidence suggests that
consumption of certain food groups for example, high energy-dense diets or highly
calorie enriched diets will increase the likelihood of obesity (Vartanian, Schwartz et al.
2007). Furthermore, additional bodies of evidence of inverse relationship between
consumption of normal goods including food and price suggest that food prices across
geographical regions are non-weak predictors of the obesity measures. Higher prices for
healthy food index are expected to have positive impact on childhood obesity while
higher prices for unhealthy food index are expected to have negative impact on childhood
obesity due to lower consumption of unhealthy food resulting from higher price. The
QFAHPD database was developed by the U.S. Department of Agriculture (USDA) to fill
the gap in available food price data and to support research on the economic determinants
of diet quality and health outcomes. It contains regional and market-level quarterly
prices presented in dollars per 100 grams of food as purchased for 52 separate food index
groups from 1999. Food prices are crucial for economic modeling of consumer food
choices, dietary patterns and health-related outcomes such as obesity. Variations in
living costs and other unobserved market conditions across the country may indicate that
analysis using national-level food prices can be inaccurate to capture the true effects of
food prices on consumer’s health-related outcomes. Since QFAHPD measures prices of
35
food index in local settings, it is more suitable in capturing the true effect of prices on
consumer behavior and well-being. QFAHPD is also preferred to an annual measure so
that price fluctuations across seasons and perishable goods including fruits and
vegetables can been correctly measured. Relative quarterly price indices of different
mixtures of food groups composed of fruits, vegetables, soda, sport drinks, and sweets
from the 1
st
quarter of each year will be used as they correspond to the quarter in ECLS-
K data and therefore best describe prices between each wave.
The metropolitan-level data on food prices from the American Chambers of
Commerce Researchers Association (ACCRA) will be utilized in addition to the ECLS-K
and QFAHPD database. The ACCRA has released their data every quarter since 1968
and played a leading role in providing information on cost-of-living differentials in about
311 metropolitan areas in the USA. The ACCRA food price information consists of a
grocery price list containing 63 items, 16 for specific foods at home. Prices of fruits and
vegetables that are widely available around the year (Potatoes, bananas, lettuce, sweet
peas, tomatoes, peaches, frozen corn) will be focused for the analysis. Similar to the
QFHAPD, ACCRA data is also preferred to an annual measure so that price fluctuations
across seasons and perishable goods including fruits and vegetables can been correctly
measured. ACCRA price data from the 1
st
quarter of each year in 2002, 2004 and 2007
will be used as they correspond to the grades in ECLS-K data and therefore best describe
prices between each wave.
In addition to the primary ECLS-K data and the data on food prices, database on
air quality from the American Lung Association and the U.S. Environmental Protection
36
Agency’s Air Quality System (AQS) were used to control for the air quality and air
pollution. Evidence from numerous epidemiological studies suggested that ambient air
quality to be one of the environmental risk factors associated with asthma among children,
therefore it was essential to control for the air qualities around the children’s resident area
in the analysis using a supplemental data. The ambient air quality including the 24-hour
averaged particle pollution (PM2.5) concentration and the daily 8-hour maximum ozone
concentration around the residency of each child were linked with the ECLS-K data.
Geographic variation is substantial for ambient air quality where different pollutants tend
to be high in different localities and it is crucial to control for these relatively different
exposures to air pollutants. Both air data from the American Lung Association and the
AQS data from the 1st quarter of each year in 2002, 2004 and 2007 are used to correctly
correspond to the grades in ECLS-K data.
3.1.3 Study Population
Inclusion/Exclusion Criteria
Cohort of kindergarten children over 1000 schools in the U.S. during the 1998-99
school year whose data were collected by several waves of follow-ups through the 8
th
grade were included in the study. Children with incomplete data on height/weight or
asthma and children with disabilities who require assistance with day-to-day functions
were excluded from the study.
37
3.2 Study Variables
3.2.1 Dependent Variable
The primary dependent variable in the analysis is the probability of ever being
diagnosed with asthma which can only be observed by an indicator variable for having
asthma from the data. The dependent variable took a value one if the child was
diagnosed with asthma by a health professional and zero otherwise. ECLS-K started
collecting data on asthma diagnosis from the parents when children were in the 3
rd
grade
with two additional waves of follow-ups with asthma data.
3.2.2 Independent Variables
Obesity
An indicator variable for obesity was derived by using the BMI from direct height
and weight measures that were not based on self-reported measures which reduces the
likelihood of measurement error in the obesity measures. According to the CDC, BMI is
a simple and one of the most frequently used measures for adiposity to classify obesity
among adults as well as in children. BMI is expressed as follows:
BMI=weight (kg) / (height (m²))
2
WHO has defined the BMI cutoffs for obesity classification although some
nations such as Japan and China have modified the definition of obesity. According to
the WHO, adults with a BMI of 30 or greater are considered obese whereas definition of
38
child obesity is evaluated differently with definitions and cutoffs for obesity that have
been developed from the 2000 sex-specific CDC growth charts for the U.S. BMI of male
and female children from ages 2 through 20 are shown in percentile are used to
categorize underweight, healthy weight, overweight, obese and severely obese. Cutoff
points for overweight is BMI for age and gender at or above 85
th
percentile and obese at
or above 95
th
percentile. BMI for age and gender above 99
th
percentile is categorized as
severely obese.2
Binary indicator variable for obesity in children population followed the 95
th
percentile BMI cutoff guidelines developed form the 2000 sex-specific CDC growth
charts whereas in young adult population followed the strict cutoff guidelines at BMI of
30 or greater. Two measures of obesity were used for the analyses, one being a binary
variable for obesity, taking a value of one if the child’s BMI was at 95
th
percentile or
greater and the actual BMI in continuous variable form. In additional to the obesity
variable, a number of other explanatory variables from the ECLS-K data were included in
the analyses which can be grouped in several categories.
Individual Demographic Variables
In this group, individual characteristic variables such as gender, age (in months),
weight at birth, and race/ethnicity (five categories) were considered. Age and weight at
birth were included as continuous variables.
39
Home and Family Variables
Variables such as household income, poverty, insurance status, maternal
employment, mother’s age at first birth, and mother’s age at birth of the child, family
type and parents’ highest level of education were considered. Information were collected
from parents on household income, measured in 4 categories and on living below poverty
level, insurance status and maternal employment, all measured by indicator variables.
Family type (two-parent family vs. single parent family) was also included in the
analyses as a control variable. Parents’ highest level of education was imputed from the
information obtained by parents on either side of parent’s highest level of education
attainment measured in four categories.
Geographical and Environmental variables
These included region of residence (four categories), size and type of residence
city (three categories) and air pollutant levels measured by ozone index and particle index
(both five categories) or 24-hour averaged particle pollution concentration and daily 8-
hour maximum ozone concentration.
Level of Exercise Variables
Parents were asked questions on their child’s level of aerobic exercise compared
to other children (three categories) and average number of days per week their child was
involved in a rapid exercise. Child’s level of aerobic exercise in comparison of other
children were categorized as having had more, less or about the same level of aerobic
40
exercise compared to the child’s peers in same gender. The average number of days for
rapid exercise in a week was included as a continuous variable.
3.3 Cross-Sectional Contemporaneous Association between
Childhood Obesity and Asthma
3.3.1 Empirical Model of Childhood Obesity and Asthma
Assume that y
i
* represents an index function which measures the propensity for
having been diagnosed with asthma, defined by the regression relationship
y
i
* = x’
1
β
1
+ δ
1
Obesity
i
+ ε
1
(1)
The vector x’
1
represents observed exogenous covariates that determine having
asthma, such as age, gender, insurance status, indicator variables for health status
(diagnosed with diabetes, cancer, etc.) and the study variable, obesity.
What we observe is a dummy variable y
i,
defined by
y
i
= 1 if y
i
* > 0,
y
i
= 0 otherwise
i indexes children and y* is an unobserved latent variable. We define y
i
as an indicator
variable that represents whether child i actually had asthma. Empirically we observe the
binary variable y that takes the value one if the child has been diagnosed with asthma (in
which case y
i
* ≥ 0) and zero otherwise (y
i
* < 0). Variable Obesity is a binary variable
taking the value one if the child is obese and zero otherwise. X’ is a vector of variables, ε
41
is an error term, and both β and δ
are vectors of coefficients where δ indicates the
measure of association effect of childhood obesity on asthma.
From the above relations, we obtain
Prob (y
i
= 1) = Prob(ε
1
> - x’
1
β
1)
= 1 – F(- x’
1
β
1
) (2)
where F is the cumulative distribution function of ε
1
. We assume that the cumulative
distribution of ε
1
is normally distributed with a zero mean and unit variance and this
distributional assumption implies a probit analysis model for regression model (1).
3.3.2 Marginal Effect of Childhood Obesity on Asthma
Marginal effect of childhood obesity on asthma can be computed for the
multivariate probit model (1) by:
Pr(Y
i
= 1) = Φ(x’
1
β
1
+ δ
1
Obesity) (3)
Where Φ is the standard normal distribution function, and Φ(x’
1
β
1
+ δ
1
Obesity) is
computed for each observation using the estimated coefficients from the bivariate probit
model. This framework gives the probability of having been diagnosed with asthma
controlling for the endogeneity of B and Y. Given that Obesity is a dummy variable, i
measures the marginal effect (M.E.) of being obese on the probability of having been
diagnosed with asthma as the sample average of changes in Φ(x’
1
β
1
+ δ
1
Obesity) with
discrete changes in Obesity while keeping all other variables X at their observed values:
42
(4)
3.4 Cross-Sectional Lagged Association between Childhood
Obesity and Asthma
3.4.1 Empirical Model of Lagged Childhood Obesity and Asthma
The model for cross-sectional lagged association between obesity and asthma
follows the similar model as regression model (1) but with a lagged effect of obesity to
analyze if being obese in previous grade level is significantly associated with developing
asthma in current grade level. Assume that y
i
* represents an index function which
measures the propensity for having been diagnosed with asthma at time t, defined by the
regression relationship with a lagged effect of child obesity
y
it
* = x’
1
β
1t
+ δ
1
Obesity
it-1
+ ε
1
(5)
The vector x’
1
still represents observed exogenous covariates for children at time t
in each grade level when the data were collected, whereas the study variable, obesity is
an indicator variable for being obese during the previous grade level from when the data
were collected.
Dummy variable y
i
is defined by
y
it
= 1 if y
it
* > 0,
43
y
it
= 0 otherwise
We define y
i
as an indicator variable that represents whether child i actually had
asthma in time t or currently grade level. We still assume that the cumulative distribution
of ε
1
is normally distributed with a zero mean and unit variance and this distributional
assumption implies a probit analysis model for regression model (5).
3.4.2 Marginal Effect of Lagged Childhood Obesity on Asthma
The marginal predicted probability of having been diagnosed with asthma is given
by similar calculation as equation 4. Given that Obesity is a dummy variable at previous
time period, (t-1), the marginal effect (M.E.) of being obese in time t-1 on the probability
of having been diagnosed with asthma in later time period, (t) as the sample average of
changes in Φ(x’
1
β
1
+ δ
1
Obesity
t-1
) with discrete changes in Obesity at time t-1 while
keeping all other variables X at time t at their observed values.
3.5 Causal Effect of Childhood Obesity on Asthma
3.5.1 Empirical Model of Childhood Obesity on Asthma
Assume that y
i
* represents an index function which measures the propensity for
having been diagnosed with asthma, defined by the regression relationship
y
i
* = x’
1
β
1
+ δ
1
Obesity + ε
1
(5)
44
The vector x’
1
represents observed exogenous covariates that determine having
asthma, such as age, gender, insurance status, indicator variables for health status
(diagnosed with diabetes, cancer, etc.) and the study variable, obesity.
What we observe is a dummy variable y
i,
defined by
y
i
= 1 if y
i
* > 0,
y
i
= 0 otherwise
We define y
i
as an indicator variable that represents whether child i actually had
asthma.
From the above relations, we obtain
Prob (y
i
= 1) = Prob(ε
1
> - x’
1
β
1)
= 1 – F(- x’
1
β
1
) (6)
where F is the cumulative distribution function of ε
1
. We assume that the cumulative
distribution of ε
1
is normally distributed with a zero mean and unit variance and this
distributional assumption implies a probit analysis model for regression model (5).
Obesity is likely to be correlated with the error term, ε
1
in regression model (5) for
asthma; unobservable factors probably affect both obesity and the likelihood of becoming
asthma due to several possible reasons.
3.5.2 Potential Endogeneity Bias
The potential problem of the analysis using the above empirical model for obesity
and asthma via multivariate probit model without correcting for the unobserved
45
differences among obese and non-obese children is the violation of exogeneity
assumption which states that vector of regressors x’s are uncorrelated with the
unobserved determinants of y. The estimates obtained under the exogeneity assumption
are unbiased and consistent, however if any regressors are endogenous, the estimators of
all regression parameters are inconsistent. Primary causes of eodogeneity include
simultaneity bias, omitted variable bias, sample selection bias and measurement error
bias (Cameron and Trivedi, 2005).
Obesity is likely to be correlated with the error term, ε
1
in regression model (4)
for asthma; unobservable factors probably affect both obesity and the likelihood of
becoming asthma due to several possible reasons. First, the direction of obesity and
asthma were ambiguous (reverse causality) which leads to simultaneity bias because it is
believed that child obesity would result in developing asthma but at the same time
children who may be obese may spend more time exercising outside to control their
weight that may cause asthma or respiratory-related problems. Second, omitted variables
such as unobserved individual heterogeneity including eating habits and soda or sugar
drink consumption, access to healthy affordable food and high-energy-dense foods at
home, and also possibly other factors such as genetic predisposition to obesity, biological
mother's weight at birth, biological mother’s age at menarche would all may lead to
endogeneity bias as those variables are unobserved in the data set. Unobserved
heterogeneity or omitted variables relating to obesity from the regression model leads to
endogeneity problem which could result in severe consequences including inconsistent
46
estimates. In the presence of endogeneity, regression models are biased and yield
unreliable estimates of the causal effect of obesity on asthma.
3.5.3 Instrumental Variable Models
Instrumental Variables
When faced with the unobserved heterogenetiy or omitted variable bias, one can
ignore the problem and suffer the consequences of biased and inconsistent estimators, use
proper panel data methods or implement the method of instrumental variables (IV) to
solve the problem of endogeneity of one or more explanatory variables. One can still
generate a consistent estimate of the effect of obesity on asthma if one identifies a set of
IVs indicated as Z that are correlated with obesity but not ε
1
, the error term in the asthma
equation (model (5)). Given Z, one can calculate an IV estimate of the effect of obesity
on asthma. However, consistent estimates can be measured using only the valid IVs.
The key issue of implementing an IV method when faced with endogeneity bias is the
choice and the validity of the instruments. In order to obtain consistent estimators, the
observable variable or the instrument has to satisfy two assumptions: (1) Instrument z
must be related, either positively or negatively to the endogenous explanatory variable,
obesity (relevancy), (2) instrument z has no partial effect on asthma but only related to
asthma through an association of obesity (once regression covariates and the omitted
variables in ε
1
are controlled for) (strict exogeneity). If the instrument fails to meet the
relevancy assumption, instrument becomes a weak instrument and may produce an
inconsistent estimator. In case of failure of strict exogeneity assumption or the
47
“identification” condition for IV estimation, IV becomes an invalid instrument. One
must choose IVs based on strong theoretical assumptions proven by evidence based
literatures and also conduct proper diagnostic methods to test for both relevancy and
strict exogeneity of an instrument. The primary issue with finding valid IVs is the
difficulty of satisfying the second assumption of strict exogeneity.
Instrumental Variable Bivariate Probit Model
The endogeneity of obesity in equation (5) can be solved using IVs in another
probit analysis model (equation (7)) indicating the obesity as a propensity function of the
observed covariates with the valid IVs.
Obesity
i
* = x’
2
β
2
+ δ
1
Z + ε
2
(7)
Obesity
i
= 1 if y
2
* > 0,
= 0 otherwise
Equation (5) and (7) together make a two-equation model called the bivariate probit
model as shown below. In order to control for endogeneity in a biavarate probit
framework requires suitable instruments Z with properties mentioned above for obesity.
The vector x’
1
represents observed exogenous covariates that determine obesity, such as
age, gender and Z represents the IVs.
y
i
* = x’
1
β
1
+ δ
1
Obesity + ε
1
(8)
Obesity
i
* = x’
2
β
2
+ δ
1
Z + ε
2
48
where,
E[ε
1
|x
1
, x
2
] = E[ε
2
|x
1
, x
2
] = 0,
Var[ε
1
|x
1
, x
2
] = Var[ε
2
|x
1
, x
2
] = 1,
Cov[ε
1,
ε
2
|x
1
, x
2
] = ρ.
(ε
1,
ε
2
) have a bivariate normal distribution with means (0,0), variance (1,1) and
correlation ρ which represents the tetrachoric correlation or correlation for binary data. It
is also the correlation that would be measured between the underlying continuous
variables if they could be observed.
Proposed Instrumental Variables
The key issue of implementing an IV method when faced with endogeneity bias is
the choice and the validity of the instruments. In order to obtain consistent estimators,
the observable variable or the instrument has to satisfy two assumptions: (1) Instrument
z must be related, either positively or negatively to the endogenous explanatory variable,
obesity (relevancy), (2) instrument z has no partial effect on asthma but only related to
asthma through an association of obesity (once regression covariates and the omitted
variables in ε
1
are controlled for) (strict exogeneity). If the instrument fails to meet the
relevancy assumption, instrument becomes a weak instrument and may produce an
inconsistent estimator. In case of failure of strict exogeneity assumption or the
“identification” condition for IV estimation, IV becomes an invalid instrument. One
must choose IVs based on strong theoretical assumptions proven by evidence based
49
literatures and also conduct proper diagnostic methods to test for both relevancy and
strict exogeneity of an instrument.
Following are the six proposed instruments for this study:
1) Market-level quarterly prices for food index groups (healthy and unhealthy
food) obtained from QFAHPD database in 1
st
quarter, year 2002, 2004, 2007
2) Metropolitan-level price index for healthy and unhealthy food obtained from
ACCRA data in 1
st
quarter, year 2002, 2004, 2007
The identifying assumption of the proposed instruments is that each instrument is
correlated with obesity and each instrument is not correlated with ε
1
in a regression
model for asthma (equation (4)).
Food prices are crucial for economic modeling of consumer food choice, dietary
patterns and health-related outcomes such as obesity. The different levels of price index
for vegetables and fruits will be used as an instrument because in general, cost of goods,
such as vegetables that are generally healthier and beneficial in maintaining healthier
weight are associated with the demand for those goods. There is a general negative
correlation between the cost and the demand of goods as proven by the theory of
elasticity of demand with respect to price with an exception of few specialty items. Local
prices of food index groups from the QFHAPD and ACCRA may better measure the
effect of prices on food choices than using national-level food prices.
American diets continue to fall short of the recommended consumption levels of
fruits and vegetables. The World Health Organization states that the key to maintaining
50
healthy weight is an abundant consumption of fresh vegetables and fruits with moderate
portion of energy dense food. Several studies have determined the negative relationship
between poor diet consisting of high calories with obesity and that it may also be
necessary to spend more money to achieve a healthy diet and living. The UK women’s
Cohort Study has explored food costs in a study of cohort of 15,191 women and has
indicated that women in the healthiest diet group spent an additional 617 pounds sterling
(≈U.S. $1,000) per year on food relative to the least-healthy diet group (Cade, Upmeier et
al. 1999). The largest amount of cost spent by the healthiest diet group was incurred by
vegetables and fruits.
Adolescent food consumption and weight outcomes
Children are now exposed to caloric dense food more than ever due to increasing
availability of fast food restaurants and commercialization of these fast food restaurants
specifically target children by including small toys with their meals. Children are more
prone to exercise poor dietary behaviors which lead to high prevalence of obesity among
children and adolescents. Numerous factors including genetic, environment,
technological changes are hypothesized as key contributors to childhood obesity.
Emerging body of literature also suggests economic contextual factors such as food
prices to be other main contributors to the obesity epidemic. In general, evidence showed
that prices for healthful foods were higher compared to those less healthful foods.
There is substantial evidence for the influence of food costs on food purchases.
Beydoun et al has found association between higher fruit and vegetable prices with
51
higher BMI (Beydoun, Powell et al. 2008). Dietary variety and the consumption of fresh
produce were generally associated with higher food costs. Individuals who are least
likely to consume healthy diets are more likely to suffer from some of the highest rates of
obesity and type 2 diabetes. Energy-dense diets or highly calorie enriched diets lack
fresh vegetables and fruits but in return, they allow for a higher consumption of food at a
lower cost. Another study assessed the influence of price changes of low-energy-density
and high-energy-density foods on mother’s food purchases and determined that purchases
of both food groups reduced when prices of each groups increased (Epstein, Dearing et al.
2007).
Powell and Chaloupka examined empirical evidence regarding the food price
sensitivity of weight outcomes based on a peer-reviewed literature search from 1990 to
2008 . The studies reviewed by the authors showed statistically significant association
between food price and weight outcome. The effects were generally small although in
some cases the effects were larger for some subpopulation including those in low-
socioeconomic status and who are in risk for overweight or obesity (Powell and
Chaloupka 2009).
Later work by Powell et al. conducted a cross-section quantile regression analysis
to assess the differential relationship of prices of fruits and vegetables across the whole
BMI distribution of middle and high school students (Powell, Han et al. 2010). Findings
suggested that fruit and vegetable prices have significant effects on adolescents. The
52
effect of relationship between fruit and vegetable prices was substantially larger among
overweight teens compared to smaller effect among normal weight adolescent group.
Sturm and Datar used panel data from the ECLS-K for children in kindergarten
through third grade and they found significant relationship between children’s weight
changes and price of fruits and vegetables (Sturm and Datar 2005). Lower real prices for
fruits and vegetables obtained from the ACCRA data were significantly associated with
lower gain in BMI while lower meat prices had the opposite effect although the effect
was smaller and insignificant. This study also confirmed the findings from previous
researches that the estimated effects were larger for children in poverty and those at risk
for overweight.
Results from a controlled laboratory experiment by Epstein et al. suggest purchase
of healthy, low energy-dense foods by youths to be price elastic (Epstein, Handley et al.
2006). The experiment manipulated the prices of healthy, low energy-dense foods and
the study results showed evidence of higher prices of these foods led to lower
consumption or vice versa. Number of other controlled field experiments also suggested
that significantly lower prices of healthy foods were linked with higher consumption.
French et al. assessed difference in consumption of fruit and salad from a high school
cafeteria by reducing 50% of price and the consumption quadrupled during the
intervention period (French, Story et al. 1997). Later work published by French et al.
examined the effects of pricing on purchases of low-fat snacks from vending machines
53
and the results showed significant increase in low-fat snack sales for both adult and
adolescent population after price reductions (French, Jeffery et al. 2001).
Identification Test for Instruments
There is no one test to directly validate any instruments but there are several
diagnostic methods to test for both relevancy and strict exogeneity assumptions of an
instrument.
Relevancy can be easily tested by estimating a simple first-stage regression
between the problematic endogenous variable and the instruments. Partial F-test from a
regression model predicting obesity as a function of covariates and an instrument has a
null hypothesis of no association and can be used to test for the relevancy assumption as
stated by many economic literatures (Nelson and Startz 1990; Staiger and Stocker 1997).
The coefficient for the instrument from the regression model indicates the association
between the endogenous variable and the instrument and if the coefficient is significantly
different from zero, then the assumption of relevancy holds. Diagnostics based on the F-
test for the joint significant of the IVs can be confirmed when F-statistic is greater than
10. Shea’s partial R
2
will also be analyzed to test for instrument relevancy (Shea 1997).
Large partial R
2
indicates valid relevancy of the instruments as R
2
quantifies the percent
of total variation in obesity that is explained by the regression of obesity on instruments
while controlling for other covariates.
One of the two requirements that needs to be satisfied in order for the instrument
to be valid is that the IV must be uncorrelated with the outcome or the error term. In
54
order to test for the exogeneity assumption, Wald-test of exogeneity, Sargan test for
overidentification and Durbin-Wu-Hausman specification tests will be performed. A
simple z-test will be performed in order to test the significance of an estimator for the
instrument in a probit regression predicting asthma as a function of covariates, obesity
and an instrument. Insignificant partial estimator for the instrument implies no partial
effect of the instrument on asthma and the instrument is assumed to be valid. Increasing
the number of overidentifying restrictions (adding too many instruments than endogenous
variables) can lead to severe biases in regression estimators, therefore IV’s must be tested
against overidentification restrictions.
Marginal Effect of Childhood Obesity on Asthma
For binary independent variables, such as obesity, the first derivative method does
not result in marginal effects but has been suggested to compare the probabilities of the
binary independent variable between two different levels, 0 and 1, while holding other
covariates at their mean values (Greene, 1996). The marginal probability of having been
diagnosed with asthma in children while controlling for endogeneity of obesity from
equation (8) is given by:
Pr(Y
i
= 1) = Φ(x’
1
β
1
+ δ
1
Obesity) (9)
Where Φ is the standard normal distribution function, and Φ(x’
1
β
1
+ δ
1
Obesity) is
computed for each observation using the estimated coefficients from the bivariate probit
model. This framework gives the probability of having been diagnosed with asthma
controlling for the endogeneity of B and Y. Given that Obesity is a dummy variable, I
55
measured the marginal effect (M.E.) of being obese on the probability of having been
diagnosed with asthma as the sample average of changes in Φ(x’
1
β
1
+ δ
1
Obesity) with
discrete changes in Obesity while keeping all other variables X at their observed values:
(10)
3.5.4 Obesity Elasticity on Asthma and Attributable Fraction
Obesity elasticity of on asthma is measured by the percentage change in asthma
caused by a percent change in obesity which can be calculated by taking the marginal
effect of obesity and multiplying by the original prevalence rates for obesity over asthma.
The final equation is derived by the following,
ɛOBESITY = [ (P02 - P01)/P01 ] / [( PA2 - PA1)/PA1 ] = [ ∆ P0/ P01 ] / [ ∆ PA/ PA1 ] (11)
= [ ∆ P0/ P01 ] * [ PA1/∆ PA ]
= dP0/ dPA * PA1/ P01
= dPA /dP0 * P01/ PA1
where,
P
02
= Prevalence of Obesity in Period 2 (For example, prevalence rate of
childhood obesity in 5
th
grade),
P
01
= Prevalence of Obesity in Period 1 (For example, prevalence rate of
childhood obesity in 3
rd
grade),
56
P
A2
= Prevalence of Asthma in Period 2 (For example, prevalence rate of asthma
in 5
th
grade),
P
A1
= Prevalence of Asthma in Period 1 (For example, prevalence rate of asthma
in 3
rd
grade).
And from the definition of marginal effect, we know that dPA /dP0 is equivalent to the
marginal effect of obesity obtained from previous bivariate models.
Attributable fraction of obesity on asthma or the absolute impact of obesity on
asthma can be calculated by taking the proportional difference in prevalence rates for
obesity and asthma between the periods of interest and multiplying it by the marginal
effect of childhood obesity on asthma obtained from equation (10)
Attributable fraction of obesity on asthma = [ (P02 - P01)/( PA2 - PA1) ] * M.E. (12)
57
Chapter 4 Results
In each survey, a nationally representative sample of U.S. children was selected
using a complex, stratified, multistage probability cluster sampling design based on
schools. Samples were weighted using the procedure recommended in the ECLS-K
documentation. The sample for this project included all children from the ECLS-K
cohorts who had not required assistance with mobility or daily activities. Figure 1
describes the sample cohort selection from the three wave panel data. Total of 13,860
children have been selected in the baseline and 10,110 and 8,740 children were included
in the analyses for 5
th
and 8
th
grades, respectively (all rounded up to the nearest tenth digit
to comply to the restricted ECLS-K data agreement). Total of 6,970 children from
13,860 children had provided complete information on asthma and bmi/obesity status for
all three grades. The only significant difference between those with complete
information for all three grades and the incomplete/lost-to-follow-up children was that
higher percentage of incomplete/lost-to-follow-ups were from large and mid-size cities or
suburbs compared to those who were followed for all three grades (p<0.001, results not
shown).
58
Figure 1: Cohort Diagram
4.1 Trends in Childhood Obesity and Asthma
4.1.1 Changes in Mean BMI and Prevalence of Childhood Obesity and
Asthma
We detected a gradual increase in the mean BMI over time with the greatest
increase in the 8
th
grade, measured in 2007 (Figure 2). The prevalence rate for
overweight showed a decreasing trend up until the 1
st
grade while the prevalence rate for
the obesity showed a more gradual increasing trend over the years (Figure 3). Greatest
change in prevalence rate for childhood obesity was observed between 1
st
grade in Spring
of 2000 and 3
rd
grade in Spring of 2002 (5.7 percentage point difference) with the second
largest change in prevalence rate for childhood obesity being observed between 3
rd
grade
in Spring 2002 and5
th
grade in Spring of 2004 (5.3 percentage point difference). The
59
steady increasing trend in obesity and the increase in mean BMI over time confirmed
previous findings of epidemiological studies on childhood obesity. Figure 4 represents
the prevalence of asthma along with the prevalence of overweight and obesity.
Differences of asthma prevalence were 4.1 percentage point and 4.5 percentage point
between 3
rd
and 5
th
grades and 5
th
and 8
th
grades, respectively.
Figure 2: Increase in Body Mass Index (BMI) Over Time (All Children)
60
Figure 3: Prevalence of Overweight and Obesity Over Time (All Children)
Figure 4: Prevalence of Asthma Over Time (All Children)
61
The results in Table 1 shows the status of asthma by 8
th
grade for the children
whose complete data on asthma were available from 3
rd
grade to 8
th
grade and were non-
asthmatic in 3
rd
grade. Table 2 shows the changes in asthma rate by different weight
group for children with complete asthma data form 3
rd
to 8
th
grade who were non-
asthmatic in both 3
rd
and 5
th
grade. Approximately 11.9% of the children who were non-
asthmatic in 3
rd
grade and obese in both 3
rd
and 5
th
grade had become asthmatic by 8
th
grade where approximately 88% of children had remained in same obese group in 8
th
grade. Higher proportion of children had become asthmatic by 8
th
grade who were obese
in 3
rd
or 5
th
grade, respectively, compared to who were never obese in 3
rd
or 5
th
grade
(approximately 8.8% vs. 8.1% vs. 6.7%, respectively). 5.3% of children who were non-
asthmatic and obese in both 3
rd
and 5
th
grade were diagnosed with asthma by 8
th
grade,
compared to approximately 3.7% who were non-asthmatic in both 3
rd
and 5
th
grade but
only obese in one of the grades (either obese in 3
rd
or obese in 5
th
grade) and only 2.7%
of children who were never obese in both 3
rd
and 5
th
grade. When children were
categorized into 3 different weight groups (normal, overweight and obese) according to
their BMIs, approximately 70% of children stayed within the same weight group from 3
rd
to 8
th
grade (result not shown). Figure 5 depicts the changes in asthma rate for children
with complete asthma data on all three grade levels who were non-asthmatic in 3
rd
grade
by obese and non-obese group. In addition, changes in prevalence rate of asthma by
weight group for all children are also depicted in Figure 6.
62
Table 1: Change in Prevalence of Asthma in 3
rd
Grade Non-asthmatic Children
(Only Children with all 3 years of Complete Asthma Data)
Table 2: Change in Prevalence of Asthma in 3
rd
and 5
th
Grade Non-Asthmatic Children
(Only Children with all 3 years of Complete Asthma Data)
63
Figure 5: Change in Prevalence of Asthma in 3
rd
Grade Non-Asthmatic Children (Only
Children with all 3 years of Complete Asthma Data)
Figure 6: Change in Prevalence of Asthma by Weight Group (All Children)
64
4.2 Baseline Descriptive Statistics
Demographic characteristics of the 3
rd
grade study population by obese and non-
obese groups are summarized in Table 3. Based on the Rao-Scott chi-square tests,
significant differences were observed among obese and non-obese groups for gender,
asthma rate, ethnicity, household income category, parents’ highest education level, and
whether the child is from a social status below a poverty level (p <0.05). Significant
differences were also observed among obese and non-obese children for the level of
exercise/activity, maternal employment status and whether the child is from a home
composed of two-parent family or single parent family. Higher percentage of male
children was obese compared to non-obese group (54.7% vs. 49.9%) and asthma was
more prevalent in the obese group (22.7% vs. 12.1%). Obese groups were composed of
significantly higher percentage of minorities, including the African American and
Hispanics compared to non-obese group. In addition, significantly lower parents’ highest
level of education attainment was observed for obese group and significantly more
children were from a family with economic social status below the poverty level (19.6%
vs. 12.5%, p<0.05). Significantly high proportion of obese children had less amount of
aerobic exercise compared to the children’s age and gender group (18.7% vs. 8.1%) and
participated in less organization sports team or league in the past months (49.4% vs.
42.8%). Also, significantly higher proportion of obese children was from a single family
than non-obese children (20.5% vs. 16.8%). No significant differences were observed for
mean age in months, insurance status, city size/type or geographical region.
65
Table 3: Baseline Demographic Characteristics (3rd Grade)
*denotes statistical significance at the 5 percent
66
Table 3: Baseline Demographic Characteristics (3rd Grade) - Continued
*denotes statistical significance at the 5 percent
67
4.3 Cross-Sectional Contemporaneous Association between
Childhood Obesity and Asthma
4.3.1 Probit Model
Results from the weight-adjusted cross-sectional probit analyses of assessing the
association between child obesity and asthma for each wave panel (3
rd
, 5
th
, and 8
th
grade)
are described in Table 4. These contained the full set of covariates and analyzed the
direct associative effect between childhood obesity and asthma without controlling for
endogeneity. Backwards induction methods and likelihood ratio tests (LR test) and the
importance of covariates based on previous studies identified the best probit models for
analyzing the contemporaneous effect association between child obesity and asthma as
shown in Tables 4. The coefficients are significant at the 5% level when the absolute
value of the z score exceeds 1.9.
In all three grade levels, the effects of the childhood obesity measures on asthma
were significant and positive with increasing effect of obesity as children got older.
Results from marginal effects of obesity indicated that obese children were on average
0.02 percentage point more likely to be diagnosed with asthma than non-obese children in
3
rd
grade with compared to 0.03 percentage point and 0.06 percentage point in 5
th
and 8
th
grade, respectively. Variables such as regional information and type of city were not
significant in all models and LR test indicated no significant improvement in models with
these variables, therefore these variables were excluded from the final probit models.
Only obesity, gender, and race for African American were significantly associated with
68
having asthma for each grade level (p<0.05) and the signs of these estimates were similar
to those from previous findings that children who are obese or male were more likely to
have asthma. Other individual characteristics had the expected sign. Household income
greater than $50,000 had significantly reduced risk for developing asthma compared to
the lowest household income group (less than $25,000) in 3
rd
grade but the household
income effect seemed to disappear throughout the 5
th
and 8
th
grade. Maternal
employment, mother’s age at child’s birth and whether the child was from a single parent
household all had positive coefficients although the effects were insignificant. Other
covariates such as living in a household below a poverty level, having no health
insurance, child’s weight at birth, and mother’s age at first birth, although not significant
showed the expected negative sign. Covariates on air quality including both ozone and
particle index were not significantly associated with asthma for all three grade levels.
69
Table 4: Weight-Adjusted Probit Estimates for Contemporaneous Effect of Obesity on
Developing Asthma
*denotes statistical significance at the 5 percent
70
4.3.2 Probit Models for Subgroup Analyses
Table 5, 6 and 7 show similar results as the previous probit model, where
continuous variable, BMI was included in the models, replacing a dichotomized variable
for obesity. The results showed significant association between BMI with asthma among
three subgroups analyses of obese and non-obese (normal weight) group, obese group,
and non-obese group for all three grade levels. Individual characteristics had the
expected sign. In particular, positive BMI effects on asthma which describe the effects of
every unit increase in BMI on asthma were positively associated for all three groups with
subgroup analyses composed of only obese children showing the largest effect of BMI on
asthma. In general, boys were more asthmatic; African American and Hispanic children
had significantly higher rate of asthma diagnosis compared to white children for all three
subgroups and three grade levels. The coefficient on age in months had a significant
positive sign in only 3
rd
grade subgroup analysis for obese and non-obese children, but
the effects were not statistically significant in other grade levels or subgroups analyses.
Both maternal employment and mother’s age at child’s birth had positive coefficients
although the effects were insignificant. Other covariates such as living in a household
below a poverty level, having no health insurance, child’s weight at birth, and mother’s
age at first birth, although not significant showed the expected negative sign. In
subgroup analyses for obese and non-obese children, coefficient for a dichotomized
variable indicating whether the child was from a single parent family was significantly
associated with asthma with an expected positive sign in 5
th
grade. The coefficients on
single parent family had the expected positive sign, but were not statistically significant
71
in 3
rd
or 8
th
grade. In 3
rd
grade, children with less level of aerobic exercise relative to
their age and gender group were more likely to develop asthma compared to children
with higher level of aerobic exercise. For all subgroup analyses, 5
th
grade children whose
parents had some year of college or obtained a college degree were more likely to
develop asthma compared to parents who had not completed high school. Covariates on
air quality including both ozone and particle index had the expected positive sign
although they were not significantly associated with asthma for all three grade levels and
three subgroup analyses.
72
Table 5: Weight-Adjusted Probit Estimates for Contemporaneous Effects of BMI on
Developing Asthma among Obese and Non-Obese (Normal Weight) Children
*denotes statistical significance at the 5 percent
73
Table 6: Weight-Adjusted Probit Estimates for Contemporaneous Effects of BMI on
Developing Asthma among Obese Children
*denotes statistical significance at the 5 percent
74
Table 7: Weight-Adjusted Probit Estimates for Contemporaneous Effects of BMI on
Developing Asthma among Non-Obese (Normal Weight) Children
*denotes statistical significance at the 5 percent
75
4.4 Cross-Sectional Lagged Association between
Childhood Obesity and Asthma
4.4.1 Probit Model
Results for the lagged impact of childhood obesity on asthma for each grade level
from the probit models are shown in Table 8. These contained the full set of covariates
for the current grade level and give the lagged effect of childhood obesity on developing
asthma without controlling for endogeneity. The dichotomized variable for childhood
obesity is referred to the obesity status of each child from data collected from previous
wave panel while other covariates indicate the current characteristics of each child data
collected for that current grade level. Lagged obese status were positively associated
with asthma among all 3
rd
, 5
th
and 8
th
grade children and the effect seemed to increase as
children got older. When compared to the current status of obesity or contemporaneous
effect of childhood obesity on asthma, the effect of lagged association between obesity
and asthma has decreased for all three grades levels. 3
rd
grade children who were obese
during the previous panel or 1
st
grade were 0.01 percentage point times likely to be
diagnosed with asthma by the current grade or 3
rd
grade, whereas the marginal effect of
lagged obesity status has increased to 0.02 percentage point among 5
th
grade children and
0.04 percentage point among 8
th
grade children. Only obesity, gender, and race for
African American were significantly associated with having asthma for each grade level
(p<0.05) and the signs of these estimates were similar to those from previous findings
that children who are obese or male were more likely to have asthma. Age in months
were only significantly associated with asthma for 3
rd
grade level and the effect became
76
insignificant in the analysis for 5
th
and 8
th
grade level. Other individual characteristics
had the expected sign. Maternal employment, mother’s age at child’s birth and whether
the child was from a single parent household all had positive coefficients although the
effects were insignificant. Other covariates such as household income, living in a
household below a poverty level, having no health insurance, child’s weight at birth, and
mother’s age at first birth, although not significant showed the expected negative sign.
Children whose parents had attained education level higher than high school were more
likely to be diagnosed with asthma only for 5
th
grade level and the coefficients became
insignificant in other grade analyses. Covariates on air quality including both ozone and
particle index were not significantly associated with asthma for all three grade levels.
77
Table 8: Weight-Adjusted Probit Estimates for Lagged Obesity Effect on Developing
Asthma
*denotes statistical significance at the 5 percent
78
4.5 Causal Effect of Childhood Obesity on Asthma
4.5.1 Validity Tests for Instrumental Variables
Table 9 and 10 show the list of proposed IVs and the results for relevancy and
exogeneity tests for the proposed IVs. Various groups of quarterly prices indices for food
groups have been considered as the IVs including different types of healthy and
unhealthy foods from two food price index data. Two IVs including the quarterly price
indices for fresh healthy food composed of fresh fruits and vegetables and unhealthy
drink group composed of carbonated soda showed strong correlation between the
instruments and child obesity and showed no significant concerns from tests for
exogeneity as shown from Table 9. Additional proposed IVs of quarterly price indices
for other types of healthy food, combining all healthy food such as a bundle of fresh and
canned fruits and vegetables from QFAHPD and a bundle of fresh and frozen fruits and
vegetables from the ACCRA database showed strong correlation but showed concerns
from exogeneity tests. Similarly, although there were strong significant correlation
between other quarterly price indices of unhealthy foods composed of different bundles
of carbonated soda, sports drink, packaged sweets, packaged snacks and ice cream,
showed significant concerns for exogeneity tests. Results from the first-stage probit
models for IVs and the results for relevancy and exogeneity are shown in Table 10 by
each grade level. Table 10 also reports the impact of the instruments on the obesity
measure from the first-stage probit models. Controlling for all the covariates, the
instruments are individually and jointly significant predictors of childhood obesity. The
79
coefficients for each IVs had the expected sign where the quarterly price index for fresh
healthy food showed a significant positive association while the quarterly price index for
carbonated soda was negatively associated with obesity for all three grade levels. The
positive association between the price index for fresh healthy food and obesity indicated
that higher price of fresh healthy food was associated with higher rate of childhood
obesity whereas the negative association between the price index for carbonated soda and
childhood obesity indicated that as the price of carbonated soda increased the rate of
childhood obesity fell. These phenomena can be explained by simple demand function
for normal goods when price goes up, the demand for the certain product goes down so
depending on the benefits or detriments associated with healthy vs. unhealthy food, BMI
or childhood obesity are affected due the changes in price for these goods. The F-
statistics from the first-stage probit models for both IVs for all three grades were well
above 10 and the Shea’s R
2
or partial R
2
for the IVs were significant for all three grade
levels. Results from the Wald-test of exogeneity and Sargan test for overidentification
showed no significant concerns. The findings demonstrated that the instruments satisfied
the non-weakness requirements of IVs.
80
Table 9: Tested IVs and Results for Relevancy and Exogeneity
81
Table 10: Results from Different Combination of Instruments using IV Probit Model
Correlation between the IVs and baseline observable characteristics which may be
correlated to the outcomes directly were investigated. Instruments were dichotomized
into two groups as high vs. low based on a simple threshold using the median for each
price index group. T-test or chi-square test statistics were calculated by instruments and
compared with the statistics across high vs. low price index group (Table 11). The only
significant difference observed across observable characteristics between the high and
low price index groups for instrument, quarterly price index for fresh healthy food, was
the geographical region where children resided in. High price index group was composed
of slightly more children from the West region compared to the low price index group
(33.3% vs. 30. 0%). The second instrument of quarterly price index for carbonated soda
resulted in no significant difference among the high and low price index group across all
observable characteristics. Correlations between the IVs and one of the known major risk
factors for asthma, environmental air pollutants were further investigated and the results
are shown in Table 12. The first instrument of quarterly price index for fresh healthy
82
food was not significantly associated with any of the air pollutant measures including the
categorical ozone and particle indices and average number of ozone and particles. When
a significant association is observed between the IVs and the air quality measures, it
raises a concern for the IVs because it may indicate for example healthy food is pricier in
highly polluted area which introduces unobservables or confounding effects of air
pollutants in asthma analyses. The results showed no significant concerns for the IVs
across observable characteristics and air quality measures as shown in Tables 11 and 12.
These findings along with results from Table 9 demonstrated the validity of the IVs.
83
Table 11: Observable Characteristics across IV1 (Price Index for Fresh Healthy Food)
for 3
rd
Grade Children
*denotes statistical significance at the 5 percent
84
Table 11: Observable Characteristics across IV1 (Price Index for Fresh Healthy Food)
for 3
rd
Grade Children - Continued
*denotes statistical significance at the 5 percent
Table 12: Relationship Between Air Pollutants and IV1 (Price Index for Fresh Healthy
Food)
85
4.5.2 Instrumental Variable Bivariate Probit Model
The key results from the IV probit and bivariate probit models are presented in
Tables 13, 14, and 15 for 3
rd
, 5
th
, and 8
th
grade children, respectively. The hypothesis of
no association is rejected for all three grade levels from the IV probit and bivariate probit
models which indicate that asthma and childhood obesity are endogenously determined
and the results from the multivariate probit models in Table 4 are biased. The bivariate
probit models for all three grade levels showed a significant positive relationship between
childhood obesity and asthma while controlling for other covariates. As shown in Table
13-15, the impact of childhood obesity on asthma became greater after applying
instrumental variables in the analyses. This implies that unobservable factors play a role
in childhood obesity and the relationship between childhood obesity and the outcome,
therefore, the impact would be significantly underestimated if the endogeneity bias were
not to be controlled in the analyses. For all three grade levels, boys were more likely to
be diagnosed with asthma compared to girls and African American children were more
like to be asthmatic. The causal effect of childhood obesity on asthma seemed to increase
as children got older and the results from the bivariate probit models for all three grade
levels seemed to be less than the IV probit models perhaps due to the reason of IV probit
mdoels inability to handle dichototomized endogenous variable. Other covariates such as
living in a household below a poverty level, having no health insurance, child’s weight at
birth, and mother’s age at first birth, although not significant showed the expected
negative sign. Covariates on air quality including both ozone and particle index were not
significantly associated with asthma for all three grade levels.
86
Table 13: Results from Instrumental Variable Model for 3
rd
Grade Children
*denotes statistical significance at the 5 percent
87
Table 14: Results from Instrumental Variable Model for 5
th
Grade Children
*denotes statistical significance at the 5 percent
88
Table 15: Results from Instrumental Variable Model for 8
th
Grade Children
*denotes statistical significance at the 5 percent
89
Results from Tables 13-15 showed a significant relationship between the
childhood obesity and asthma for all 3
rd
, 5
th
, and 8
th
grade children. Marginal effects of
childhood obesity on asthma were calculated following Greene (1996). The marginal
effects for the childhood obesity based on different BMI thresholds for all three grade
levels are reported in Table 16. The marginal effects of binary variables such as
childhood obesity are interpreted as the increased or decreased probability a child would
be asthmatic if the binary variable is true, i.e. obese. For continuous variables, marginal
effects are easily calculated by taking the partial derivatives of the regression coefficients,
whereas the marginal effect for binary variables is calculated using equation (10).
Table 16: Marginal Effects from Instrumental Variable Model
For each of the multivariate bivariate probit models, we estimated the marginal
effects for each grade level. The marginal effects yield the percentage point change in
the prevalence of asthma generated by changes caused in obesity. Marginal effects for
childhood obesity are consistent through all grade levels and different BMI thresholds
(Table 16). As measured by the marginal effects, childhood obesity increased the
probability of being diagnosed with asthma and the positive impact of childhood obesity
90
on asthma increased as children got older. Obese children in 3
rd
grade had a higher
probability of being asthmatic where childhood obesity increased the probability of
asthma by 0.10 while holding other covariates constant and using a standard 95% BMI
threshold was used to define childhood obesity. Similarly, being obese in 5
th
and 8
th
grade increased the probability by 0.14 and 0.19 percentage points when same 95% BMI
threshold was used to define childhood obesity. Both marginal effects of being
overweight, defined by BMI threshold between 85
th
percentile and 95
th
percentile, and
obese or overweight, defined by BMI threshold of 85
th
percentile or greater, were slightly
lower than marginal effects of obesity for all three grade levels.
Marginal effects for obesity have increased when a higher threshold of BMI was
used to define childhood obesity (Table 17). There was an increase of 0.02 in marginal
effect for obesity with using a 97
th
percentile BMI threshold for 3
rd
grade children while
holding other covariates constant at their mean. In 5
th
and 8
th
grade children, the
marginal effects of obesity on asthma were increased by 4% and 3%, respectively by
increasing the BMI threshold from 95
th
percentile to 97
th
percentile. Overall, the average
predicted probabilities of asthma have increased as the BMI threshold which defines
childhood obesity increased.
91
Table 17: Marginal Effects from Instrumental Variable Model using Different Obesity
BMI Threshold Level
92
4.5.3 Obesity Elasticity on Asthma and Attributable Fraction
Equation (11) was used to calculate the obesity elasticity of asthma which
measures the percentage change in asthma caused by a percent change in obesity.
Between 3
rd
and 5
th
grade, obesity prevalence increased by 5.3 percentage point while
asthma prevalence increased by 4.1 percentage point. Marginal effect of obesity revealed
that each percentage point increase in obesity led to an increase of 0.10 percentage points
in asthma during the same period. Changes in prevalence percentage points for obesity
and asthma were slightly higher between 5
th
and 8
th
grade where obesity prevalence
increased by 5.9 percentage points and asthma prevalence increased by 4.5 percentage
points. Obesity elasticity of asthma were 0.163, 0.206, and 0.274 for 3
rd
, 5
th
and 8
th
grade,
respectively, which in return can be interpreted as a 10% increase in obesity resulting in
1.63%, 2.06%, and 2.74% increase in asthma for 3
rd
, 5
th
, and 8
th
grade, respectively.
Attributable fraction of obesity on asthma which is the same as the absolute impact of
obesity on asthma in our study from equation (12) showed that 13.0% of the increase in
asthma between 3
rd
and 5
th
grade were explained by the increase in obesity while 18.4%
of the increase in asthma between 5
th
and 8
th
grade were explained by the increase in
obesity.
93
Chapter 5 Discussions and Conclusions
5.1 Summary of Findings
The paper examined the trends of obesity and the impact of childhood obesity on
asthma among young and adolescent children in U.S. This is the first study to produce
estimates which controlled for the endogeneity of childhood obesity and asthma using an
IV method approach. The steady increasing trend for obesity and the increase in mean
BMI over time confirmed the previous findings of epidemiological studies on childhood
obesity. Baseline estimates using multivariate probit models to analyze the
contemporaneous associative relationship between childhood obesity and asthma that did
not control for unobservables or endogeneity bias around childhood were also presented
in the study. The results suggested a small positive effect of childhood obesity on asthma
for all three grade levels and the effects seemed to increase as children got older. Boys
and African Americans were significantly at higher risk for developing asthma compared
to girls and white children while controlling for other covariates. Few covariates such as
parents’ highest level of education attainment, certain household income levels, and
certain level of aerobic exercises compared to the child’s peer and gender, were
significantly associated with childhood obesity only among 3
rd
grade children; however
the significant associative relationships disappeared among 5
th
and 8
th
children. Other
covariates such as living in a household below a poverty level, having no health
insurance, child’s weight at birth, and mother’s age at first birth, although not significant
showed the expected negative sign. Covariates on air quality including both ozone and
94
particle index were not significantly associated with asthma for all three grade levels.
Similar results were obtained when an indicator variable for childhood obesity was
replaced with a continuous variable, BMI. The results showed significant association
between BMI with asthma among three subgroups analyses of obese and non-obese
(normal weight) group, obese group, and non-obese group for all three grade levels.
Individual characteristics had the expected sign.
In addition to the contemporaneous associative effect of childhood obesity on
asthma, lagged associative effect of childhood obesity on asthma was further investigated
using multivariate probit models. These contained the full set of covariates for the
current grade level and reported the lagged effect of childhood obesity on developing
asthma without controlling for endogeneity. An indicator variable for obese status during
the previous wave panel was included to analyze the relationship between previous
obesity status and current asthma status. Lagged obese status were positively associated
with asthma among all 3
rd
, 5
th
and 8
th
grade children and the effect seemed to increase as
children got older. When compared to the current status of obesity or contemporaneous
effect of childhood obesity on asthma, the effect of lagged association between obesity
and asthma were smaller for all three grades levels.
This paper demonstrated that childhood obesity was endogenously determined
with asthma by using bivariate probit models. Two price indices for healthy and
unhealthy food were identified as valid IVs with strong theoretical basis for the choice of
IVs and rigorous testing for relevancy and exogeneity assumptions of IVs. In contrast to
95
the baseline multivariate probit models, the bivariate probit estimates which controlled
for endogeneity showed that childhood obesity exerted greater statistically significant
positive impact on asthma for all grade levels. The findings were robust to the use of
different sets of covariates. The greater magnitude of childhood obesity impact on
asthma after controlling for unobservables around obesity implied that the obesity effect
may have been underestimated if endogeneity was not properly controlled for in the
analyses. The positive effect of childhood obesity was greater for the severely obese,
defined by higher threshold of BMI than standard threshold of 95
th
percentile of BMI
while the positive effect of overweight status on asthma was smaller than the obesity
effect.
Moreover, further investigation on percentage change in asthma caused by a
percent change in obesity and changes in asthma attributed by changes in obesity has
been conducted in the study. Results showed that a 10% increase in obesity resulted in
1.63%, 2.06%, and 2.74% increase in asthma for 3
rd
, 5
th
, and 8
th
grade, respectively
which seemed to lead to an ample percentage increase in asthma resulting from increase
in obesity. Attributable fraction of obesity on asthma which is the same as the absolute
impact of obesity on asthma in our study showed that 13.0% of the increase in asthma
between 3
rd
and 5
th
grade were explained by the increase in obesity while 18.4% of the
increase in asthma between 5
th
and 8
th
grade were explained by the increase in obesity.
96
5.2 Study Limitations
There are several limitations to be considered for this study. First, the study
sample size was not sufficient for detailed stratified analyses by different population
subgroups. Although the results did not reveal significant differences in asthma status
among racial and ethnic groups except for African American children, this does not mean
that such racial or ethnic effects are absent. It could also mean that the study was not
large enough to detect the smaller scale of differences among different ethnic groups.
Second, despite the effort to control for the environmental risk factors for asthma using
size of city and difference in air quality by linking tertiary air quality data, we could not
obtain all related environmental factors from the data. Other environmental risk factors
such as exposure to smoke by smoking habits by family members, indoor air quality, or
other health conditions including allergies or genetic predisposition that may exacerbate
asthma could be not observed due to the nature of the retrospective observational data
analyses. Second, there is a concern for the direction of causality between childhood
obesity and asthma where obesity may cause asthma and asthmatic children could also
become obese due to inability to do adequate amount of aerobic exercise. However, this
study tried to control for the simultaneity problem by implementing the instrumental
variables technique with few limitations of the approach on its own.
Estimates obtained from the bivariate probit models using the IV technique were
believed to be unbiased but with tremendous efforts to find the valid instrumental
variables. The instrumental variables were chosen based on strong theoretical basis
97
which were shown to be highly correlated with childhood obesity but not with the
outcome of asthma. Despite the large number of statistical tests which have been
developed to test the validity of the instruments there is still a lack of direct validity test.
It is often with difficulty to argue the validity of the instruments and this approach should
be done with extensive literature reviews to show strong theoretical basis. It is our belief
that food price indices to be highly correlated with childhood obesity but not directly
correlated with asthma which were shown through rigorous statistical testing. Some may
argue that food price indices may be directly correlated with other factors that may
impact asthma such as level of exercise and ambient air quality. However, these potential
observable risk factors were further investigated by different food price index groups and
were determined to be similar in nature. Although health conditions of children were not
appropriately adjusted in the models and other potential limitations of instrumental
variables approach still exist, the estimates obtained from the instrumental variables
technique are believed to be closer to the true impact on asthma compared to the biased
estimates obtained from general multivariate probit models without controlling for
endogeneity bias.
This study used an IV approach based on the associated food prices of healthy and
unhealthy food which have been previously shown by numerous studies to be highly
correlated to obesity. Similar to the previous studies examining the effect of food prices
on BMI or obesity status, this study was also unable to observe the actual consumption of
healthy and unhealthy food but applied the general theory of elasticity and supply and
demand. Actual consumption or demand for food may greatly differ by individual
98
preference, taste, religion or culture and not solely reflected by differences in food prices.
However, previous researches on consumption of healthy food including fruits and
vegetables reported that low-income families in the U.S. spent significantly less on fruits
and vegetables in a given period and were twice as likely to not purchase any fruits or
vegetables compared to higher-income families (Blisard, Stewart et al. 2004). Results
showed greater difference in purchasing behavior with strong response to prices than
small changes in income. Furthermore, a study by Huang and Lin estimated that a 10%
reduction in price for fruits and vegetables led to an approximately 8% increase in
consumption (Huang & Lin, 2000).
5.3 Policy Implications
This showed a causal link between childhood obesity and asthma and on could
expect that an intervention targeting childhood to eventually lead to a reduction in asthma.
Policy reforms through improved exercise programs at school by mandating certain hours
of active physical education for all students will be greatly beneficial in combating
childhood obesity. This study also found the food price to play an important role in
childhood obesity so it would be greatly beneficial to increase subsidies and food
assistance programs for healthy food especially among low-socioeconomic status
families. Also, making unhealthy food options less available at school grounds and
mandating clearer labeling of calorie counts at chain restaurants under the provisions
included under the affordable care would be helpful in addressing childhood obesity
problems. Increasing negative effect of childhood obesity on asthma as shown by greater
99
marginal effect also indicates that policies in addressing childhood obesity should target
older children rather than younger children if the funding is limited.
This study focused on the effects of childhood obesity on asthma for a variety of
reasons. First, the impact of childhood obesity has been getting enormous attention
especially in the past few years and the CDC has already recognized obesity as one of the
winnable battles which can be prevented with proper behavioral changes. Childhood
obesity has reached epidemic levels in developed countries including the U.S. While
childhood obesity carries a great magnitude of health problems and is called a gateway
condition to other potential diseases including cardiovascular diseases, hypertension,
diabetes as well as other orthopedic, psychological and social problems, asthma has also
been on the rise which is the number one reason on children’s absenteeism from school
and plays a huge burden on family and society. Childhood obesity and asthma account
for a large number of lost school days and deprive children of both academic
achievement and social interaction. Both conditions place great strain on healthcare
resources as a result of doctor and hospital visits in addition to other costs for treatments.
Furthermore, childhood obesity has been found to have a strong positive
association with asthma in a wide range of research studies. However, none of the
existing literature examined the causality of childhood obesity on asthma. Childhood
obesity can be affected by policy reforms through improved exercise programs at school,
increasing subsidies and food assistance programs for healthy food for low-income
families and by providing healthier food options for breakfast and lunch at school. In
100
addition, making unhealthy food options with saturated fat and high contents of sugar less
available at school grounds and neighborhoods can be beneficial in combatting childhood
obesity. For the 2002-03 school year, the U.S. Department of Agriculture started a pilot
program which provided fresh and dried fruits and vegetables free to children in about
100 schools. This pilot program has since become a permanent program that was
expanded to cover selected schools in all 50 states, as part of the 2008 Farm Bill. An
extensive evaluation of the program is warranted to determine the potential effectiveness
of such program to combat childhood obesity.
Informational interventions to change consumption behavior by increasing the
value a consumer places on fruits and vegetables may also be helpful, especially among
high risk children who are overweight and at borderline obese. Our findings of a greater
effect of childhood obesity on asthma of African American ethnicity raise important
policy issues. In order to reduce health disparities, the program should target obese
children from low socio-economic status who are less likely to register for the program.
Policy makers can also improve healthcare infrastructures for early detection of
childhood obesity by providing programs to properly screen and treat the condition early
on before it leads to other severe problematic health implications. Effective
comprehensive weight-management programs incorporating counseling and other
interventions that target diet and physical activity as well as subsidies for healthier food
consumption directed at certain target or risk groups could further reduce some of the
differentials in asthma status among the African American children. Interventions should
101
also include behavioral management techniques and parental involvement in such
programs should be highly encouraged.
One of many potential reasons for deterring from adopting such intervention
program may be the uncertainty of return on investment due frequently unclear future
benefits gained from the interventions. Reducing childhood obesity is a long-term
investment which aims at preventing future obesity-related medical conditions and costs.
Other potential deterrent to the adoption is the poor cost measurement. It is especially
difficult to measure health benefits of childhood obesity interventions since the benefits
are rarely revealed until adulthood. Moreover, there exists a difficulty with assessing the
future medical costs saved from the reduced prevalence of childhood obesity. With more
rigorous approaches in measuring benefits and costs associated with such interventions
along with studies evaluating effectiveness or cost-effectiveness of these interventions
will be beneficial in making policy decisions and to make appropriate resource allocation.
Researchers, policy makers and clinicians should work side by side to improve current
health care infrastructure by targeting appropriate at risk children who are obese or
overweight through proper screening and to develop guidelines for effective concomitant
treatment of childhood obesity and its related health implications.
5.4 Conclusion
Gradual increases in mean BMI and prevalence of childhood obesity along with
prevalence of asthma were observed in this study. Children who were exposed to obesity
at any point in time during their kindergarten years were at higher risk for developing
102
asthma compared to those children who were never obese during the same time. This
may warrant not only a contemporaneous effect of childhood obesity on asthma but also a
lagged effect of obesity status on developing asthma later in life.
No other existing literatures have examined the causal relationship of
childhood obesity on asthma due to its difficulty in controlling for potential endogeneity
bias around childhood obesity. This study focused on analyzing the causality of
childhood obesity on asthma while controlling for not only observable but also
unobservable factors using the instrumental variables approach. Through this approach,
the study determined that childhood obesity significantly increases the risk for
developing asthma especially in boys and African American children. We suspected that
the instrumental variables approach was a superior method to investigate a causal
relationship between childhood obesity and asthma over panel analyses due to infrequent
changes in one’s weight or obesity and asthma status. Economic evaluation is one of
many tools that helps inform policy makers in the policy making process. This study
proposed an economic evaluation technique to determine causality of childhood obesity
on asthma and this analysis raised few questions which can be explored in future research.
First, more specific modeling approach can be done to estimate actual food consumption
of healthy vs. unhealthy food and its impact on childhood obesity using a data from pilot
studies. Second, obtaining certain risk factors for asthma, such as atopic disease, having
a pet indoor, exposure to maternal smoking and other genetic disposition, can provide an
opportunity to extend this project with more robust results using an instrumental variable
approach for endogenous variables.
103
Bibliography
Ahmad, N., S. Biswas, S. Bae, K. E. Meador, R. Huang and K. P. Singh (2009).
"Association between obesity and asthma in US children and adolescents." CORD
Conference Proceedings 46(7): 642-646.
Aligne, C. A., P. Auinger, R. S. Byrd and M. Weitzman (2000). "Risk factors for
pediatric asthma. Contributions of poverty, race, and urban residence." Am J Respir Crit
Care Med 162(3): 873-877.
Anderson, P. M., K. F. Butcher and P. B. Levine (2003). "Maternal employment and
overweight children." J Health Econ 22(3): 28-28.
Barlow, S. E. and W. H. Dietz (2002). "Management of child and adolescent obesity:
summary and recommendations based on reports from pediatricians, pediatric nurse
practitioners, and registered dietitians." Pediatrics 110(1): 236-238.
Beydoun, M. A., L. M. Powell and Y. Wang (2008). "The association of fast food, fruit
and vegetable prices with dietary intakes among US adults: is there modification by
family income?" Soc Sci Med 66(11): 2218-2229.
Bhattacharya, J. and M. K. Bundorf (2009). "The incidence of the healthcare costs of
obesity." J Health Econ 28(3): 649-658.
Bibbins-Domingo, K., P. Coxson, M. J. Pletcher, J. Lightwood and L. Goldman (2007).
"Adolescent overweight and future adult coronary heart disease." N Engl J Med 357(23):
2371-2379.
Blisard, Noel & Stewart, Hayden & Jolliffe, Dean, 2004. "Low-Income Households'
Expenditures on Fruits and Vegetables," Agricultural Information Bulletins 33755,
United States Department of Agriculture, Economic Research Service. Retrieved Jan 12,
2013 from http://ideas.repec.org/p/ags/uerser/34041.html#cites.
Brauer, M., G. Hoek, H. A. Smit, J. C. de Jongste, J. Gerritsen, D. S. Postma, M. Kerkhof
and B. Brunekreef (2007). "Air pollution and development of asthma, allergy and
infections in a birth cohort." Eur Respir J 29(5): 879-888.
Cade, J., H. Upmeier, C. Calvert and D. Greenwood (1999). "Costs of a healthy diet:
analysis from the UK Women's Cohort Study." Public Health Nutr 2(4): 505-512.
104
Cameron, A.and Trivedi P. Microeconometrics: Methods and Applications. Cambridge
University Press, May 2005. ISBN 0521848059.
Camilo, D. F., J. D. Ribeiro, A. D. Toro, E. C. Baracat and A. A. Barros Filho (2010).
"Obesity and asthma: association or coincidence?" J Pediatr (Rio J) 86(1): 6-14.
Cassol, V. E., T. M. Rizzato, S. P. Teche, D. F. Basso, D. F. Centenaro, M. Maldonado, E.
Z. Moraes, V. N. Hirakata, D. Solé and S. S. Menna-Barreto (2006). "Obesity and its
relationship with asthma prevalence and severity in adolescents from southern Brazil." J
Asthma 43(1): 57-60.
Castro-Rodríguez, J. A., C. J. Holberg, W. J. Morgan, A. L. Wright and F. D. Martinez
(2001). "Increased incidence of asthmalike symptoms in girls who become overweight or
obese during the school years." Am J Respir Crit Care Med 163(6): 1344-1349.
Cawley, J. and F. Liu (2012). "Maternal employment and childhood obesity: A search for
mechanisms in time use data." Econ Hum Biol 10(4): 352-364.
Center for Disease Control and Prevention (2011). Children's Food Environment State
Indicator Report, 2011. Retrieved March 13, 2012, from
http://www.cdc.gov/obesity/downloads/childrensfoodenvironment.pdf.
Connor, S. (2006). "Food-related advertising on preschool television: building brand
recognition in young viewers." Pediatrics 118(4): 1478-1485.
Cottrell, L., W. A. Neal, C. Ice, M. K. Perez and G. Piedimonte (2010). "Metabolic
Abnormalities in Children with Asthma." Am J Respir Crit Care Med.
Cox, B. D. and M. Whichelow (1996). "Ratio of waist circumference to height is better
predictor of death than body mass index." BMJ 313(7070): 1487-1487.
Dietz, W. H. and T. N. Robinson (1998). "Use of the body mass index (BMI) as a
measure of overweight in children and adolescents." J Pediatr 132(2): 191-193.
Dong, G. H., Y. Cao, H. L. Ding, Y. N. Ma, J. Jin, Y. D. Zhao and Q. C. He (2007).
"Effects of environmental tobacco smoke on respiratory health of boys and girls from
kindergarten: results from 15 districts of northern China." Indoor Air 17(6): 475-483.
Epstein, L. H., K. K. Dearing, R. A. Paluch, J. N. Roemmich and D. Cho (2007). "Price
and maternal obesity influence purchasing of low- and high-energy-dense foods." Am J
Clin Nutr 86(4): 914-922.
105
Epstein, L. H., E. A. Handley, K. K. Dearing, D. D. Cho, J. N. Roemmich, R. A. Paluch,
S. Raja, Y. Pak and B. Spring (2006). "Purchases of food in youth. Influence of price and
income." Psychol Sci 17(1): 82-89.
Epstein, L. H., M. D. Myers, H. A. Raynor and B. E. Saelens (1998). "Treatment of
pediatric obesity." Pediatrics 101(3): 554-570.
Figueroa-Muñoz, J. I., S. Chinn and R. J. Rona (2001). "Association between obesity
and asthma in 4-11 year old children in the UK." Thorax 56(2): 133-137.
Finkelstein, E. A., I. C. Fiebelkorn and G. Wang (2003). "National medical spending
attributable to overweight and obesity: how much, and who's paying?" Health Aff
(Millwood) Suppl Web Exclusives: W3-26.
Finkelstein, E. A., I. C. Fiebelkorn and G. Wang (2004). "State-level estimates of annual
medical expenditures attributable to obesity." Obes Res 12(1): 18-24.
Flegal, K. M., M. D. Carroll, C. L. Ogden and L. R. Curtin (2010). "Prevalence and
trends in obesity among US adults, 1999-2008." JAMA 303(3): 235-241.
Freedman, D. S., W. H. Dietz, S. R. Srinivasan and G. S. Berenson (1999). "The relation
of overweight to cardiovascular risk factors among children and adolescents: the
Bogalusa Heart Study." Pediatrics 103(6 Pt 1): 1175-1182.
French, S. A., R. W. Jeffery, M. Story, K. K. Breitlow, J. S. Baxter, P. Hannan and M. P.
Snyder (2001). "Pricing and promotion effects on low-fat vending snack purchases: the
CHIPS Study." Am J Public Health 91(1): 112-117.
French, S. A., M. Story, R. W. Jeffery, P. Snyder, M. Eisenberg, A. Sidebottom and D.
Murray (1997). "Pricing strategy to promote fruit and vegetable purchase in high school
cafeterias." J Am Diet Assoc 97(9): 1008-1010.
FX, P.-S. (1991). "Health implications of obesity." Am J Clin Nutr 53(6 Suppl): 1595S-
1603S.
Gable, S. and Lutz, S. (2000), Household, Parent, and Child Contributions to Childhood
Obesity. Family Relations, 49: 293–300.
Gehring, U., J. Cyrys, G. Sedlmeir, B. Brunekreef, T. Bellander, P. Fischer, C. P. Bauer,
D. Reinhardt, H. E. Wichmann and J. Heinrich (2002). "Traffic-related air pollution and
respiratory health during the first 2 yrs of life." Eur Respir J 19(4): 690-698.
Giamila, G. F. (2005). "Adipose tissue, adipokines, and inflammation." J Allergy Clin
Immunol 115(5): 911-920.
106
Gibson, L. Y., S. M. Byrne, E. A. Davis, E. Blair, P. Jacoby and S. R. Zubrick (2007).
"The role of family and maternal factors in childhood obesity." Med J Aust 186(11): 591-
595.
Gilliland, F. D., K. Berhane, T. Islam, R. McConnell, W. J. Gauderman, S. S. Gilliland, E.
Avol and J. M. Peters (2003). "Obesity and the risk of newly diagnosed asthma in school-
age children." Am J Epidemiol 158(5): 406-415.
Gortmaker, S. L., A. Must, J. M. Perrin, A. M. Sobol and W. H. Dietz (1993). "Social and
economic consequences of overweight in adolescence and young adulthood." N Engl J
Med 329(14): 1008-1012.
Gray, D. S. and K. Fujioka (1991). "Use of relative weight and Body Mass Index for the
determination of adiposity." J Clin Epidemiol 44(6): 545-550.
Greene, W., Marginal Effects in the Bivariate Probit Model (June 1996). NYU Working
Paper No EC-96-11. 1996. Retrieved May 1, 2012 from
http://ssrn.com/abstract=1293106.
Han, J. C., D. A. Lawlor and S. Y. Kimm (2010). "Childhood obesity." 375(9727): 1737-
1748.
Hasler, G., P. J. Gergen, V. Ajdacic, A. Gamma, D. Eich, W. Rössler and J. Angst
(2006). "Asthma and body weight change: a 20-year prospective community study of
young adults." Int J Obes (Lond) 30(7): 1111-1118.
Hedley, A. A., C. L. Ogden, C. L. Johnson, M. D. Carroll, L. R. Curtin and K. M. Flegal
(2004). "Prevalence of overweight and obesity among US children, adolescents, and
adults, 1999-2002." JAMA 291(23): 2847-2850.
Huang, K., Lin, B. Estlimation of food demandand nutrient elasticities from household
survey data. Technical Bulletin No. 1887. Food and Rural Economics Division,
Economic Research Service, U.S. Department of Agriculture, 2000. Retrieved Jan 12,
2013 from http://ageconsearch.umn.edu/bitstream/33579/1/tb001887.pdf.
Huffman, F. G., S. Kanikireddy and M. Patel (2010). "Parenthood--a contributing factor
to childhood obesity." Int J Environ Res Public Health 7(7): 2800-2810.
Infante-Rivard, C. C. (1993). "Childhood asthma and indoor environmental risk factors."
Am J Epidemiol 137(8): 834-844.
Jeong, Y., K. Jung-Choi, J. H. Lee, H. Y. Lee, E. A. Park, Y. J. Kim, E. Ha, S. Y. Oh and
H. Park (2010). "Body weight at birth and at age three and respiratory illness in preschool
children." J Prev Med Public Health 43(5): 369-376.
107
Jusot, F., M. Khlat, T. Rochereau and C. Serme (2008). "Job loss from poor health,
smoking and obesity: a national prospective survey in France." JECH 62(4): 332-337.
Kajbaf, T. Z., S. Asar and M. R. Alipoor (2011). "Relationship between obesity and
asthma symptoms among children in Ahvaz, Iran: a cross sectional study." Ital J Pediatr
37(1): 1-1.
Kim, K. W., Y. H. Shin, K. E. Lee, E. S. Kim, M. H. Sohn and K. E. Kim (2008).
"Relationship between adipokines and manifestations of childhood asthma." CORD
Conference Proceedings 19(6): 535-540.
Kusunoki, T., T. Morimoto, R. Nishikomori, T. Heike, M. Ito, S. Hosoi and T. Nakahata
(2008). "Obesity and the prevalence of allergic diseases in schoolchildren." Pediatr
Allergy Immunol 19(6): 527-534.
Lin, S., M. I. Gomez, S. A. Hwang, J. P. Munsie and E. F. Fitzgerald (2008). "Self-
reported home environmental risk factors for childhood asthma: a cross-sectional study of
children in Buffalo, New York." J Asthma 45(4): 325-332.
Lu, F. L., C. J. Hsieh, J. L. Caffrey, M. H. Lin, Y. S. Lin, C. C. Lin, M. S. Tsai, W. C. Ho,
P. C. Chen, F. C. Sung and R. S. Lin (2012). "Body mass index may modify asthma
prevalence among low-birth-weight children." Am J Epidemiol 176(1): 32-42.
Luttikhuis, H. O., L. Baur, H. Jansen, V. A. Shrewsbury, C. O'Malley, R. P. Stolk, C. D.
Summerbell and G. Cochrane Heart (2009). "Interventions for treating obesity in
children." Sao Paulo Med J 127(5): 321-321.
Mai, X. M., P. O. Gäddlin, L. Nilsson and I. Leijon (2005). "Early rapid weight gain and
current overweight in relation to asthma in adolescents born with very low birth weight."
Pediatr Allergy Immunol 16(5): 380-385.
Matricardi, P. M., C. Grüber, U. Wahn and S. Lau (2007). "The asthma-obesity link in
childhood: open questions, complex evidence, a few answers only." Clin Exp Allergy
37(4): 476-484.
McConnell, R., T. Islam, K. Shankardass, M. Jerrett, F. Lurmann, F. Gilliland, J.
Gauderman, E. Avol, N. Künzli, L. Yao, J. Peters and K. Berhane (2010). "Childhood
incident asthma and traffic-related air pollution at home and school." 118(7): 1021-1026.
Michelson, P. H., L. W. Williams, D. K. Benjamin and A. E. Barnato (2009). "Obesity,
inflammation, and asthma severity in childhood: data from the National Health and
Nutrition Examination Survey 2001-2004." Ann Allergy Asthma Immunol 103(5): 381-
385.
108
Mokdad, A. H., E. S. Ford, B. A. Bowman, W. H. Dietz, F. Vinicor, V. S. Bales and J. S.
Marks (2003). "Prevalence of obesity, diabetes, and obesity-related health risk factors,
2001." JAMA 289(1): 76-79.
Murray, C. S., D. Canoy, I. Buchan, A. Woodcock, A. Simpson and A. Custovic (2011).
"Body mass index in young children and allergic disease: gender differences in a
longitudinal study." CORD Conference Proceedings 41(1): 78-85.
Musaad, S. M., T. Patterson, M. Ericksen, M. Lindsey, K. Dietrich, P. Succop and G. K.
Khurana Hershey (2009). "Comparison of anthropometric measures of obesity in
childhood allergic asthma: central obesity is most relevant." CORD Conference
Proceedings 123(6): 1321-1321e1312.
Must, A., J. Spadano, E. H. Coakley, A. E. Field, G. Colditz and W. H. Dietz (1999).
"The disease burden associated with overweight and obesity." JAMA 282(16): 1523-
1529.
Ogden, C. L., M. D. Carroll, L. R. Curtin, M. A. McDowell, C. J. Tabak and K. M. Flegal
(2006). "Prevalence of overweight and obesity in the United States, 1999-2004." JAMA
295(13): 1549-1555.
Paraponaris, A., B. Saliba and B. Ventelou (2005). "Obesity, weight status and
employability: empirical evidence from a French national survey." Econ Hum Biol 3(2):
241-258.
Pi-Sunyer (1991). "Health implications of obesity." Am J Clin Nutr 53(6 Suppl): 1595S-
1603S.
Powell, L. M. and F. J. Chaloupka (2009). "Food prices and obesity: evidence and policy
implications for taxes and subsidies." Milbank Q 87(1): 229-257.
Powell, L. M., E. Han and F. J. Chaloupka (2010). "Economic contextual factors, food
consumption, and obesity among U.S. adolescents." J Nutr 140(6): 1175-1180.
Rana, A. R., M. P. Michalsky, S. Teich, J. I. Groner, D. A. Caniano and D. P. Schuster
(2009). "Childhood obesity: a risk factor for injuries observed at a level-1 trauma center."
J Pediatr Surg 44(8): 1601-1605.
Reilly, J. J. and J. Kelly (2010). "Long-term impact of overweight and obesity in
childhood and adolescence on morbidity and premature mortality in adulthood:
systematic review." Int J Obes (Lond) 35(7): 891-898.
Reilly, J. J., E. Methven, Z. C. McDowell, B. Hacking, D. Alexander, L. Stewart and C. J.
Kelnar (2003). "Health consequences of obesity." Arch Dis Child 88(9): 748-752.
109
Robinson, T. (2001). "Television viewing and childhood obesity." Pediatr Clin North Am
48(4): 1017-1025.
Roland, R. S. (2002). "The effects of obesity, smoking, and drinking on medical
problems and costs." Health Aff (Millwood) 21(2): 245-253.
Saha, C., M. E. Riner and G. Liu (2005). "Individual and neighborhood-level factors in
predicting asthma." Arch Pediatr Adolesc Med 159(8): 759-763.
Sibbald, B., H. R. Anderson and S. McGuigan (1992). "Asthma and employment in
young adults." Thorax 47(1): 19-24.
Sorof, J. and S. Daniels (2002). "Obesity hypertension in children: a problem of epidemic
proportions." Neurology Today 40(4): 441-447.
Strain, G. W. and B. Zumoff (1992). "The relationship of weight-height indices of
obesity to body fat content." J Am Coll Nutr 11(6): 715-718.
Sturm, R. and A. Datar (2005). Body mass index in elementary school children,
metropolitan area food prices and food outlet density. Public Health. England. 119: 1059-
1068.
Tounian, P., Y. Aggoun, B. Dubern, V. Varille, B. Guy-Grand, D. Sidi, J. P. Girardet and
D. Bonnet (2001). "Presence of increased stiffness of the common carotid artery and
endothelial dysfunction in severely obese children: a prospective study." 358(9291):
1400-1404.
U.S. Department of Health and Human Services (2008). Physical Activity, Guidelines
Adiisory, Committee Report, 2008. Retrieved August 15, 2012 from
http://www.health.gov/paguidelines/report/pdf/committeereport.pdf
Vartanian, L. R., M. B. Schwartz and K. D. Brownell (2007). "Effects of soft drink
consumption on nutrition and health: a systematic review and meta-analysis." Am J
Public Health (N Y) 97(4): 667-675.
von Mutius, E., T. Nicolai and F. D. Martinez (1993). "Prematurity as a risk factor for
asthma in preadolescent children." J Pediatr 123(2): 223-229.
Wang, G. and W. H. Dietz (2002). "Economic burden of obesity in youths aged 6 to 17
years: 1979-1999." Pediatrics 109(5): E81-E81.
Wang, Y., M. A. Beydoun, L. Liang, B. Caballero and S. K. Kumanyika (2008). "Will all
Americans become overweight or obese? estimating the progression and cost of the US
obesity epidemic." Obesity (Silver Spring) 16(10): 2323-2330.
110
Zhang, Z., H. J. Lai, K. A. Roberg, R. E. Gangnon, M. D. Evans, E. L. Anderson, T. E.
Pappas, D. F. Dasilva, C. J. Tisler, L. P. Salazar, J. E. Gern and R. F. Lemanske (2010).
"Early childhood weight status in relation to asthma development in high-risk children."
CORD Conference Proceedings 126(6): 1157-1162.
Zimmerman, F. J. and J. F. Bell (2010). "Associations of television content type and
obesity in children." Am J Public Health (N Y) 100(2): 334-340.
Abstract (if available)
Abstract
OBJECTIVES: (1) To examine the trends of childhood obesity and changes in body mass index (BMI) over time, (2) To analyze the contemporaneous/lagged association between obesity and asthma, and (3) To analyze the contemporaneous/lagged causal effect of obesity on asthma by implementing instrumental variable method to control for endogeneity bias. ❧ Methods: A Cohort of young and adolescent children was identified from the Early Childhood Longitudinal Study Kindergarten Class of 1998-99 (ECLS-K) data from the U.S. Department of Education. A secondary data source on food prices has been used to identify valid instruments for obesity to reflect the price variations across different food groups within specific geographical unit or region. Changes in BMI obtained from the actual measured height and weight and self-reported asthma status were analyzed from a longitudinal data from 3rd to 8th grade for an associative as well as a causal effect using bivariate probit models. ❧ RESULTS: Two price indices for healthy and unhealthy food were identified as valid IVs. The marginal effects for childhood obesity were consistent through all grade levels and different BMI thresholds. The results indicated that childhood obesity increased the probability of being diagnosed with asthma and the harmful effects of childhood obesity on asthma increased as children got older. Obese children in 3rd grade had a higher probability of being asthmatic where childhood obesity increased the probability of asthma by 0.10 percentage points while holding other covariates constant and using a standard 95% BMI threshold was used to define childhood obesity. Similarly, being obese in 5th and 8th grade increased the probability by 0.14 and 0.19 percentage points. There was an increase of 0.02 in marginal effect for obesity with using a 97th percentile BMI threshold for 3rd grade children while holding other covariates constant at their mean. In 5th and 8th grade children, the marginal effects of obesity on asthma were increased by 4% and 3%, respectively by increasing the BMI threshold from 95th percentile to 97th percentile. Attributable fraction of obesity on asthma which is the same as the absolute impact of obesity on asthma in our study showed that 13.0% of the increase in asthma between 3rd and 5th grade were explained by the increase in obesity while 18.4% of the increase in asthma between 5th and 8th grade were explained by the increase in obesity. ❧ CONCLUSION: This paper demonstrated that childhood obesity was endogenously determined with asthma. Gradual increases in mean BMI and prevalence of childhood obesity along with prevalence of asthma were observed in this study. Children who were exposed to obesity at any point in time during their kindergarten years were at higher risk for developing asthma compared to those children who were never obese during the same time. This may warrant not only a contemporaneous effect of childhood obesity on asthma but also a lagged effect of obesity status on developing asthma later in life. This study focused on analyzing the causality of childhood obesity on asthma while controlling for not only observable but also unobservable factors using the instrumental variables approach. Through this approach, the study determined that childhood obesity significantly increases the risk for developing asthma especially in boys and African American children. We suspected that the instrumental variables approach was a superior method to investigate a causal relationship between childhood obesity and asthma over panel analyses due to infrequent changes in one’s weight or obesity and asthma status.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Economic aspects of obesity
PDF
Discriminating changes in health using patient-reported outcomes
PDF
Effects of stress and the social environment on childhood asthma in the children' s health study
PDF
Understanding primary nonadherence to medications and its associated healthcare outcomes: a retrospective analysis of electronic medical records in an integrated healthcare setting
PDF
Burden of illness in hemophilia A: taking the patient’s perspective
PDF
Healthcare provider recommendations: a panacea to improving influenza vaccination rates?
PDF
Air pollution and childhood obesity
PDF
Impact of pharmacy-based transitional care on healthcare utilization and costs
PDF
Influences of specific environmental domains on childhood obesity and related behaviors
PDF
Autonomic and metabolic effects of obstructive sleep apnea in childhood obesity
PDF
Metabolic consequences of obesity-associated inflammation during puberty and perinatal development
PDF
A cohort study of air-pollution and childhood obesity incidence
PDF
Effects of a formulary expansion on the use of atypical antipsychotics and health care services by patients with schizophrenia in the California Medicaid Program
PDF
Three essays on comparative economic and clinical outcomes of antipsychotics using retrospective claims data
PDF
Essays in pharmaceutical and health economics
PDF
Essays on the use of microsimulation for health and economic policy analysis
PDF
Advances and applications for economic evaluation methods in health technology assessment (HTA)
PDF
Assessment of the impact of second-generation antipscyhotics in Medi-Cal patients with bipolar disorder using panel data fixed effect models
PDF
Effects of near-roadway air pollution exposure on obesity, obesity-related behavior, and neurobehavioral deficits during peripuberty
PDF
Native American ancestry among Hispanic Whites is associated with higher risk of childhood obesity: a longitudinal analysis of Children’s Health Study data
Asset Metadata
Creator
Chung, Janice Kayla
(author)
Core Title
The causal-effect of childhood obesity on asthma in young and adolescent children
School
School of Pharmacy
Degree
Doctor of Philosophy
Degree Program
Pharmaceutical Economics and Policy
Publication Date
06/27/2013
Defense Date
05/01/2013
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
asthma,childhood obesity,OAI-PMH Harvest
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Nichol, Michael B. (
committee chair
), McCombs, Jeffrey (
committee member
), Sood, Neeraj (
committee member
)
Creator Email
janice0405@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-280405
Unique identifier
UC11293583
Identifier
etd-ChungJanic-1709.pdf (filename),usctheses-c3-280405 (legacy record id)
Legacy Identifier
etd-ChungJanic-1709-1.pdf
Dmrecord
280405
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Chung, Janice Kayla
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
childhood obesity