Levy jordan thesis

advertisement
Obesity and Mental Health in Adolescents
Jordan Levy
Honors thesis completed in partial fulfillment of the
requirements of the Honors Program in the
Psychological Sciences
Under the direction of David Schlundt, Ph.D.
Vanderbilt University
April 2012
Abstract
The study examined the relationship between obesity and depression, conduct
disorder, and substance abuse in adolescents using data from the 2008 National
Longitudinal Study of Adolescent Health. Depression at baseline was associated with
higher overall z-BMI over time, and display of antisocial behavior at baseline was
associated with higher overall z-BMI at baseline but lower overall z-BMI at wave 4.
There was no significant association between substance abuse and obesity. Risk factors
for adult obesity include being male, having little social support, being sedentary, being
Latino, being black, being American-Indian, and perceiving oneself as either overweight
or very overweight at baseline. Protective factors against adult obesity include being
Asian, perceiving oneself as being at the right weight at baseline, trying to gain or
maintain weight, eating two or more servings of vegetables daily, and having a positive
outlook on life. The results of the study suggest that mental health interventions as well
as nutritional and physical activity interventions may be needed in the primary prevention
of obesity.
2
Introduction
Obesity has become a growing concern among adolescents in the United States in
recent years. Between 1976 and 1980, 5% of adolescents ages 12-19 were obese. By
2008, obesity in the same age group had risen to 18.1% (Ogden & Carroll, 2010). Obesity
in adolescents elevates the risk for future development of health problems and disease
(Ogden & Carroll, 2010), so the increased rates of obesity in U.S. adolescents are a grave
cause of concern.
One reason that childhood obesity is so foreboding is that overweight children are
very likely to become obese adults, and adolescents who become overweight early in life
are generally more obese as adults than those who become obese as adults (Freedman et
al., 2001). Overweight children are more at risk for adult morbidity and mortality
(Freedman et al., 2001), demonstrating the severity of the problem.
Obese adolescents face an increased risk of both physical and psychosocial health
complications. Obese adolescents typically mature earlier than non-overweight children,
increasing their risk for development of eating disorders later in life (Dietz, 1998).
Obesity is also linked with increased blood lipids, glucose intolerance and diabetes, high
concentrations of liver enzymes, hypertension, sleep apnea, and polycystic ovary
syndrome (Dietz, 1998). Obese adolescents are often victims of verbal harassment about
body weight, which can lead to low body satisfaction, low self-esteem, high depressive
symptoms, and thoughts of suicide (Eisenberg, 2003). They also typically develop a
negative self-image that persists into adulthood (Dietz, 1998).
Aside from being a time of risk for the development of obesity, adolescence is a
period marked by many developmental changes. These changes are often accompanied
3
by feelings of distress and unhappiness, which can be manifested as depression. The
reported percentage of adolescents who have been diagnosed with depression varies
widely, from near 0% in a large nonclinical study of children 10-11 years old to 57% in a
clinical sample of children ages 8-13 (Peterson et al., 1993). Many of the pressures of
adolescent life can greatly contribute to the development of depression. For example,
poor body image and low self-esteem, poor relationships with parents, stressful life
events, difficulty in school, and low popularity among peers are issues faced by many
adolescents that can contribute to the development of depression.
Another mental health disorder that adolescents face is conduct disorder, which is
marked by the display of chronic, severe antisocial behavior that typically begins in early
childhood and can extend into adulthood (Robins & Ratcliff, 1979). As described in the
DSM-IV, conduct disorder is typically manifested in any of four categories: aggression to
people or animals, destruction of property, deceitfulness or theft, and serious violation of
rules (Searight et al., 2001). The prevalence of conduct disorder is estimated at 2-9% for
girls and 6-16% for boys (Searight et al., 2001). Conduct disorder presents a serious
problem to society because it has been shown to increase the risk of many public health
concerns, including violence, use of weapons, teenage pregnancy, substance abuse, and
dropping out of school (Searight et al., 2001). Conduct disorder in adolescence can
develop into antisocial personality disorder in adulthood (Holmes et al., 2001), furthering
the risk that the disorder places on society.
Finally, early adolescence is the most common time for individuals to experiment
with drugs, with peak drug use typically occurring between the ages of 18-22 (Kandel
and Logan, 1984). According to the 2006 National Survey on Drug Use and Health, it is
4
estimated that 8% of adolescents meet the criteria set by the American Psychological
Association for substance abuse or dependence, 5% meet the criteria for alcohol abuse or
dependence, and at least 11% show signs of “problematic use” of alcohol or drugs.
Substance abuse among adolescents is especially problematic because it has been linked
to a plethora of other issues. Studies have demonstrated a significant association between
substance abuse and suicidal behavior (Crumley, 1990), affective disorders, conduct and
antisocial personality disorder, attention-deficit hyperactivity disorders, and anxiety
disorders (Bukstein et al., 1989). Consequently, it is evident that although substance
abuse is a problem in itself, it can also have a profound impact on the development of
many other psychiatric disorders.
Existing research tends to demonstrate that there is a relationship between obesity
and depression, conduct disorders and substance abuse in adolescents. However, research
has yielded conflicting results, and as a result, the direction of causation is unknown for
each of these variables.
Some studies suggest that depressed adolescents are at increased risk for the
development of obesity (Goodman and Whitaker, 2002; Erermis et al., 2004), while
others found no relationship between adolescent depression and obesity later in life
(Bardone et al., 1998; Brewis, 2003; Eisenberg et al., 2003). Still others found that
obesity was associated with fewer feelings of depression (Isnard et al., 2003), so the true
relationship between obesity and depression in adolescents is very unclear.
Research on the relationship between obesity and conduct disorder in adolescents
tends to show that antisocial behavior is associated with greater overall BMI, but very
few studies have been done on the subject. Studies have shown that conduct disorder is
5
related to having a higher BMI (Pine et al., 1997), with one study showing that females
with conduct disorders likely to have BMIs 0.23-0.33 units higher than females without
conduct disorders, and males with conduct disorder likely to have BMI .20 units higher
than males without conduct disorder (Anderson et al., 2006). However, since there are
very few studies in existence on the relationship between obesity and conduct disorder in
adolescents, more research is necessary to corroborate these findings.
Finally, existing research on the relationship between obesity and substance abuse
in adolescents tends to be very conflicted. Volkow and Wise (2005) suggest that both
drugs and food can activate the same reward pathway in the brain, meaning that
individuals who suffer from problems with substance abuse may also be more likely to
have problems with food. However, studies that actually examine this relationship have
demonstrated that obese adults are actually less likely to suffer from issues of substance
abuse than normal-weight adults (Simon et al., 2006). Among adolescents, the
relationship is even less clear, as the few studies that exist have also yielded contradictory
results: though it has been shown that obesity is associated with an approximately 25%
lower odds of substance use disorders in adults (Simon et al., 2006), in adolescents,
different studies have shown both no relationship between obesity and substance abuse as
well as a positive relationship between obesity and substance abuse in females. As a
result, significantly more research is necessary in order to provide more conclusive data
about the relationship between obesity and substance abuse in adolescents.
Of additional interest are the gender differences that appear with obesity and each
of the mental health issues. Obesity was significantly associated with increased
symptoms of depression (Erickson et al., 2000) in females, but these effects were less
6
significant or nonexistent in males. These gender differences in symptomatology seem to
be indicative of gender differences in self-perception as overweight and body
dissatisfaction. In females, increased BMI was significantly associated with increased
weight concerns (Erickson et al., 2000), which were associated with increased depressive
symptoms. As a result, it seems plausible that obese females are more susceptible to
depression than obese males because of body pressures created by American society and
the body dissatisfaction that these expectations can provoke. Males seem to be somewhat
shielded from these body pressures, and as a result, obese males experience less
depression associated with poor body image.
For the most part, the abovementioned studies demonstrate that although a
relationship likely exists between each of the mental health issues and obesity in
adolescents, it remains unclear what the direction of the relationship is. Additionally,
existing studies have not demonstrated whether obesity predisposes adolescents to these
conditions or if adolescents who have these mental health disorders are more likely to
become obese. Consequently, there is a great need for further research on the direction of
causality between obesity and depression, conduct disorders and substance abuse in
adolescents.
As a result, my research project focuses on these areas of research. In my study, I
performed statistical analysis on the 2008 National Longitudinal Study of Adolescent
Health (Add Health) data set in order to investigate the relationship between obesity,
depression, conduct disorder, and substance abuse in adolescents. I expected to not only
demonstrate a positive relationship between these variables, but that these psychological
disorders can be predictive factors for the development of obesity in adolescents.
7
Research Hypotheses
1. There will be a positive relationship between obesity and depression, conduct
disorder, and substance abuse in adolescents.
2. Depression, conduct disorder, and substance abuse precede the development of obesity
in adolescents.
Method
Participants
Participants in the study consisted of subjects of the 2008 Add Health study,
which initially enrolled children in grades 7-12 in 1994 and followed them longitudinally
for 14 years. There are four waves of data collection available, and data on depression,
substance abuse, conduct disorders, height and weight were measured at each wave by
self-report data, questionnaires, and personal interviews.
A sample of 80 high schools and 52 middle schools from the U.S. was selected
with unequal probability of selection, ensuring that the sample was representative of US
schools with respect to region of the country, urbanicity, school size, school type, and
ethnicity (Harris et. al, 2009). The core sample consisted of the nationally representative
sample of 12,105 adolescents. A wide range of additional subsamples was drawn to
provide meaningful samples of children who were disabled, blacks from well-educated
families, Chinese, Cuban, Puerto Rican, adolescents residing together, siblings of various
levels of relationship (twins, full siblings, half-siblings, nonrelated adolescents residing
together, siblings of twins). These subsamples added up to a survey of more than 20,000
youth in Add Health Wave I (Harris et. al, 2009). During Wave II, the sample was 31.0%
8
white non-Hispanic, 21.8% black non-Hispanic, 0.8% American Indian non-Hispanic,
15.7% Hispanic, 8.8% Mexican, 2.5% Cuban, 3.0% Puerto Rican, 1.5% Central/South
American, 1.3% Other Hispanic, 6.8% Asian, 2.8% Filipino non-Hispanic, 1.8% Chinese
non-Hispanic, and 2.2% Asian non-Hispanic (Popkin and Udry, 1998).
Students in the study filled out a brief questionnaire at school, followed by
extensive at-home interviews with students and their parents conducted by researchers.
These interviews were repeated one year later, again at ages 18-26, and finally at ages 2432. School administrators provided information regarding participants’ schools, and
existing data were compiled to describe neighborhoods and communities. Information
from high school transcripts is also available to the study if participants allowed.
Pregnant women were excluded from the study, as were disabled children who used a
walking-aid device.
Procedure
I first downloaded the appropriate data files corresponding to each wave of the
study using SPSS (Statistical Package for the Social Sciences) and identified the variables
I needed to answer my research questions. I then created a data set and merged the four
waves of data, linking the files using participant identification codes. I transformed height
and weights into age and gender standardized body mass index, or z-BMI. Z-BMI
adjusts measured body mass index against the World Health Organization norms for
gender and age since BMI naturally changes as children get older. A z-BMI of zero
would reflect the expected mean value given the child’s age and gender. The value above
and below the mean is in standard deviation units that also control for natural changes
over time.
9
I then created scales in SPSS to correspond with each psychological disorder for
each wave of the study, paying careful attention to how each of the above psychological
disorders were assessed, and including measures of diagnosis as well as symptom
severity.
Once the data set was prepared, I learned about using mixed linear models to
analyze nested data sets. In this case, longitudinal observations were nested within
participants. I developed z-BMI trajectory models, and saw how individual
psychopathology scores related to the slopes and intercepts of these models.
I then prepared and described the data using SPSS and created mixed linear
models using the HLM (hierarchical linear models) program. Individual difference
models controlled for demographic and socioeconomic differences. I created a linear
model of BMI change, then examined how individual difference variables modified the
intercept or slope of this model. After controlling for individual differences in
demographics and socioeconomic status, I used the mental health variables one at a time
to test hypotheses about their relationship to the slopes and intercept of the z-BMI
equation.
Results
Sample Demographics
The sample size at baseline was 6,291 individuals (see Table 1, Table 2, and
Table 3) 49.1% of subjects were male and 50.9% were female. The sample was 66.6%
white, 24.5% black, 11.3% Hispanic, 3.6% American Indian, 4.2% Asian and 6.5% other
ethnicity, and ranged from 11-19 years of age at wave 1. The mean z-BMI of the sample
10
was 0.33 (S.D. = 1.022), meaning that they were slightly overweight. The participants
came from predominantly educated families, with the majority of parents having
completed high school or earned a GED (at least 88.9% of mothers and at least 89.1% of
fathers). The mean annual household income was $47,700 (S.D. = 56.355), with 22.1%
of participants’ families earning $0-24,999 annually, 25.1% between $25,000-49,999,
17.3% earning between $50,000-74,999, 6.3% earning between $74,000-99,999, and
14.3% earning $100,000 or more annually.
Male z-BMI increased all four waves across all races (Table 4, Figure 1).
American Indian males had the highest overall z-BMI scores (0.585, 0.652, 1.030 and
1.255 at waves 1 through 4, respectively), followed by males of other ethnicities (0.416,
0.444, 0.900, 1.38), Hispanic males (0.451, 0.448, 0.857, 1.349), black males (0.437,
0.431, 0.803, 1.044), white males (0.343, 0.366, 0.759, 1.057) and Asian males (0.198,
0.288, 0.711, 0.879). American Indian males had the highest z-BMI scores at waves 1
through 3, but at wave 4 were surpassed by males of other ethnicities and Hispanic males.
Asian males had the lowest z-BMI scores at all four waves.
Female z-BMI also increased across all races (Table 5, Figure 2). American
Indian females had the highest overall z-BMI scores (0.563, 0.507, 0.894 and 1.368 at
waves 1 through 4, respectively), followed by black females (0.530, 0.539, 0.776, 1.289),
females of other ethnicities (0.344, 0.435, 0.638, 1.082), Hispanic females (0.385, 0.379,
0.607, 1.053), white females (0.227, 0.230, 0.502, 0.924) and Asian females (-0.094, 0.082, 0.202, 0.632). American-Indian females had the highest z-BMI at wave 1, were
slightly surpassed by black females at wave 2, then returned to the highest z-BMI scores
for waves 3 and 4. Asians had the lowest z-BMI scores at all four waves.
11
Mixed Models Analysis
Using the Hierarchical Linear Models (HLM) program, a model was developed to
represent change in z-BMI among the participants. The level-1 model was a simple linear
equation that expressed change in z-BMI over time. The equation is:
Z_BMIti = π0i + π1i*(WAVEti) + eti
Where π0i is the intercept, π1i is the slope, and eti is the error.
The level-2 intercept as outcomes model was:
π0i = β00 + β01*(SEXi) + β02*(SOCIALSUi) + β03*(SEDENTARi) +
β04*(LATINOi) + β05*(ISBLACKi) + β06*(ISINDIANi) + β07*(ISASIANi)
+ β08*(ISOTHERi) + r0i,
The level-2 slopes as outcomes model was:
π1i = β10 + β11*(DEPRESSIi) + β12*(ANTISOCIi) + β13*(RIGHTWEIi) +
β14*(OVERWEIGi) + β15*(VERYOVERi) + β16*(LOSEWEIGi) +
β17*(EATVEGGIi) + β18*(POSITIVEi) + r1i.
This means that the variables found to affect the coefficient π0i are sex (p<0.001), level
of social support (p=0.017), sedentary level (p=0.001), being Latino (p=0.025), being
black (p<0.001), being Indian (p<0.001), and being Asian (p=0.002). Being of another
ethnicity was found to not be statistically significant (p=0.080).
The linear slope π1i is affected by the variables depression (p=0.035), antisocial
tendencies (p=0.002), perceiving oneself as the “right weight” at baseline (p<0.001),
perceiving oneself as “overweight” at baseline (p<0.001), “perceiving oneself as “very
overweight” at baseline (p<0.001), attempting to change one’s weight at baseline
(p<0.001), how many servings of vegetables participants ate each day (p=0.002), and
12
having a positive outlook on life (p=0.034). Consequently, the mixed model was:
Z_BMIti = β00 + β01*SEXi + β02*SOCIALSUi + β03*SEDENTARi +
β04*LATINOi + β05*ISBLACKi + β06*ISINDIANi +β07*ISASIANi +
β08*ISOTHERi + β10*WAVEti + β11*DEPRESSIi*WAVEti +
β12*ANTISOCIi*WAVEti + β13*RIGHTWEIi*WAVEti +
β14*OVERWEIGi*WAVEti + β15*VERYOVERi*WAVEti +
β16*LOSEWEIGi*WAVEti + β17*EATVEGGIi*WAVEti +
β18*POSITIVEi*WAVEti + r0i + r1i*WAVEti + eti.
Intercepts
The sex coefficient is -0.249, meaning that men have overall higher z-BMI scores
than women (Table 6). The male overall z-BMI was 0.5877 (S.D.=0.989) and the female
overall z-BMI was 0.5075 (S.D.=0.911) (Table 7, Figure 3).
The social support coefficient is -0.005, meaning that individuals with greater
levels of social support had lower z-BMI scores. The lower 25th percentile of social
support had the highest z-BMI at 0.597 (S.D.=0.967), the 25-50th percentile had a z-BMI
of 0.552 (S.D.=0.941), the 50-75th percentile had a z-BMI of 0.517 (S.D.=0.947) and the
upper quartile had a z-BMI of 0.521 (S.D.=0.947) (Table 8, Figure 4).
The sedentary coefficient is 0.001, meaning that individuals who were more
sedentary had higher z-BMIs than those who were less sedentary. The lower quartile of
sedentary activity had an overall z-BMI of 0.463 (S.D.=0.938), the 25-50th percentile had
a z-BMI of 0.511 (S.D.=0.935), the 50-75th percentile had a z-BMI of 0.554
(S.D.=0.938), and the upper quartile had a z-BMI of 0.664 (S.D.=0.982) (Table 9, Figure
5).
13
The Latino coefficient is 0.093, meaning that Latinos have higher z-BMI scores
than non-Latinos. The overall Latino z-BMI score is 0.635 (S.D.=0.923) and the overall
non-Latino z-BMI is 0.536 (S.D.=0.954) (Table 10, Figure 6).
The black coefficient is 0.211, meaning that black individuals had higher z-BMI
scores than non-blacks. The overall z-BMI for blacks was 0.678 (S.D.=0.962) and the
overall z-BMI for non-blacks was 0.504 (S.D.=0.94361) (Table 11, Figure 7).
The American Indian coefficient is 0.241, meaning that American Indian
individuals had higher z-BMI scores than non-American-Indians. The overall z-BMI for
American-Indians was 0.785 (S.D.=1.021) and the overall z-BMI for non-AmericanIndians was 0.538 (S.D.=0.947) (Table 12, Figure 8).
The Asian coefficient is -0.165, meaning that Asian individuals had lower z-BMI
scores than non-Asian individuals. The overall z-BMI for Asians was 0.261 (S.D.=0.983)
and the overall z-BMI for non-Asians was 0.559 (S.D.=0.948) (Table 13, Figure 9).
Wave Slope
The depression coefficient was found to be -0.001, meaning that being depressed
at baseline resulted in a higher overall z-BMI at wave 4. At wave 1, the upper quartile of
depression had the highest z-BMI at 0.375 (S.D.=1.040), followed by the 50-75%
percentile at 0.348 (S.D.=1.017), the 25-50th percentile at 0.314 (S.D.=1.018) and the
lower quartile at 0.294 (S.D.=1.011) (Table 14, Figure 10). By wave 4, the 50-75th
percentile had the highest z-BMI at 1.075 (S.D. 1.048), followed by the upper quartile at
1.061 (S.D.=1.057), the 25-50th percentile at 0.992 (S.D.=1.036), and the lower quartile at
0.972 (S.D.=1.036). The upper quartile and 50-75th percentile had the highest overall zBMI scores at all four waves, and the 25-50th percentile and the lower quartile had the
14
lowest overall z-BMI scores at all four waves. There was no easily discernible difference
in slope between any of the four quartiles, though the negative coefficient implies a slight
tendency for effect of depression to diminish over time.
The antisocial coefficient, which measures diagnostic criteria for conduct
disorder, is -0.001, meaning that displaying more antisocial tendencies at baseline tended
to result in a lower overall z-BMI at wave 4. At wave 1, the upper quartile of antisocial
behavior had the highest z-BMI at 0.384 (S.D.=1.001), followed by the 50-75th percentile
at 0.351 (S.D.=1.015), the lower quartile at 0.301 (S.D.=1.035) and the 25-50th percentile
at 0.296 (S.D.=1.035) (Table 15, Figure 11). By wave 4, the lower quartile had the
highest z-BMI at 1.064 (S.D.=1.036), followed closely by the 50-75th percentile at 1.063
(S.D.=1.014), the upper quartile at 0.989 (S.D.=1.071) and the 25-50th percentile at 0.974
(S.D.=1.057). While the upper quartile had the highest overall z-BMI at wave 1, they had
the second lowest z-BMI by wave 4. In contrast, the lower quartile had close to the
lowest overall z-BMI at wave 1, but by wave 4 they had the highest overall z-BMI scores.
The coefficient for perceives self as at the “right weight” at baseline is 0.091.
Perceiving oneself as being at the “right weight” at baseline is predictive of having a
lower z-BMI at wave 4. Participants who felt that they were at the right weight at
baseline had an overall z-BMI of 0.151 (S.D.=0.787) and participants who felt that they
were not at the right weight had an overall z-BMI of 0.536 (S.D.=1.201) (Table 16,
Figure 12). At wave 4, those who perceived themselves as the right weight had a z-BMI
of 0.957 (S.D.=0.957) and those perceived themselves as not being the right weight had a
z-BMI of 1.194 (S.D.=1.108). Those who felt that they were not at the right weight had
the highest z-BMI scores at all four waves, and there was no discernable difference
15
between their slope and the slope of those who perceived themselves as the right weight,
though the positive coefficient implies that the protective effect decreases slightly with
time.
The coefficient for perceives self as “overweight” at baseline is 0.215, which
demonstrates that perceiving oneself as being “overweight” at baseline is predictive of
having a significantly higher z-BMI at wave 4. Participants who felt that they were
overweight at baseline had an overall z-BMI of 1.078 (S.D.=0.739) and those who felt
that they were not overweight at baseline had a z-BMI of 0.052 (S.D.=0.972) (Table 17,
Figure 13). By wave 4, those who perceived themselves as overweight had a z-BMI of
1.615 (S.D.=0.772) and those who did not perceive themselves as overweight had a zBMI of 0.793 (S.D.=1.047). Both those who perceived themselves as overweight and
those who did not followed similar trajectory patterns, but those who perceived
themselves as overweight at baseline had significantly higher z-BMI scores at all four
waves.
The coefficient for perceives self as “very overweight” at baseline is 0.301, which
demonstrates that perceiving oneself as being “very overweight” at baseline is predictive
of having a significantly greater z-BMI at wave 4. Participants who felt that they were
very overweight at baseline had an overall z-BMI of 1.755 (S.D.=0.679) and those who
did not feel that they were very overweight had an overall z-BMI of 0.284 (S.D.=0.996)
(Table 18, Figure 14). By wave 4, those who perceived themselves as very overweight
had a z-BMI of 1.999 (S.D.=0.655) and those who did not perceive themselves as very
overweight had a z-BMI of 0.990 (S.D.=1.040). Those who perceived themselves as very
overweight at baseline had significantly higher z-BMI scores than those who did not at all
16
four waves. Both those who perceived themselves and those who did not had similar
slopes between waves 1 and 2, but between waves 3 and 4 those who did not perceive
themselves as very overweight increased in z-BMI much more than those who did, who
showed a more gradual increase.
The coefficient for attempts to change weight is 0.043, meaning that those who
were trying to lose weight at baseline had a higher z-BMI by wave 4, those who were
trying to maintain their weight had a lower z-BMI and those who were trying to gain
weight at baseline had an ever lower z-BMI by wave 4. Participants who were trying to
gain weight at baseline had an overall z-BMI of -0.121 (S.D.=0.996) at baseline, those
who were trying to maintain their weight had a z-BMI of 0.157 (S.D.=0.917), and those
who were trying to lose weight had an overall z-BMI of 0.943 (S.D.=0.841) (Table 19,
Figure 15). By wave 4, those who were trying to gain weight had a mean z-BMI of 0.623
(S.D.=1.080), those who were trying to maintain their weight had an overall z-BMI of
0.904 (S.D.=1.009) and those who were trying to lose weight had a mean z-BMI of 1.509
(S.D.=0.841). Those who were trying to lose weight had the highest overall z-BMIs at all
four waves and those who were trying to gain weight had the lowest z-BMI scores at all
four waves. All three groups showed similar slopes.
The coefficient for servings of vegetables eaten daily is -0.012, meaning eating
more servings of vegetables daily at baseline is predictive of having a lower overall zBMI at wave 4. Participants who ate zero servings of vegetables daily had an overall zBMI of 0.380 (S.D.=1.030) at baseline, those who ate one serving of vegetables daily had
an overall z-BMI of 0.311 (S.D.=1.03) at baseline, and those who ate two or more
servings of vegetables daily had a z-BMI of 0.310 (S.D.=1.001) at baseline (Table 20,
17
Figure 16). By wave 4, those who ate zero servings of vegetables daily had an overall zBMI of 1.098 (S.D.=1.051), those who ate one serving of vegetables daily had a mean zBMI of 1.029 (S.D.=1.023) and those who ate two or more servings of vegetables daily
had an overall z-BMI of 0.939 (S.D.=1.061). Individuals who ate zero servings of
vegetables daily had the highest overall z-BMI scores at all four waves, and individuals
who ate two or more servings of vegetables daily had the lowest overall z-BMI scores at
all four waves.
The coefficient for positive outlook on life is 0.003, which means that having a
positive outlook on life at baseline is associated with having a lower overall z-BMI at
wave 4. At wave 1, the 50-75th percentile of positive outlook had the highest overall zBMI at 0.361 (S.D.=1.000), followed by the 25-50th percentile at 0.353 (S.D.=1.033), the
lower quartile at 0.344 (S.D.=1.023) and the upper quartile at 0.268 (S.D.=1.039) (Table
21, Figure 17). By wave 4, the 50-75th percentile remained at the highest overall z-BMI at
1.041 (S.D.=1.022), followed by the 25-50th percentile at 1.036 (S.D.=1.044), the lower
quartile 1.0347 (S.D.=1.071) and the upper quartile at 0.985 (S.D.=1.042). The lower
three quartiles had very similar z-BMI scores at both wave 1 and wave 4, and only
diverged slightly in between. The upper quartile had a lower z-BMI at wave 1 and wave
4, and was only surpassed once at wave 3 by the lower quartile.
Discussion
Hypotheses
The results support the hypothesis that depression is associated with a greater
overall z-BMI. Individuals who scored in the upper two quartiles for depression at wave
18
1 had higher z-BMI scores at all four waves than those who scored in the lower two
quartiles for depression. Since all four quartiles of depression have nearly identical
slopes, depression does not affect the rate of weight gain much at all. Instead, depressed
and less depressed individuals tended to gain weight at the same rate, but depressed
individuals were heavier for the duration of the study. Thus, the hypothesis that
depression in adolescents precedes the onset of obesity is incorrect; depressed individuals
were heavier from the start of the study and did not appear to gain more any more weight
than less depressed individuals.
The hypothesis that conduct disorder is associated with a greater overall z-BMI
was not supported by the data. Individuals who scored in the upper two quartiles for
antisocial behavior, the critical marker of conduct disorder, were heavier at baseline than
the lower two quartiles of antisocial behavior. However, by wave 4, the upper quartile
and the 25-50th percentile had dropped to the lowest z-BMI scores, while the lower
quartile and the 50-75th percentile had the highest z-BMI scores. Thus, individuals who
displayed the most antisocial behavior at baseline had nearly the lowest z-BMIs at wave
4, and those who displayed the least antisocial behavior at baseline had nearly the highest
z-BMIs at wave 4. Consequently, the hypothesis is not correct because antisocial
behavior at baseline tended to result in lower overall z-BMI at wave 4. Additionally, the
hypothesis that conduct disorder precedes the onset of obesity is also incorrect, because
while the most antisocial individuals were the heaviest initially, by wave 4 they were the
least heavy.
19
Finally, the hypotheses that substance abuse is associated with a greater overall zBMI and that substance abuse precedes the onset of obesity are incorrect. Substance
abuse was not found to be a significant factor in predicting z-BMI.
However, through the development of the mixed model for z-BMI, many nonmental-health variables were found to be significant predictors in determining z-BMI.
Being male, having little social support, being sedentary, being Latino, being black, being
American-Indian, and perceiving oneself as either overweight or very overweight at
baseline were all found to be risk factors for having a higher z-BMI at wave 4. Similarly,
being Asian, perceiving oneself as being at the right weight, trying to gain or maintain
weight, eating two or more servings of vegetables daily, and having a positive outlook on
life were all found to be protective factors against having a higher z-BMI at wave 4.
Many of these factors are consistent with current government recommendations to
prevent adolescent obesity, which include getting at least 60 minutes of exercise weekly,
reducing “screen” time in front of the television and computer to less than two hours
daily, and consuming at least five servings of fruits and vegetables daily (NIH).
Overall, z-BMI tended to be fairly stable between waves 1 and 2 before steadily
increasing between waves 3 and 4. During waves 1 and 2, participants were between ages
11-20, were ages 18-26 at wave 3 and were between ages 24-32 at wave 4. Z-BMI
stability between waves 1 and 2 is likely attributed to the single year separating data
collection periods. Weight gain tended to begin around age 18-26 and continued on
through ages 24-32, which is fairly consistent with existing models that show aging as a
contributing factor to weight gain in early adulthood (Roberts and Williamson, 2002).
During the transition period between adolescence and young adulthood, a significant
20
proportion of individuals becomes and remains obese during adulthood (Gordon-Larsen
et al., 2004), and significant weight gain often begins in early 20s (Sheehan et al., 2003).
Existing literature tends to have conflicting results on the relationship between
obesity and many mental health variables, so it is difficult to compare these to the results
of my study. Studies relating obesity and depression have notoriously inconsistent
findings (Wardle et al., 2005). Though a study by Richardson et al. (2003) found that
depression in late adolescence is associated with later obesity in females, another study
by Pine et al. (1997) found that adolescent BMI was inversely related to adult depressive
symptoms in males. The same study also found that BMI was positively related to
adolescent symptoms of conduct disorder, and Mustillo et al. (2003) found that conduct
disorder in adolescent males was associated with chronic obesity, which contradicts the
results of my study. Consequently, the results of my study agree with some existing
literature and contradict the findings of other studies, which is representative of the
inconsistent nature of existing literature on the subject.
An important aspect of the study to note is the small effect sizes of each of the
variables related to z-BMI scores. The study did not find any single variable that is
responsible for large changes in z-BMI score trajectory; rather, z-BMI is affected by
many variables that have small effects in z-BMI level and change over time.
The results of the study have a profound impact on the future prevention of
obesity in adolescents, and demonstrate that additional attention should be paid to certain
mental health variables. Since depressed individuals tend to have higher z-BMI scores
over time than non-depressed individuals, primary prevention of obesity should
incorporate assessment and treatment of depression in adolescents. Similarly, since
21
having a positive outlook is associated with having a lower z-BMI over time, prevention
of obesity should ensure that adolescents feel optimistic about the future. Feelings of
hopelessness and pessimism should be recognized as risk factors and addressed during
adolescence in order to reduce the risk of developing obesity later in life.
In addition to these mental health variables, primary prevention of obesity should
continue to include emphasis on proper nutrition and ample physical activity. Individuals
who perceived themselves as being the right weight at baseline had lower z-BMI scores
over time, while those who perceived themselves as overweight or very overweight had
significantly higher z-BMI scores over time. Additionally, those who were attempting to
lose weight at baseline had much higher z-BMI scores over time than individuals who
were trying to maintain or gain weight. As a result, maintaining a healthy weight during
adolescence can be a potential protective factor against the development of obesity in
adulthood. In accordance with current government recommendations, consuming two or
more servings of vegetables daily can also protect against the development of obesity in
adulthood.
Strengths and Limitations
Unquestionably, the greatest strength of this study is that heights and weights
were measured at each wave, as opposed to being self-reported. This eliminates all
possibility for falsification of height and weight data and means that the dependent
variable is very strong. The time range of the study is also advantageous, as it follows
participants from early adolescence to adulthood, which is when weight gain tends to
begin. Though there was some attrition within the Add Health study, only individuals
with data on at least two waves were included in analysis, so the sample has relatively
22
complete data. Additionally, the sample is very diverse and includes a random sample
from both public and private schools from all regions of the United States.
However, there are several limitations of this study as well. The first limitation is
that, although the dependent variable is very strong, the independent variables are
somewhat less reliable. The Add Health study did not use a clinical assessment, so
measures of depression, conduct disorder, and substance abuse are dependent on scores
on a scale, as opposed to actual diagnoses made by a clinician. While the scales do assess
depressive, antisocial, and substance abuse symptomatology, they are not true
representations of depression, conduct disorder, and substance abuse as actual diagnosed
disorders.
Additionally, the Add Health study contains scales for a limited number of mental
health variables. I had originally wanted to look at the relationship between obesity and
depression, anxiety disorders, conduct disorder, suicidal thoughts, and substance abuse,
but there were insufficient questions to develop a reliable scale for anxiety and suicide.
Future Research Directions
As a result of the limitations of this study, future research should focus on a
clinical sample with individuals who have been diagnosed with depression, conduct
disorder, or substance abuse. It would be worthwhile to investigate if the results of this
study would differ if conducted with participants who meet the diagnostic criteria for
these disorders. Additionally, since there were insufficient questions in the Add Health
study to form reliable scales for anxiety and suicidal ideation, future research should
incorporate these disorders into a clinical sample for assessment. Given the inconclusive
nature of the current research on obesity and mental health disorders, it would be
23
advisable to replicate this study with a clinical sample in order to obtain more data on the
relationship between these variables.
24
References
Anderson, S. E., Cohen, P., Naumova, E.N. & Must, A. (2006). Association of depression
and anxiety disorders with weight change in a prospective community-based
study of children followed up into adulthood. Archives of Pediatrics & Adolescent
Medicine, 160, 285-291.
Bardone, A.M., Moffitt, T.E., Caspi, A. et al. (1998). Adult physical health outcomes of
adolescent girls with conduct disorder, depression, and anxiety. Journal of the
American Academy of Child and Adolescent Psychiatry, 37, 594-601.
Brewis, A. (2003). Biocultural aspects of obesity in young Mexican schoolchildren.
American Journal of Human Biology, 15, 446–460.
Bukstein, O. G., Brent, D.A., & Kaminer, Y. (1989). Comorbidity of substance abuse and
other psychiatric disorders in adolescents. American Journal of Psychiatry, 146,
1131-1141.
Centers for Disease Control and Prevention (2011). Tips for parents: ideas to help
children maintain a healthy weight. [ONLINE] Available at:
http://www.cdc.gov/healthyweight/children/. [Last Accessed 7 April 2012].
Crumley, F. E. (1990). Substance abuse and adolescent suicidal behavior. Journal of the
American Medical Association, 263, 3051-3056.
Dietz, W. H. (1998). Health consequences of obesity in youth: Childhood predictors of
adult disease. Pediatrics, 101, 518-525.
Eisenberg, M. E. & Neumark-Sztainer, D. N. (2003). Associations of weight-based
teasing and emotional well-being among adolescents. Archives of Pediatric
Adolescent Medicine, 157, 733-738.
25
Erermis, S., Cetin, N., Tamar, M., Bukusoglu, N., Akdeniz, F., & Goksen, D. (2004). Is
obesity a risk factor for psychopathology among adolescents? Pediatrics
International, 46, 296–301.
Erickson, S. R., Robinson, T.N., Haydel, F. & Killen, J.D. (2000). Are overweight
children unhappy? Archives of Pediatrics & Adolescent Medicine, 154, 931-935.
Freedman, D. S., L. K. Khan, et al. (2001). Relationship of childhood obesity to coronary
heart disease risk factors in adulthood: The Bogalusa Heart Study. Pediatrics,
108, 712-718.
Goodman, E. and Whitaker, R.C. (2002). A prospective study on the role of depression in
the development and persistence of adolescent obesity. Pediatrics, 110, 497-504.
Gordon-Larsen, P., Adair, L.S., Nelson, M.C., & Popkin, B.M. (2004). Five-year obesity
incidence in the transition period between adolescence and adulthood: the national
longitudinal study of adolescent health. American Journal of Clinical Nutrition,
80, 569-75.
Harris, K.M., Halpern, C.T., Whitsel, E., Hussey, J., Tabor, J., Entzel, P., & Udry, J.R.
(2009). The national longitudinal study of adolescent health: research design
[WWW document]. URL: http://www.cpc.unc.edu/projects/addhealth/design.
Holmes, S. E., Slaughter, J. R., & Kashani, J. (2001). Risk factors in childhood that lead
to the development of conduct disorder and antisocial personality disorder. Child
Psychiatry and Human Development, 31, 183-193.
Isnard, P., Michel, G., Frelut, M.L., Vila, G., Falissard, B., Naja, W., Navarro, J. and
Mouren-Simeoni, M.C. (2003). Binge eating and psychopathology in severely
obese adolescents. International Journal of Eating Disorders, 34, 235–243.
26
Kandel, D.B., Logan, J.A. (1984). Patterns of drug use from adolescence to young
adulthood: periods of risk for initiation, stabilization and decline in use. American
Journal of Public Health, 74, 660-666.
Mustillo, S., Worthman, C., Erkanli, A., Keeler, G., Angold, A., & Costello, E. (2003).
Obesity and psychiatric disorders: developmental trajectories. Pediatrics, 111, 851859.
Ogden, C. and Carroll, M. (2010). Prevalence of obesity among children and adolescents:
United States trends 1963-1965 through 2007-2008. Center for Disease Control,
1-4.
Petersen, et al. (1993). Depression in adolescence. American Psychologist, 48, 155–168.
Robins, L. N. and Ratcliff, K. A. (1979). Risk factors in the continuation of childhood
antisocial behaviour into adulthood. International Journal of Mental Health, 7,
96-116.
Pine, D.S., Cohen, P., Brook, J. & Coplan, J.D (1997). Psychiatric symptoms in
adolescence as predictors of obesity in early adulthood: a longitudinal study.
American Journal of Public Health, 87, 1303-1310.
Popkin, B.M. and Udry, J.R. (1998). Adolescent obesity significantly increases in second
and third generation U.S. immigrants: The national longitudinal study of
adolescent health. The Journal of Nutrition, 128, 701-706.
Richardson, L.P. et al. (2003). A longitudinal evaluation of adolescent depression and
adult obesity. Archives of Pediatrics & Adolescent Medicine, 157, 739-745.
Roberts, S.B. and Williamson, D.F. (2002). Causes of adult weight gain. The Journal of
Nutrition, 132.
27
Searight, H. R., Rottnek, F., & Abby, S.T. (2001). Conduct disorder: diagnosis and
treatment in primary care. American Family Physician, 63, 1579-1589.
Simon, G. E., Korff, M.V., Saunders, K., Miglioetti, D.L., Crane, P.K., van Belle, G. &
Kessler, R.C. (2003). Association between obesity and psychiatric disorders in the
U.S. adult population. Archives of General Psychiatry, 63, 824-830.
Sheehan, T.J., DuBrava, S., Dechello, L.M., & Fang, Z. (2003). Rates of weight change
for black and white Americans over a twenty year period. International Journal of
Obesity and Related Metabolic Disorders, 27, 498–504.
Substance Abuse and Mental Health Services Administration. (2007). Results from the
2006 National Survey on Drug Use and Health: National Findings (Office of
Applied Studies, NSDUH Series H-32, DHHS Publication No. SMA 07-4293).
Rockville, MD.
Volkow, N. D. and Wise, R.A. (2005). How can drug addiction help us understand
obesity? Nature Neuroscience, 8, 555-560.
Wardle, J. et al. (2006) Depression in adolescent obesity: cultural moderators of the
association between obesity and depressive symptoms. International Journal of
Obesity, 30, 634-643.
28
Tables
Table 1
Demographic Characteristics of Add Health Wave 1 Sample (n = 6291)
Freque
ncy
Perc
ent
3087
3204
49.1
50.9
Cumulative
Percent
Gender
Male
Female
49.1
100
Ethnicity
Non-Hispanic White
Non-Hispanic Black
Hispanic
American Indian
Asian
Other
Age
11
12
13
14
15
16
17
18
19
4
159
779
952
1072
1166
1115
904
140
0.1
2.5
12.4
15.1
17
18.5
17.7
14.4
2.2
0.1
2.6
15
30.1
47.1
65.7
83.4
97.8
100
zBMI
-7.00 - -6.01
-6.00 - -5.01
-5.00 - -4.01
-4.00 - -3.01
-3.00 - -2.01
-2.00 - -1.01
-1.00 - 0
0.01 - 1.00
1.01-2.00
2.01-3.00
3.01-4.00
1
1
7
13
76
470
1689
2375
1313
291
4
0
0
0.1
0.2
1.2
7.5
27.1
38.1
21
4.7
0.1
0
0
0.1
0.4
1.6
9.1
36.2
74.2
95.3
99.9
100
4189
66.6
66.6
1543
24.5
91.1
708
11.3
102.4
227
3.6
106
264
4.2
110.2
407
6.5
116.7
(multiple selections possible)
Mean = 0.3327
Standard Deviation = 1.02198
Education Level
of Parents
29
Mother
Father
Household
Income
8th grade or less
>8th grade/didn't graduate
high school
High school graduate
Completed a GED
Business/trade/voc. school
after high school
College/didn't graduate
Graduated from
college/university
Prof training beyond 4-year
college/univ
Went to school/but I don't
know level
She never went to school
I don't know if she went to
school
8th grade or less
>8th grade/didn't graduate
high school
High school graduate
Completed a GED
Business/trade/voc. school
after high school
College/didn't graduate
Graduated from
college/university
Prof training beyond 4-year
college/univ
Went to school/but I don't
know level
He never went to school
I don't know if he went to
school
0 - 24,999
25,000 - 49,999
50,000 - 74,999
75,000 - 99,999
100,000 or greater
95
1.5
2.3
316
1193
179
4.9
18.3
2.8
9.8
38.3
42.6
228
393
3.5
6
48.1
57.5
1073
16.5
83.1
397
6.1
92.6
265
9
4.1
0.1
98.9
99.2
35
92
0.5
1.4
100
2.7
265
895
98
4.1
13.8
1.5
10.5
36.7
39.6
202
273
3.1
4.2
45.5
53.5
812
12.5
77.3
427
6.6
89.8
297
6
4.6
0.1
98.5
98.7
45
0.7
100
1437
1635
1122
409
927
22.1
25.1
17.3
6.3
14.3
26
55.6
75.8
83.2
100
30
Table 2
Demographic Characteristics of Add Health Wave 1 Sample by Race and Gender, Male
NH-White NH-Black
Age
11
1
1
12
37
18
13
237
100
14
322
88
15
357
128
16
372
161
17
400
123
18
292
110
19
49
21
TOTAL
2067
750
z-BMI
-7.00 - -6.01
1
0
-6.00 - -5.01
0
1
-5.00 - -4.01
2
1
-4.00 - -3.01
4
2
-3.00 - -2.01
31
6
-2.00 - -1.01
164
43
-1.00 - 0
533
188
0.01 - 1.00
767
296
1.01-2.00
442
160
2.01-3.00
103
49
3.01-4.00
3
0
TOTAL
2050
746
Mother Education Level 8th grade or less
21
2
>8th grade/didn't graduate high school
82
11
High school graduate
410
106
Completed a GED
56
12
Business/trade/voc. school after high school
81
24
College/didn't graduate
123
36
Graduated from college/university
298
194
Prof training beyond 4-year college/univ
125
73
Went to school/but I don't know level
87
17
She never went to school
3
1
I don't know if she went to school
17
2
TOTAL
1303
478
Father Education Level 8th grade or less
26
3
>8th grade/didn't graduate high school
98
16
High school graduate
319
57
Completed a GED
33
8
Business/trade/voc. school after high school
71
22
College/didn't graduate
101
28
Graduated from college/university
282
100
Prof training beyond 4-year college/univ
147
46
31
His
Income
Went to school/but I don't know level
He never went to school
I don't know if he went to school
TOTAL
0 - 24,999
25,000 - 49,999
50,000 - 74,999
75,000 - 99,999
100,000 or greater
TOTAL
81
2
9
1169
390
584
439
152
290
1855
32
18
0
8
306
238
166
90
39
101
634
Table 3
Demographic Characteristics of Add Health Wave 1 Sample by Race and Gender, Female
NH-White NH-Black Hispanic Americ
Age
11
57
0
0
12
284
32
11
13
343
91
43
14
367
127
62
15
371
146
59
16
350
141
78
17
329
134
45
18
20
100
57
19
19
22
7
TOTAL
2140
793
362
z-BMI
-7.00 - -6.01
0
0
0
-6.00 - -5.01
0
0
0
-5.00 - -4.01
2
0
1
-4.00 - -3.01
2
1
0
-3.00 - -2.01
17
7
4
-2.00 - -1.01
163
52
24
-1.00 - 0
675
146
85
0.01 - 1.00
813
323
146
1.01-2.00
374
204
88
2.01-3.00
61
52
9
3.01-4.00
0
0
0
TOTAL
2107
785
357
Mother ed 8th grade or less
30
6
26
>8th grade/didn't graduate high school
127
54
44
High school graduate
478
139
54
Completed a GED
77
25
10
Business/trade/voc. school after high school
83
31
2
College/didn't graduate
158
60
17
Graduated from college/university
313
202
24
Prof training beyond 4-year college/univ
115
69
2
Went to school/but I don't know level
82
23
28
She never went to school
0
0
1
I don't know if she went to school
8
4
0
TOTAL
1471
613
208
Father Ed 8th grade or less
32
4
17
>8th grade/didn't graduate high school
91
27
30
High school graduate
383
92
34
Completed a GED
35
16
5
Business/trade/voc. school after high school
78
16
9
College/didn't graduate
102
28
9
Graduated from college/university
262
120
13
Prof training beyond 4-year college/univ
155
50
4
33
Income
Went to school/but I don't know level
He never went to school
I don't know if he went to school
TOTAL
0 - 24,999
25,000 - 49,999
50,000 - 74,999
75,000 - 99,999
100,000 or greater
TOTAL
108
0
14
1260
404
573
436
167
330
1910
34
0
3
390
247
181
108
35
112
683
25
3
7
156
130
89
35
11
41
306
34
Table 4
Mean Male z-BMI Over Time by Race
NH-White NH-Black Hispanic
Wave 1
0.3425
0.4374
0.4506
Wave 2
0.3656
0.4313
0.4477
Wave 3
0.7589
0.8028
0.8569
Wave 4
1.0566
1.0444
1.3493
American Indian
0.5848
0.6515
1.0303
1.2548
Asian
0.1977
0.2884
0.711
0.8785
Other
0.4163
0.4441
0.9002
1.38
35
Table 5
Mean Female z-BMI Over Time by Race
NH-White NH-Black Hispanic
Wave 1
0.2269
0.5301
0.3846
Wave 2
0.2301
0.5392
0.3791
Wave 3
0.5016
0.7762
0.6074
Wave 4
0.9236
1.2887
1.0525
American Indian
0.5634
0.507
0.8939
1.3684
Asian
-0.0941
-0.0815
0.2023
0.6323
Other
0.3439
0.4351
0.6379
1.0818
36
Table 6
Final Estimation of Fixed Effects (With Robust Standard Errors)
Fixed Effect
For INTRCPT1, π0
INTRCPT2, β00
SEX, β01
SOCIALSU, β02
SEDENTAR,
β03
LATINO, β04
ISBLACK, β05
ISINDIAN, β06
ISASIAN, β07
ISOTHER, β08
For WAVE slope, π1
INTRCPT2, β10
DEPRESSI, β11
ANTISOCI, β12
RIGHTWEI,
β13
OVERWEIG,
β14
VERYOVER,
β15
LOSEWEIG,
β16
EATVEGGI, β17
POSITIVE, β18
Coefficien
t
Standard
error
0.343324
0.038238
-0.248689
-0.004965
t-ratio
Approx.
d.f.
pvalue
5784
<0.001
0.021698
0.002088
8.979
11.461
-2.378
5784
5784
<0.001
0.017
0.001078
0.093141
0.210927
0.241019
-0.165064
0.092885
0.000345
0.041912
0.025501
0.059732
0.050522
0.051946
3.124
2.222
8.271
4.035
-3.267
1.788
5784
5784
5784
5784
5784
5784
0.002
0.026
<0.001
<0.001
0.001
0.074
0.0642
-0.000892
-0.001155
0.009636
0.000447
0.000415
6.663
-1.995
-2.782
5784
5784
5784
<0.001
0.046
0.005
0.090923
0.009923
9.163
5784
<0.001
0.215255
0.011547
18.642
5784
<0.001
0.301322
0.016265
18.526
5784
<0.001
0.043212
-0.011947
0.002654
0.004635
0.003865
0.001307
9.323
-3.091
2.031
5784
5784
5784
<0.001
0.002
0.042
37
Table 7
Overall z-BMI by Gender
Overall z-BMI
Male
Female
Total
N
0.5877
0.5075
0.5469
3087
3204
6291
Std. Deviation
0.98932
0.91082
0.95092
38
Table 8
Overall z-BMI by Social Support Quartiles
Social Support Quartile
Overall z-BMI
lower 25%ile
0.5965
25-50%ile
0.5522
50-75%ile
0.5173
Upper 25%ile
0.5214
Total
0.5469
N
1572
1574
1573
1572
6291
Std. Deviation
0.96674
0.94095
0.94751
0.9471
0.95092
39
Table 9
Overall z-BMI by Sedentary Quartile
Sedentary Quartile
Overall z-BMI
lower 25%ile
0.4627
25-50%ile
0.5114
50-75%ile
0.5538
Upper 25%ile
0.6635
Total
0.5469
N
1602
1552
1604
1533
6291
Std. Deviation
0.93853
0.93482
0.93835
0.98169
0.95092
40
Table 10
Overall z-BMI by Latino
Race
Overall z-BMI
Non-Latino
0.5357
Latino
0.6348
Total
0.5469
N
5583
708
6291
Std. Deviation
0.95388
0.92323
0.95092
41
Table 11
Overall z-BMI by Black
Race
Overall z-BMI
Non-Black
0.5043
Black
0.6778
Total
0.5469
N
4748
1543
6291
Std. Deviation
0.94361
0.96165
0.95092
42
Table 12
Overall z-BMI by American Indian
Race
Non American Indian
American Indian
Total
Mean
0.5379
0.7853
0.5469
N
6064
227
6291
Std. Deviation
0.94714
1.02059
0.95092
43
Table 13
Overall z-BMI by Asian
Overall z-BMI
Non Asian
0.5594
Asian
0.261
Total
0.5469
N
6027
264
6291
Std. Deviation
0.94758
0.98343
0.95092
44
Table 14
z-BMI Over Time by Depression
Depression Quartile
lower 25%ile Mean
N
Std. Deviation
25-50%ile
Mean
N
Std. Deviation
50-75%ile
Mean
N
Std. Deviation
Upper 25%ile Mean
N
Std. Deviation
Total
Mean
N
Std. Deviation
Quartile
Wave 1
0.2935
1565
1.01147
0.3143
1581
1.01849
0.3484
1572
1.01718
0.3748
1573
1.03957
0.3327
6291
1.02198
Wave 2
0.2955
1154
0.99808
0.2998
1160
1.01053
0.3546
1132
1.04184
0.3789
1119
1.04752
0.3317
4565
1.02471
Wave 3
0.6284
914
1.00616
0.6034
928
1.0143
0.6888
879
1.03112
0.6958
868
1.04377
0.653
3589
1.02391
Wave 4
0.972
1203
1.03642
0.9919
1246
1.03591
1.0748
1225
1.04822
1.0613
1217
1.05724
1.025
4891
1.04506
45
Table 15
z-BMI Over Time by Antisocial
Antisocial Percentile
lower 25%ile Mean
N
Std. Deviation
25-50%ile
Mean
N
Std. Deviation
50-75%ile
Mean
N
Std. Deviation
Upper 25%ile Mean
N
Std. Deviation
Total
Mean
N
Std. Deviation
Wave 1
0.3011
1798
1.03459
0.2956
1387
1.03474
0.3507
1535
1.01541
0.3843
1571
1.00064
0.3327
6291
1.02198
Wave 2
0.3235
1268
1.03893
0.3125
989
1.02554
0.3253
1118
0.988
0.3624
1190
1.04307
0.3317
4565
1.02471
Wave 3
0.6765
1008
1.02898
0.6319
799
1.01906
0.6343
886
1.01843
0.6639
896
1.02888
0.653
3589
1.02391
Wave 4
1.0638
1391
1.03629
0.9736
1086
1.05739
1.0633
1190
1.01395
0.9893
1224
1.07149
1.025
4891
1.04506
46
Table 16
z-BMI Over Time by Perceives Self as "Right Weight" at Baseline
Wave 1 Wave 2 Wave 3 Wave 4
Not Right Weight Mean
0.5362
0.5406
0.8331
1.1937
N
2972
2140
1706
2359
Std. Deviation
1.20055 1.18403 1.13028
1.10802
Right Weight
Mean
0.1506
0.1474
0.4898
0.8679
N
3319
2425
1883
2532
Std. Deviation
0.78683 0.81695 0.88617
0.9566
Total
Mean
0.3327
0.3317
0.653
1.025
N
6291
4565
3589
4891
Std. Deviation
1.02198 1.02471 1.02391
1.04506
47
Table 17
z-BMI Over Time by Perceives Self as "Overweight" at Baseline
Wave 1
Wave 2
Wave 3
Not Overweight
Mean
0.0522
0.0613
0.4264
N
4571
3324
2603
Std. Deviation
0.97159 0.97116
1.0017
Overweight
Mean
1.0784
1.0559
1.2513
N
1720
1241
986
Std. Deviation
0.7392 0.78507 0.82106
Total
Mean
0.3327
0.3317
0.653
N
6291
4565
3589
Std. Deviation
1.02198 1.02471 1.02391
Wave 4
0.7929
3510
1.04709
1.6151
1381
0.77239
1.025
4891
1.04506
48
Table 18
Z-BMI Over Time by Perceives Self as "Very Overweight" at Baseline
Wave 1 Wave 2 Wave 3
Not Very Overweight
Mean
0.2839
0.2843
0.611
N
6082
4416
3470
Std. Deviation 0.99635 0.99811 1.00659
Very Overweight
Mean
1.7545
1.737
1.8787
N
209
149
119
Std. Deviation 0.67877 0.77877 0.72377
Total
Mean
0.3327
0.3317
0.653
N
6291
4565
3589
Std. Deviation 1.02198 1.02471 1.02391
Wave 4
0.9897
4720
1.03954
1.9993
171
0.65489
1.025
4891
1.04506
49
Table 19
z-BMI Over Time by Trying to Gain, Lose or Maintain Weight
Wave 1 Wave 2 Wave 3
Trying to Gain
Mean
-0.1205 -0.0883
0.2648
N
1996
1432
1121
Std. Deviation 0.99594
0.9791 0.99663
Trying to Maintain Mean
0.1567
0.1478
0.5195
N
2182
1612
1263
Std. Deviation
0.9167 0.94288 0.97636
Trying to Lose
Mean
0.9427
0.9219
1.1541
N
2113
1521
1205
Std. Deviation 0.84111
0.8703 0.88921
Total
Mean
0.3327
0.3317
0.653
N
6291
4565
3589
Std. Deviation 1.02198 1.02471 1.02391
Wave 4
0.6227
1515
1.07974
0.9035
1692
1.00897
1.5091
1684
0.84058
1.025
4891
1.04506
50
Table 20
z-BMI Over Time by Daily Servings of Vegetables
Wave 1
Wave 2
0 Mean
0.3798
0.3709
N
2028
1448
Std. Deviation
1.02992
1.03105
1 Mean
0.3107
0.3199
N
2462
1788
Std. Deviation
1.03
1.02199
2 or more
Mean
0.3099
0.3049
N
1801
1329
Std. Deviation
1.00052
1.02093
Total
Mean
0.3327
0.3317
N
6291
4565
Std. Deviation
1.02198
1.02471
Wave 3
0.7019
1112
1.01517
0.6471
1420
1.00253
0.6095
1057
1.05965
0.653
3589
1.02391
Wave 4
1.0982
1574
1.05152
1.0292
1890
1.02274
0.9387
1427
1.06137
1.025
4891
1.04506
51
Table 21
z-BMI Over Time by Positive Outlook Quartile
Positive Outlook Quartile
Wave 1 Wave 2
lower 25%ile Mean
0.3435
0.3174
N
1685
1220
Std. Deviation
1.02583 1.04628
25-50%ile
Mean
0.3532
0.3439
N
1543
1106
Std. Deviation
1.03271 1.03195
50-75%ile
Mean
0.3607
0.3712
N
1603
1180
Std. Deviation
0.99973 1.00136
Upper 25%ile Mean
0.268
0.2913
N
1460
1059
Std. Deviation
1.02856
1.0173
Total
Mean
0.3327
0.3317
N
6291
4565
Std. Deviation
1.02198 1.02471
Wave 3
0.6183
926
1.04562
0.6951
883
1.04216
0.6783
938
1.00294
0.6189
842
1.00272
0.653
3589
1.02391
Wave 4
1.0347
1294
1.07077
1.0361
1183
1.04434
1.0413
1266
1.02157
0.9848
1148
1.04242
1.025
4891
1.04506
52
Figures
Figure 1: Mean Male z-BMI Over Time
Figure 2: Mean Female z-BMI Over Time
Figure 3: Overall z-BMI by Gender
Figure 4: Overall z-BMI by Social Support Quartiles
Figure 5: Overall z-BMI by Sedentary Quartile
Figure 6: Overall z-BMI by Latino
Figure 7: Overall z-BMI by Black
Figure 8: Overall z-BMI by American-Indian
Figure 9: Overall z-BMI by Asian
Figure 10: z-BMI Over Time by Depression Quartile
Figure 11: Overall z-BMI by Antisocial Quartile
Figure 12: z-BMI Over Time by Perceives Self as “Right Weight” at Baseline
Figure 13: z-BMI Over Time by Perceives Self as “Overweight” at Baseline
Figure 14: z-BMI Over Time by Perceives Self as “Very Overweight” at Baseline
Figure 15: z-BMI Over Time by Attempt to Gain, Lose or Maintain Weight at
Baseline
Figure 16: z-BMI Over Time by Daily Servings of Vegetables
Figure 17: z-BMI Over Time by Positive Outlook Quartile
53
Figure 1
Mean Male z-BMI Over Time
1.6
1.4
1.2
Mean zBMI
1
0.8
0.6
0.4
0.2
0
Wave 1
Wave 2
Wave 3
Wave 4
Male NH-White
Male NH-Black
Male Hispanic
Male American Indian
Male Asian
Male Other
54
Figure 2
Mean Female z-BMI Over Time
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
Wave 1
Wave 2
Wave 3
Wave 4
-0.2
NH-White
NH-Black
Hispanic
American Indian
Asian
Other
55
Figure 3
Overall z-BMI by Gender
0.6
0.5877
0.58
0.56
0.54
0.52
0.5075
0.5
0.48
0.46
Male
Female
56
Figure 4
Overall z-BMI by Social Support
Quartiles
0.62
0.6
0.5965
0.58
0.56
0.5522
0.54
0.5173
0.52
0.5214
0.5
0.48
0.46
lower 25%ile
25-50%ile
50-75%ile
Upper 25%ile
57
Figure 5
Overall z-BMI by Sedentary Quartile
0.7
0.6635
0.6
0.5538
0.5114
0.5
0.4627
0.4
0.3
0.2
0.1
0
lower 25%ile
25-50%ile
50-75%ile
Upper 25%ile
58
Figure 6
Overall z-BMI by Latino
0.66
0.6348
0.64
0.62
0.6
0.58
0.56
0.54
0.5357
0.52
0.5
0.48
Non Latino
Latino
59
Figure 7
Overall z-BMI by Black
0.8
0.6778
0.7
0.6
0.5
0.5043
0.4
0.3
0.2
0.1
0
Non Black
Black
60
Figure 8
Overall z-BMI by American-Indian
0.9
0.7853
0.8
0.7
0.6
0.5379
0.5
0.4
0.3
0.2
0.1
0
Non American Indian
American Indian
61
Figure 9
Overall z-BMI by Asian
0.6
0.5594
0.5
0.4
0.3
0.261
0.2
0.1
0
Non Asian
Asian
62
Figure 10
1.2
z-BMI Over Time by Depression Quartile
1
0.8
0.6
0.4
0.2
0
1
2
3
4
Wave
lower 25%ile
25-50%ile
50-75%ile
Upper 25%ile
63
Figure 11
Overall z-BMI by Antisocial Quartile
1.2
1
0.8
0.6
0.4
0.2
0
1
2
3
4
Wave
lower 25%ile
25-50%ile
50-75%ile
Upper 25%ile
64
Figure 12
z-BMI Over Time by Perceives Self as
"Right Weight" at Baseline
1.4
1.2
1
0.8
0.6
0.4
0.2
0
1
2
3
4
Wave
Not Right Weight
Right Weight
65
Figure 13
z-BMI Over Time by Perceives Self as
"Overweight" at Baseline
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
1
2
3
4
Wave
Not Overweight
Overweight
66
Figure 14
z-BMI Over Time by Perceives Self as
"Very Overweight" at Baseline
2.5
2
1.5
1
0.5
0
1
2
3
4
Wave
Not Very Overweight
Very Overweight
67
Figure 15
z-BMI Over Time by Attempt to Gain,
Lose or Maintain Weight at Baseline
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
1
2
3
4
-0.2
-0.4
Wave
Trying to Gain
Trying to Maintain
Trying to Lose
68
Figure 16
z-BMI Over Time by Daily Servings of
Vegetables
1.2
1
0.8
0.6
0.4
0.2
0
1
2
3
4
Wave
0
1
2 or more
69
Figure 17
z-BMI Over Time by Positive Outlook
Quartile
1.2
1
0.8
0.6
0.4
0.2
0
1
2
3
4
Wave
lower 25%ile
25-50%ile
50-75%ile
Upper 25%ile
70
Download