Chat with us, powered by LiveChat Annotated Bibliography | Abc Paper
+1(978)310-4246 credencewriters@gmail.com
  

Complete an annotated bibliography for the following 5 articles attached. 
DO NOT USE OUTSIDE SOURCES OTHER THAN WHAT I HAVE PROVIDED. 
My topic: Causes of Eating Disorders Among Teens

Follow this outline exactly—
 1) Provide ALL three elements:
a. Citation (APA format)
b. Summary of article – usually 1 or 2 paragraphs
c. Your analysis – usually 1 or 2 sentences NOT a full critique

Annotated Bibliography Worksheet

EXAMPLE

Specify your selected Topic here: How to provide Autism Therapy Dogs for comfort during meltdowns and elopement issues to people and families affected by autism in Michigan.

Annotation #1

APA Citation:

Harrison, K. L., & Zane, T. (2017). Is there science behind that? Autism service dogs. 
Science in Autism Treatment

, 14(3), 31-36

Summary: This article will be used in my Proposal as it is an excellent resource of information in regards to Autism Service Dogs. This article not only provides a ton of information on its own, but it also provides many references to scientific studies that have been completed on the subject of autism and the effects of Autism Service Dogs. It explains some of the hypotheses regarding this subject as well as some of the therapy’s history. It also explains the costs associated with this type of therapy and what the projected outlook is in the future. A few of the recommended references are also contacts to other organizations that are set up in other states that may be contacted or referenced later on.

Annotation #2

APA Citation:

Cirulli, F., Borgi, M., Berry, A., Berry, A., Francia, N., & Alleva, E. (2011). Animal-assisted interventions as innovative tools for mental health. Ann lst Super Sanita, 47(4), 341-348.

Summary: The purpose of this article is how animal-assisted methods benefit humans. In order to conduct this study the authors collected statistics in regards to different mental disabilities and different types of animal-assisted therapy in order to reach their results. The authors concluded that animal-assisted therapy is useful for many people. They found that communication increased and negative behaviors decreased due to the use of animal-assisted therapy. The largest limitations regarding this study are that it was conducted only within the Italian population and that it was based solely upon statistics. A recommendation for further research includes following a specific population over time to gather evidence.

Annotation #3

APA Citation:

Hall, P. L., & Malpus, Z. (2000). Pets as therapy: Effects on social interaction in long-stay psychiatry. British Journal of Nursing, 9(21), 2220-2225.

Summary: The purpose of this article was to find out if the verbal and nonverbal social skills of patients increased due to therapy canines. The author completed the study within an inpatient psychiatric nursing home. There were 10 residents who consented to the therapy dogs’ visits and the study. The author used observation and a quasi-experimental design. The author found that there was a notable increase in verbal and non-verbal interaction and communication from the patients when the dog handler was not around. The limitation of this study was the very small sample size of 10 people. In the future is using a larger sample size and to complete some visits with only the dog handler being present without the dog to test the validity of the human factor.

Annotation #4

APA Citation:

Rederfer, L. A., & Goodman, J. F. (1989). Brief report: Pet-facilitated therapy with autistic children. Journal of Autism and Developmental Disorders, 19(3), 461-467.

Summary: The purpose of this article was to find out is animal-assisted therapy helped to decrease the symptoms of children on the autism spectrum. The authors conducted the study through a systematic search and analysis of research studies that were already completed. They used 20 studies with a total of 330 participants between the ages of 3 and 16. The authors found that 8 out of the 20 studies found that animal-assisted therapy helped to decrease the symptoms associated with autism. The limitations for this study is that 90% of the studies that were used were considered to be insufficient. In the future the author should attempt to use studies that provide enough sufficient detail in order to replicate the studies or complete their own study in this fashion.

Annotation #5

APA Citation:

Sams, M. J., Fortney, E. V., & Willenbring, S. (2006). Occupational therapy incorporating animals for children with autism: A pilot investigation. American Journal of Occupational Therapy, 60(3), 268-274.

Summary: The purpose of this article to see how often social interactions and language use in occupational therapy with animals compared with occupational therapy sessions using typical techniques within the field. The author hypothesized that the children would show an increase in social interaction. In order to conduct the study they gathered 22 participants through a school-based OT program. The authors found that the children’s social interactions were increased, but more research in this area is needed. The limitation in regards to this study is figuring out whether the animal-assisted therapy is what increased the social interaction of the children or if it was the motivation of reward. It is recommended that this becomes the basis for further research.

Annotated Bibliography

This assignment gives you the opportunity to begin to construct your literature review by first creating an annotated bibliography of the literature you are reviewing for your final project. For the product that you turn in for review and grading, you will want to:
1) Have at least 4-6 articles/publications
2) Choose those that you plan to include in your literature review – if it is not relevant do not put it in your annotated bib.
3) Provide ALL three elements Prof. David Taylor outlines:
a. Citation (APA format)
b. Summary of article – usually 1 or 2 paragraphs
c. Your analysis – usually 1 or 2 sentences NOT a full critique
4) Remember that the more effort you put in here, the easier your writing will be.

Rubric:

Area

Possible Points

At least 4-6 articles

1

APA style citations

1

Thorough summary of studies including purpose, hypothesis, methods, major conclusions

2

Provides a critique (a sentence or two) of each study

1

TOTAL

5

Five-Year Longitudinal Predictive Factors for Disordered
Eating in a Population-Based Sample of Overweight

Adolescents: Implications for Prevention and Treatment

Dianne Neumark-Sztainer,
PhD, MPH, RD1*

Melanie Wall, PhD2

Mary Story, PhD, RD1

Nancy E. Sherwood, PhD1,3

ABSTRACT

Objective: The objective of this study is

to identify predictors of prevalence

and incidence of disordered eating

(binge eating and extreme weight

control behaviors) among overweight

adolescents.

Method: Five-year longitudinal associa-

tions were examined in 412 overweight

adolescents who participated in Project

EAT-I and II.

Results: Among both overweight males

and females, risk factors for disordered

eating included exposure to weight loss

magazine articles, higher weight impor-

tance, and unhealthy weight control

behaviors, while family connectedness,

body satisfaction, and regular meals

were protective factors, although there

were some differences in predictors of

prevalence (total cases) versus incidence

(new cases) of disordered eating. Among

males, poor eating patterns, including

fast food and sweetened beverage intake,

increased risk for disordered eating, and

the use of healthy weight control behav-

iors was protective.

Discussion: Attention should be

directed toward decreasing disordered

eating among overweight adolescents.

Findings suggest the importance of pro-

moting positive family relationships, psy-

chological health, and regular meals, and

steering adolescents away from overem-

phasizing weight and using unhealthy

weight control behaviors. VVC 2009 by

Wiley Periodicals, Inc.

Keywords: overweight; etiology; binge

eating; disordered eating; risk factors

(Int J Eat Disord 2009; 42:664–672)

Introduction

Obesity among adolescents is of public health con-
cern given its high prevalence and potential
adverse medical consequences.1,2 The use of disor-
dered eating behaviors, such as binge eating and
unhealthy weight control behaviors, is also a seri-
ous problem given that these behaviors are com-
monly used by adolescents and are associated with
poor eating patterns and dietary quality,3,4 eating

disorders,5 and depression,6,7 and have been found
to longitudinally increase risk for weight gain and
obesity.8–11 In previous analyses of the Project EAT
study population, which is utilized in the current
study, the co-occurrence of overweight status and
use of disordered eating behaviors was found to be
high, particularly among female adolescents.12,13

Thus, while an important goal is to prevent obesity
in adolescents, it may be even more important to
prevent the co-occurrence of obesity and disor-
dered eating in youth.

In the design of interventions aimed at the pre-
vention and treatment of disordered eating, it is
important to identify and address factors that are
associated with increased risk for these behaviors.
Furthermore, in the design of interventions for
overweight adolescents, it is important to know
whether there are different subgroups which may
be at particularly high risk for disordered eating
and need tailored interventions. Studies to date
examining disordered eating in overweight adoles-
cents have tended to focus on clinical samples.14–23

However, clinical samples of overweight youth may
differ from non–treatment-seeking overweight
youth in terms of their level of preoccupation with
their weight, prevalence of disordered eating,

Supported by R40 MC 00319 from the Maternal and Child Health

Bureau (Title V, Social Security Act; PI: D. Neumark-Sztainer),

Health Resources and Services Administration, Department of

Health and Human Services.

*Correspondence to: Prof. Dianne Neumark-Sztainer, Division of

Epidemiology and Community Health, School of Public Health,

University of Minnesota, 1300 South Second Street, Suite 300,

Minneapolis, Minnesota 55454. E-mail: [email protected]

Accepted 20 June 2009

1 Division of Epidemiology and Community Health, School of

Public Health, University of Minnesota, Minneapolis, Minnesota
2 Division of Biostatistics, School of Public Health, University of

Minnesota, Minneapolis, Minnesota
3 HealthPartners Research Foundation, Minneapolis, Minnesota

Published online 29 July 2009 in Wiley InterScience

(www.interscience.wiley.com). DOI: 10.1002/eat.20733

VVC 2009 Wiley Periodicals, Inc.

664 International Journal of Eating Disorders 42:7 664–672 2009

CE ACTIVITY

psychological well-being, and sociodemographic
characteristics; therefore, generalizations from
studies on clinical samples of adolescents need to
be made cautiously. Furthermore, studies have
mostly been cross-sectional in nature and thus do
not allow for a determination of temporality in
determining whether factors identified as risk
factors preceded or followed the onset of disor-
dered eating behaviors.14–26 Additionally, in a
review of the literature on disordered eating atti-
tudes and behaviors in overweight youth, Gold-
schmidt et al. indicate there has been limited
research in ethnically diverse populations, and few
studies have had a large enough sample of boys to
examine factors associated with disordered eating
in overweight boys.27 They concluded that
although risk factors for eating disorder pathology
have been identified in the general population, fac-
tors predicting the development of eating disorder
pathology specifically in overweight youth remain
largely unknown.27

We embarked upon the current study to iden-
tify risk and protective factors for disordered
eating behaviors among overweight adolescents.
This study builds upon the existing literature, by
examining risk and protective factors for disor-
dered eating among overweight adolescents using
a longitudinal study design and a nonclinical
study population of adolescent males and females
from diverse ethnic/racial and socioeconomic
backgrounds. The specific aim of the current
study is to identify factors that predict the preva-
lence and incidence of disordered eating among
overweight adolescents. Disordered eating behav-
iors that are examined include binge eating with
loss of control and extreme weight control behav-
iors (i.e., self-induced vomiting and use of diet
pills, laxatives, or diuretics). We hypothesized that
predictors for disordered eating among over-
weight adolescents would be similar to those
found in the general adolescent population and
would include weight-specific variables from
within the domains of socioenvironmental factors
(e.g., weight-related pressures from families and
friends), personal factors (e.g., high importance of
weight), and behavioral factors (e.g., dieting and
unhealthy weight control behaviors).12 We also
hypothesized that other more global variables
from within these domains would be predictive of
disordered eating, including lower family connect-
edness, poorer psychological well-being, and
poorer eating patterns.12 Five-year longitudinal
associations are examined between an array of
socioenvironmental, personal, and behavioral fac-
tors and disordered eating among overweight

male and female adolescents who participated in
a large, population-based study.

Method

Study Population and Design

Project EAT is a study of socioenvironmental, perso-

nal, and behavioral factors of potential relevance to die-

tary intake and weight-related outcomes in adoles-

cents.13,28 There were two measurement time points; the

first (Project EAT-I: Time 1) occurring when participants

were in middle school and high school and the second

(Project EAT-II: Time 2) occurring 5 years later. Project

EAT-I surveyed middle school and high school adoles-

cents from 31 public schools in the Minneapolis/St Paul

area during the 1998–1999 academic year. Participants

completed in-class surveys and anthropometric meas-

ures. Five years later (2003–2004), Project EAT-II resur-

veyed participants by mail. Details on study design and

response rates have been previously described.
8,29

All

applicable institutional and governmental regulations

concerning the ethical use of human volunteers were

followed during this research, and the University of Min-

nesota’s Institutional Review Board Human Subjects

Committee approved all study protocols.

The total study population included 2,516 ethnically

and socioeconomically diverse adolescents (1,386 females

and 1,130 males) who completed surveys in both Project

EAT-I and Project EAT-II. The current analysis includes

only adolescents who were overweight (BMI �85th per-
centile for age and gender)30,31 at both assessments: N 5

180 male adolescents and 232 female adolescents. After

weighting the data to account for attrition over the 5-year

study period (see Statistical Analyses section below), the

ethnic/racial breakdown of this subsample of overweight

participants was as follows: 45% Caucasian, 24% African-

American, 16% Hispanic, 6% Asian, 5% Native American,

and 4% mixed or other race. Forty-two percent of this

sample was of low or low–middle socioeconomic status

(SES). A third of the participants were in middle school in

Project EAT-I (mean age: 12.7 6 0.8 years at Time 1 and

17.3 6 0.6 years at Time 2) and two-thirds of the partici-

pants were in high school in Project EAT-I (mean age: 15.9

6 0.9 years at Time 1 and 20.0 6 0.9 years at Time 2).

Survey Development and Measures

The development of the Project EAT survey was guided

by 21 focus groups with adolescents,32 Social Cognitive

Theory,33,34 a review of existing instruments, reviews by

adolescents and experts, and several pilot tests of the sur-

vey. The survey was designed to assess issues of relevance

to a broad spectrum of weight-related problems in ado-

lescents. Measures used in the current analysis are

described in Table 1.

OVERWEIGHT AND DISORDERED EATING

International Journal of Eating Disorders 42:7 664–672 2009 665

TABLE 1. Description of measures

Variables Description of Survey Item(s)

Weight status and disordered eating
Overweight status Height and weight were self-reported and body mass index (BMI; kg/m2) was calculated. Adolescents with

BMI values above the 85th percentile for gender and age were classified as overweight.30,31

Binge eating with loss of control ‘‘In the past year, have you ever eaten so much food in a short period of time that you would be
embarrassed if others saw you (binge eating)?’’ and ‘‘During the times when you ate this way did you
feel you couldn’t stop eating or control what or how much you were eating?’’ Respondents who
answered affirmatively to both of these questions were classified as engaging in binge eating with loss
of control.35

Extreme weight control behaviors ‘‘Have you done any of the following things in order to lose weight or keep from gaining weight in the
past year?’’ (a) used laxatives; (b) took diet pills; (c) made myself vomit; and (d) used diuretics.
Respondents were categorized as having used extreme weight control behaviors if they reported any of
these behaviors.36

Socioenvironmental variables
Weight-related norms
Maternal/paternal weight

concerns/behaviors
‘‘My mother (father) diets to lose weight or keep from gaining weight.’’ ‘‘My mother (father) encourages me

to diet to control my weight.’’ Four responses ranging from ‘‘not at all’’ to ‘‘very much.’’ Responses for
mother and father were summed (range 5 4–16) such that higher scores indicated greater parental
concern about weight. Cronbach’s a 5 0.77.

Peer dieting behaviors ‘‘Many of my friends diet to lose weight or keep from gaining weight.’’ Four responses ranging from
‘‘not at all’’ to ‘‘very much.’’

Weight teasing by family members ‘‘Have you ever been teased or made fun of by family members because of your weight?’’ (yes/no)
Respondents who indicated they had been teased and also bothered by family members teasing them
about their weight in a separate question were categorized as having been teased.37

Weight teasing by peers Same questions as above about teasing by other kids.
Media exposure
Magazines on weight loss ‘‘How often do you read magazine articles in which dieting or weight loss are discussed?’’ Four response

categories ranged from ‘‘never’’ to ‘‘often.’’
Television viewing ‘‘In your free time on an average weekday (Monday–Friday), how many hours do you spend watching TV

and videos?’’ A similar question was asked about weekend use.38,39 Responses for weekday and
weekend use were combined (5 3 weekday hours 1 2 3 weekend hours) to compute total viewing
hours per week (range 5 0–35).

Family meals
Family meal frequency ‘‘During the past seven days, how many times did all, or most, of your family living in your house eat a

meal together?’’
Family meal atmosphere ‘‘How strongly do you agree with the following statements? (a) I enjoy eating meals with my family; (b) In

my family, eating brings people together in an enjoyable way; (c) In my family, mealtime is a time for
talking with other family members; and (d) In my family, dinner time is about more than just getting
food, we all talk with each other.’’ Items were rated on a four-point scale and responses were summed
(range 5 4–16) such that higher scores indicated a more positive atmosphere. Cronbach’s a 5 0.73.

Relationships
Family connectedness ‘‘How much do you feel you can talk to your mother (father) about your problems?’’ ‘‘How much do you

feel your mother (father) cares about you?’’ Five response categories. Responses for mother and father
were summed (range 5 4–20) such that higher scores indicated more communication and greater
caring. Cronbach’s a 5 0.69.

Friend connectedness ‘‘Do you have one or more close friends who you can talk to about your problems?’’ Response categories
were ‘‘yes, always,’’ ‘‘yes, sometimes,’’ and ‘‘no.’’

Personal variables
Body image and weight concerns
Body satisfaction Body satisfaction scale assessing satisfaction with different body parts (height, weight, body shape, waist,

hips, thighs, stomach, face, body build, shoulders). Five response categories. Responses to the ten items
were summed (range 5 10–50) such that higher scores indicated greater body satisfaction. Cronbach’s
a 5 0.92.40

Weight concern ‘‘How strongly do you agree with the following statements? (a) I think a lot about being thinner; and (b) I
am worried about gaining weight.’’ Items were rated on a four-point scale and responses were summed
such that a higher score indicated greater concern about weight (range 5 2–8). Cronbach’s a 5 0.87.

Weight importance ‘‘During the past six months, how important has your weight or shape been in how you feel about
yourself?’’ Four responses ranging from ‘‘weight and shape were not very important’’ to ‘‘weight and
shape were the most important things that affected how I felt about myself’’.35

Psychological well-being
Depressive symptoms Kandel and Davies’ (1982) six-item scale assessing depressive mood. Items were rated on a three-point

scale and responses were summed (range 5 6–18) such that a higher score indicated a more depressed
mood. Cronbach’s a 5 0.82.41

Self-esteem Shortened six-item version of Rosenberg’s Self-Esteem Scale. Items were rated on a four-point scale and
responses were summed such that a higher score indicated greater self-esteem (range 5 6–24).
Cronbach’s a 5 0.79.42

(Continued)

NEUMARK-SZTAINER ET AL.

666 International Journal of Eating Disorders 42:7 664–672 2009

Body mass index (BMI) and overweight status are

based on self-reported heights and weights. Although at

Time 2, only self-reported height and weight was

assessed, at Time 1 both measured and self-reported

height and weight were assessed and BMI values were

highly correlated (females: r 5 0.85; males: r 5 0.89).
48

Disordered eating behaviors in the past year included

binge eating with loss of control and extreme weight con-

trol behaviors (self-induced vomiting and use of diet

pills, laxatives, and diuretics). This variable was coded

dichotomously; adolescents using at least one of the dis-

ordered eating behaviors in the past year were consid-

ered to have engaged in disordered eating behaviors.

Table 1 also includes sociodemographic variables and

an array of socioenvironmental, personal, and behavioral

variables of potential relevance to disordered eating.

Statistical Analyses

All analyses were performed on the subset of adoles-

cents who were overweight at both Time 1 and Time 2

and who had complete information on disordered eating

at both times. Analyses were conducted separately for

female and male adolescents because of previously iden-

tified differences in the prevalence and predictors of dis-

ordered eating across gender.12,13 We first examined and

calculated the prevalence, incidence, and persistence of

disordered eating behaviors (i.e. binge eating or engaging

in extreme weight control behaviors). Logistic regression

was used to identify socioenvironmental, personal, and

behavioral predictors of prevalence and incidence of

Time 2 disordered eating adjusted for sociodemographic

characteristics (age cohort, race/ethnicity, SES). Separate

regressions were run for each predictor. Predictors of

Time 2 prevalence (i.e., total cases) of disordered eating

were identified using logistic regression on all of the ado-

lescents identified as overweight at Time 1 and Time 2,

while analyses for predictors of Time 2 incidence (i.e.,

new cases) of disordered eating included only those over-

weight adolescence who were not engaging in disordered

eating at Time 1. All analyses were stratified by gender

and conducted using SAS software (version 9.1, 2003,

SAS, Cary, NC).

To account for differential response rates across socio-

demographic groups in the longitudinal sample, data were

weighted using the response propensity method,49 where

the inverse of the estimated probability that an individual

responded at Time 2 is used as the weight. After adjusting

for sociodemographics and weighting, there were no sig-

nificant differences found for Time 1 overweight status,

TABLE 1. (Continued)

Variables Description of Survey Item(s)

Behavioral variables
Weight control practices
Dieting ‘‘How often have you gone on a diet in the past year? By ‘diet’ we mean changing the way you eat so you

can lose weight.’’ Responses were categorized dichotomously.
Healthy weight control behaviors ‘‘Have you done any of the following things in order to lose weight or keep from gaining weight during

the past year? (a) exercise; (b) ate more fruits and vegetables; (c) ate less high-fat foods; (d) ate less
sweets’’ Respondents were categorized as having used only healthy weight control behaviors if they
used any of these behaviors but did not use any unhealthy weight control behaviors.

Unhealthy weight control
behaviors

‘‘Have you done any of the following things in order to lose weight or keep from gaining weight during
the past year? (a) fasted; (b) ate very little food; (c) used food substitute (e.g., slimfast); (d) skipped
meals.’’ Respondents were categorized as having used unhealthy weight control behaviors if they used
any of these behaviors.

Eating/physical activity patterns
Breakfast frequency ‘‘During the past week, how many days did you eat breakfast?’’
Lunch frequency ‘‘During the past week, how many days did you eat lunch?’’
Dinner frequency ‘‘During the past week, how many days did you eat dinner?’’
Fast food ‘‘In the past week, how often did you eat something from a fast food restaurant (like McDonald’s,

Burger King, Hardee’s, etc.)?’’
Dietary intake Intake of fruits and vegetables, sugar-sweetened beverages, and diet soda were assessed with the

149-item, semiquantitative Youth and Adolescent Food Frequency Questionnaire.43,44

Moderate-to-vigorous
physical activity

Modified version of the Leisure time Exercise Questionnaire.45–47 Responses on two questions assessing
hours spent in strenuous activity and moderate physical activity in a usual week were summed
(range 5 0–16 h).

Demographic variables
Sex Are you. . .1) male 2) female.
Ethnicity/race ‘‘Do you think of yourself as. . .White, Black or African American, Hispanic or Latino, Asian American,

Hawaiian or Pacific Islander, or American Indian or Native American.’’ Subjects could choose more
than one category; those responses indicating multiple categories were coded as ‘‘mixed/other’’.

Age What is your age?
Socioeconomic status Composite variable based primarily on parental level of education, defined by the higher level of either

parent. In cases of missing data on educational level for both parents, other variables used included
eligibility for public assistance, eligibility for free or reduced-cost school meals, and parental
employment status.28

OVERWEIGHT AND DISORDERED EATING

International Journal of Eating Disorders 42:7 664–672 2009 667

binge eating, or extreme weight control behaviors between

Time 2 responders and nonresponders.

Results

Disordered Eating Prevalence, Incidence, and

Tracking among Overweight Adolescents

Among the 232 overweight female adolescents,
30.8% (n 5 71) engaged in disordered eating at
Time 1 and that number increased to 40.1% (n 5
93) at Time 2. Among the 161 overweight females
who were not engaging in disordered eating at
Time 1, about one-third of them (34.1%, n 5 55)
started to engage in disordered eating behaviors by
Time 2. Among the 71 females who were already
engaging in disordered eating at Time 1, roughly
half (53.6%, n 5 38) were still engaging in those
behaviors at Time 2. Among the 93 overweight ado-
lescent females reporting disordered eating behav-
iors at Time 2, almost all (89.2%, n 5 83) reported
the use of at least one extreme weight control
behavior and one-third (32.2%, n 5 30) reported
binge eating with loss of control.

Among the 180 overweight male adolescents,
13.4% (n 5 24) engaged in disordered eating at
Time 1 and that number increased to 20.2% (n 5
36) at Time 2. Among the 24 overweight males
engaging in disordered eating behaviors at Time 1,
37.6% (n 5 9) continued to engage in these behav-
iors at Time 2, while 17.5% (n 5 27) of the over-
weight males previously not engaged in disordered
eating at Time 1 had begun by Time 2. Of the 36
overweight males engaging in disordered eating at
Time 2, the majority (77.8%, n 5 28) reported the
use of at least one extreme weight control behavior
and almost one-third (30.6%, n 5 11) reported
binge eating with loss of control. Similar to the
females, overweight males were more likely to use
extreme weight control behaviors than to engage in
binge eating.

At Time 2, females engaging in disordered eating
behaviors had higher BMI levels (M 5 32.3; SD 5
5.6) than females not engaging in disordered eating
(M 5 30.9; SD 5 4.6) (p .041). Similarly, among
male adolescents, those with disordered eating
behaviors had higher BMI levels (M 5 34.5; SD 5
6.9) than those not engaging in disordered eating
(M 5 31.8; SD 5 4.1) (p 5 0.003).

Predictors of Disordered Eating

In overweight adolescent females, a number of
socioenvironmental, personal, and behavioral Time
1 variables were identified that predicted the preva-

lence and/or incidence of disordered eating at
Time 2 (Table 2). For variables within the socioen-
vironmental domain, exposure to magazine articles
about weight loss was associated with increased
prevalence and incidence of disordered eating,
while a positive atmosphere at family meals and
greater family connectedness was associated with a
lower prevalence of disordered eating. For personal
factors, higher levels of weight concerns were asso-
ciated with increased prevalence and incidence of
disordered eating, higher levels of weight impor-
tance were associated with higher prevalence of
disordered eating, and higher levels of body satis-
faction and self-esteem were associated with a
lower prevalence of disordered eating. Behaviors
associated with increases in prevalence and inci-
dence of disordered eating included dieting,
unhealthy weight control behaviors (skipping
meals, eating very little, smoking, fasting, and using
food substitutes), and increased hours of
moderate-to-vigorous physical activity. Eating
lunch and dinner on a regular basis were protective
against disordered eating in overweight female
adolescents.

Among overweight male adolescents, within the
domain of socioenvironmental factors, peer dieting
predicted a higher prevalence of disordered eating,
exposure to magazine articles about weight loss
was associated with higher prevalence and inci-
dence of disordered eating, and family connected-
ness was a protective factor for both prevalence
and incidence of disordered eating (Table 2). Per-
sonal factors that predicted higher prevalence and
incidence of disordered eating included greater
weight importance and depressive symptoms,
while body satisfaction was protective for preva-
lence of disordered eating. Behavioral risk factors
predicting higher levels of disordered eating
included unhealthy weight control behaviors, eat-
ing fast food, and drinking sweetened beverages.
The use of healthy weight control behaviors and
eating meals on a regular basis predicted lower
prevalence of disordered eating in overweight male
adolescents.

Discussion

The aim of the current study was to identify risk
and protective factors for disordered eating among
a population-based sample of overweight adoles-
cents. Although similar proportions of adolescent
females and males were overweight, disordered
eating was about twice as common among

NEUMARK-SZTAINER ET AL.

668 International Journal of Eating Disorders 42:7 664–672 2009

overweight females than overweight males at
Time 2. A number of socioenvironmental, personal,
and behavioral variables were identified that
longitudinally predicted disordered eating in this
population. While there were some differences
across prevalence and incidence analyses, and
between males and females, a number of fairly
consistent patterns emerged. Within the realm of
socioenvironmental factors, reading magazines
that had articles on dieting and weight loss was a
risk factor for disordered eating, while family con-
nectedness was a protective factor. Personal factors
that predicted disordered eating in both females
and males included weight importance, while body
satisfaction was protective. Higher self-esteem was

protective in female adolescents, and the existence
of depressive symptoms was a risk factor in male
adolescents. Unhealthy weight control behaviors
(including skipping meals, eating very little, fasting,
using food substitutes, and smoking) increased risk
for disordered eating, while eating meals on a regu-
lar basis was protective in both male and female
adolescents. Given the high prevalence of over-
weight adolescents engaging in disordered eating
behaviors and the potentially harmful consequen-
ces associated with these behaviors,3–11 it is crucial
to develop interventions to prevent and reduce dis-
ordered eating in this population. Findings from
the current study can help guide the implementa-
tion of such interventions.

TABLE 2. Five-year longitudinal predictors of prevalence (total cases) and incidence (new cases) of disordered eating
at Time 2 among overweight adolescents (at Time 1 and Time 2)a,b

Females Males

OR CI
p-value

(prevalence)
p-valuec

(incidence) OR CI
p-value

(prevalence)
p-valuec

(incidence)

Socioenvironmental factors
Weight-related norms
Maternal weight concerns/behaviors 1.23 0.93, 1.63 .151 .117 1.18 0.78, 1.77 .430 .904
Paternal weight concerns/behaviors 1.06 0.78, 1.45 .711 .366 1.05 0.69, 1.60 .830 .504
Peer dieting behaviors 1.16 0.94, 1.43 .165 .374 1.51 1.09, 2.10 .014 .575
Weight-teasing by family 1.36 0.76, 2.43 .302 .217 0.64 0.23, 1.76 .383 .088
Weight-teasing by peers 1.58 0.89, 2.82 .122 .878 0.68 0.30, 1.53 .350 .061

Media exposure
Magazines on weight loss 1.55 1.12, 2.15 .009 .004 1.80 1.18, 2.73 .006 .020
Television viewing 0.99 0.96, 1.02 .691 .903 0.99 0.95, 1.03 .484 .218

Family meals
Family meal frequency 0.94 0.84, 1.05 .253 .944 0.93 0.79, 1.10 .413 .742
Family meal atmosphere 0.61 0.44, 0.86 .005 .146 0.75 0.44, 1.28 .294 .455

Relationships
Family connectedness 0.90 0.83, 0.98 .010 .322 0.86 0.77, 0.96 .009 .024
Friend connectedness 0.85 0.55, 1.29 .438 .934 1.17 0.64, 2.14 .605 .195

Personal factors
Body image and weight concerns
Body satisfaction 0.94 0.91, 0.98 .001 .086 0.96 0.91, 1.00 .049 .178
Weight concern 1.90 1.27, 2.83 .002 .021 1.29 0.82, 2.02 .280 .577
Weight importance 1.44 1.08, 1.93 .014 .710 1.88 1.20, 2.95 .006 .042

Psychological well-being
Self-esteem 0.91 0.84, 0.99 .022 .214 0.92 0.81, 1.04 .184 …

Journal of Consulting and Oinical Psychology
1996, Vol. 64, No. 5,936-940

Copyright 1996 by the American Psychological Association, Inc.
0022-006X/96/$3.00

Weight Concerns Influence the Development of Eating Disorders:
A 4-Year Prospective Study

Joel D. Killen, C. Barr Taylor, Chris Hayward, K. Parish Haydel, Darrell M. Wilson, Larry Hammer,
Helena Kraemer, Anne Blair-Greiner, and Diane Strachowski

Stanford University School of Medicine

The authors examined factors prospectively associated with age of onset of partial syndrome eating

disorders over a 4-year interval in a community sample (N = 877) of high school-age adolescent
girls. Four percent developed a partial syndrome eating disorder over the interval. A measure of
weight concerns was significantly associated with onset in a multivariate Cox proportional hazard

analysis (p < .001). Girls scoring in the highest quartile on the measure of weight concerns had the

highest incidence (10%) of partial syndrome onset, whereas none of the girls in the lowest quartile
developed eating disorder symptoms. This finding is consistent with both theoretical and clinical

perspectives and may represent a useful step toward the establishment of a rational basis for the
choice of a prevention intervention target.

Eating disorders are a significant public health concern

(Herzog & Copeland, 1985). Consequently, leading clinicians

and researchers have called for work aimed at the development

of eating disorder prevention interventions (Shisslak, Crago, &

Estes, in press; Striegel-Moore, Silberstein, Frensch, & Rodin,

1989; Wadden & Stunkard, 1985). However, before effective

prevention interventions can be developed, rational interven-

tion targets need to be identified. Currently, the development of

prevention interventions is hampered by limited understanding

of the risk factors that may influence the development of eating

disorders (Agras & Kirkley, 1986; Fairburn & Beglin, 1990;

Strober, 1986).

Social learning models suggest that society's preoccupation

with weight and shape may serve as an important risk factor for

the development of eating disorders (Striegel-Moore, Silb-

erstein, & Rodin, 1986). Societal norms that link beauty, suc-

cess, and happiness to a thin body shape may produce pressures

to maintain a slender physique that can lead to the development

of excessive dieting and other un healthful weight regulation

Joel D. Killen and K. Parish Haydel, Department of Medicine, Stan-
ford University School of Medicine; Darrell M. Wilson and Larry Ham-

mer, Department of Pediatrics, Stanford University School of Medicine;
C. Barr Taylor, Chris Hayward, Helena Kraemer, Anne Blair-Greiner,
and Diane Strachowski, Department of Psychiatry and Behavioral Sci-

ences, Stanford University School of Medicine.
This study was ftmded by Public Health Service Grant RO1 HD24240-

01 from the National Institute for Child Health and Development.
We thank the following people for their assistance in conducting this

study: Darby Cunning and Suzanne Taborski-Barba from Stanford
University; Don Flohr, Jack Scardina, Rod Adams, Cliff Harris, Sherry
Garvey, Barbara SemenofT, Wanda Wong, and Laura Arnold of the
Santa Clara Unified School District; and Jim Warren and Joanna Laird

of the Fremont Union High School District.
Correspondence concerning this article should be addressed to Joel

D. Killen, Stanford University School of Medicine, 1000 Welch Road,
Palo Alto, California 94304. Electronic mail may be sent via Internet to
[email protected]

practices. However, although weight concerns and attendant be-

haviors such as dieting may be pandemic among girls and

women, comparatively few go on to develop severe clinical syn-

dromes (Strober, 1986). Only a few prospective studies have

examined linkages between preoccupations with weight and

body shape, excessive dieting, and the subsequent onset of eat-

ing disorder symptoms (Attie & Brooks-Gunn, 1989; Graber,

Brooks-Gunn. Paikoff, & Warren, 1994; Striegel-Moore et al.,

1989). These efforts have used comparatively small sample

sizes(Attie & Brooks-Gunn, 1989; Graber etal., 1994)or stud-

ied the development of symptoms over relatively short intervals

of time (Attie & Brooks-Gunn, 1989; Striegel-Moore, et al.,

1989) and used paper-and-pencil instruments rather than clini-

cal interviews to document symptomatology.

Previously, we presented evidence of linkage between weight

and shape concerns and eating disorder symptoms in a prospec-

tive study of young adolescent girls enrolled in the sixth and

seventh grades at the time of study entry (Killen et al., 1994).

Over a 3-year interval, 4% developed symptoms reflective of a

partial syndrome eating disorder (Shisslak, et al., 1995). A

measure of weight and shape concerns, developed for use in that

study, predicted development of partial syndrome eating disor-

ders over the 3-year period.

In this article, we extend the earlier finding through an anal-

ysis of data obtained from a new 4-year prospective study of

potential risk factors for the development of eating disorders in

a sample of high-school-age girls, for this study, unlike in the

previous trial, a structured clinical interview was used to assess

eating disorder symptoms.

In addition to analyzing the effects of weight and shape con-

cerns, we examined the relationship between eating disorders,

"emotionality" and the two other early emerging, inherited per-

sonality traits or temperaments ("activity," "sociability") iden-

tified by Buss and Plomin (1984). Our interest in emotionality

as a risk factor stems from work that indicates that those suffer-

ing from eating disorders also manifest higher rates of behaviors

reflective of poor impulse control (Casper, Eckert, Halmi,

936

WEIGHT CONCERNS AND EATING DISORDERS 937

Goldberg, & Davis, 1980; Garfinkel, Moldofsky, & Garner,

1980; Strober, 1984). A temperament model of the develop-

ment of eating disorders suggests that emotionality may provide

a common pathway for both impulse control problems and dis-

ordered eating behaviors. Bulimics, for example, may be tem-

peramentally predisposed to emotionality, which is expressed

as a susceptibility to become easily and intensely distressed. Bu-

limic behavior may stem from this emotionality in impulsive

individuals with problems of control. Although personality

problems consistent with this hypothesis have been shown to

coexist with eating disorders, earlier studies were largely cross-

sectional and, thus, of limited value for describing the process

of development of eating disorders.

The focus of this article, as in our previous work, is on partial

syndrome eating disorders. This focus reflects the growing rec-

ognition that a broad spectrum of eating disturbances exist in

the community and that these "partial syndrome" disorders are

not adequately described by current diagnostic critera (Shisslak

et al., 1995). Clinical researchers have emphasized the impor-

tance of longitudinal research with community samples to bet-

ter characterize the spectrum of eating disturbances. Such re-

search may help in identifying risk factors that may serve as

targets for prevention efforts and enhance the understanding of

psychopathological mechanisms (Fairburn & Beglin, 1990;

Striegel-Moore et al., 1989).

Method

Participants

A total of 877 ninth-grade girls enrolled in four northern California

high schools participated iri the study, which was approved by the Stan-
ford University School of Medicine Committee for the Protection of

Human Subjects in Research. At baseline, the average age of the girls in
the sample was 14.9 years. Ethnic distribution was as follows: White,
46%; Hispanic, 14%; Asian (Cambodian, Vietnamese, Chinese, Japa-
nese, Thai), 23%; Black, 3%; Pacific Islander. 6%; Native American, 2%;

other, 6%.

Procedure

Baseline and follow-up assessments were conducted over a 4-year pe-
riod by staff trained by the principal investigators. Paper-and-pencil
measures and structured clinical interviews were completed during the

students' regular classroom periods. Students were identified by a spe-
cial identification number to ensure confidentiality.

Factors Assessed at Baseline and Follow- Ups

Weight concerns. This instrument is designed to ascertain partici-
pants' fear of weight gain, worry over weight and body shape, impor-

tance of weight, diet history, and perceived fatness. Previous psychomet-
ric work indicated that the measure has excellent stability (test-retest r
= .71 fora7-monthinterval)andgoodsensitivity(BCillenetal., 1994).

Excellent stability was demonstrated again in this sample ( r = .75 for a
12-month interval).

Eating Disorder Inventory (EDI). The EDI is designed to assess a
variety of psychological and behavioral characteristics common in an-
orexia nervosa and bulimia nervosa (Gamer, Olmstead, & Polivy,
1983). The instrument consists of eight subscales.

Dietary restraint. Dietary restraint was measured using the revised
Restraint scale developed by Herman et al. (Herman, Polivy, Pliner, &

Threlkeld, 1978). The instrument yields a total score and two subscale
scores: Weight Fluctuation (WF) and Concern for Dieting (CD).

Temperament. Early emerging, inherited personality traits were as-
sessed with the 20-item instrument assessing emotionality (defined by

three components: distress, fear, and anger), activity, and sociability (E-
A-S; Buss & Plomin, 1984). Sample items include "I frequently get

distressed (E)," "When I get scared I panic (F)," "I usually seem to be
in a hurry (Act)," and "I like to be with people (S)." Test-retest reli-
abilities from previous work range from .75 to .85 for the various com-

ponents, and correlational studies suggest that the components are in-
dependent (Buss & Plomin, 1984). The response choices were modified

slightly from the original to improve comprehension for this target
sample.

Height. Standing height was measured to the nearest millimeter us-

ing a portable direct reading stadiometer. Students were measured with

shoes removed, and the body was positioned so that the heels and but-
tocks were against the vertical support of the stadiometer and the head
was aligned so that the auditory canal and the lower rim of the orbit

were in a horizontal plane. Two measures of height were obtained, and
the average was used in data analyses.

Weight. Body weight was determined to the nearest 0.1 kg, using

digital scales with the participants wearing light indoor clothing without
shoes or coats. Two measures of weight were obtained and the average

used in data analyses.
Body mass index (BMI). This index was computed from the for-

mula "weight (kg) divided by height (m) squared," which is generally

considered to be the preferred index of relative body weight as a reflec-
tion of adiposity (Kraemer, Berkowitz, & Hammer, 1990).

Although restraint, drive for thinness, and body dissatisfaction and

the measure of weight concerns were appreciably intercorrelated (rs
ranged from .47 to .78), all were retained as independent variables in

subsequent analyses. Restraint and the two EDI variables were retained
because extensive research has been conducted with these variables.

The weight concerns variable was included because of its success in pre-
dicting onset of symptoms in out previous work and its excellent psy-
chometric properties (Killen et al., 1994).

Self-reported frequency of drinking. Frequency of drinking in the

last month was assessed using an item developed by the University of
Southern California health behavior research group. The question is

phrased "How many alcoholic drinks have you had in the last month
(30 days) ?" (Graham, Flay, & Johnson, 1984). Response choices range
from 1 (none) to 7 (20 + drinks).

Structured Clinical Interview

A structured clinical interview was used to assess eating disorder

symptoms. The interview featured diagnostic criteria from the Diag-
nostic and Statistical Manual of Menial Disorders (3rd ed., rev.; DSM-

III-R; American Psychiatric Association, 1987) definition of bulimia
nervosa and the Eating Disorder Examination (EDE; Cooper, Cooper,

& Fairburn, 1989). The interview was adapted for adolescents from the
original EDE designed for use with adult clinical populations.

Female interviewers were recruited from seven graduate schools in
either counseling psychology or clinical psychology programs. In-

terviewers undertook an extensive training seminar that consisted of
two 8-hr training sessions, during which the criteria for eating disorders
were reviewed and a videotape of a simulated interview was observed.
Participatory instruction was facilitated through role playing. At the
completion of training, interviewers demonstrated knowledge of DSM
criteria for eating disorders and competent interviewing style.

Structured interviews were completed during regular classroom pe-
riods. Individual students were escorted by an interviewer to a private
location for the structured interview. At the onset of the interview, stu-

dents were informed that the interview was confidential unless the stu-

938 KILLEN ET AL.

dent disclosed information regarding imminent danger to themselves or
others.

Interview forms were reviewed on site for quality assurance. Periodi-
cally, interviewers were supervised conducting an interview. Frequent
feedback was given throughout the study, maintaining form accuracy

and overall diagnostic appraisal. All interviewers served a probationary
period to demonstrate professionalism and interviewing competency.

Group meetings were held daily with supervisors and interviewers to

address logistical and conceptual issues. Supervisors were available at
all times to answer any questions that interviewers may have had.

The structured interview consisted of 18 questions addressing the en-

dorsement of binge eating, compensatory behaviors, and measures of
weight and shape concern. Eight of the questions required the student
to give numerical frequency or rating responses from a printed card.

Use of the card was provided to increase students' accuracy. Objective
bulimic episodes were assessed in accordance with the DSA/definition
of large amounts of food accompanied by experienced "loss of control."

Interviewers rated the amount of food eaten during a binge as 1 (not
large), 2 (more than average), or 3 (large). Reports of Compensatory

bulimic behaviors (excessive exercise, self-induced vomiting, laxative
abuse, and diuretic misuse) were obtained through the interview. We

measured frequency scores for objective binge episodes and compensa-

tory behaviors using the students' responses to a printed card with nu-
merical categories. Response choices ranged from 0 (never) to 7 (every
day for the past 3 months). Presence of dietary rules was determined by

the interviewer on the basis of the EDE criteria for rigid rules. Adher-
ence to dietary rules for purposes other than weight and shape were

ruled out. The importance of weight and shape was determined by the

level of significance the student placed on this index. Students were
asked how important their weight and shape was in relation to other
areas of their life. For clarification, students were asked, "Is it more or

less important than getting good grades or doing things with friends?"
and "Do you only feel good about yourself when you are at a certain

weight?" Interviewer ratings for overall importance of weight and shape
were made on a scale ranging from 0 (no importance) to 6 (most

important).

Definition of Partial Syndrome

Because weight and shape concerns are core features of clinical eating

disorders, there is necessarily some overlap between the independent
variable and the dependent variable in this study. However, to be char-
acterized as suffering from a partial syndrome eating disorder, girls had

to present with a range of symptoms beyond weight and shape concerns.
Specifically, girls had to manifest the following during the 3 months

before the day of the interview: (a) binge eating episodes in which the

amount of food consumed was above average or large; (b) compensa-
tory behavior specifically to prevent weight gain, including vomiting,
using laxatives, diuretics, exercise (at least five times in the preceding

3 months), or following strict dietary rules; and (c) overconcem and
preoccupation with body weight and shape ("weight/shape is one of
the main aspects of self-worth," or "weight/shape is the most important

factor in determining self-worth") or a feeling of lack of control of eat-
ing during a binge episode.

Results

Of the total sample (N = 877), 825 girls were available for

our analyses. Of the 52 girls who were not available, 1 did not

provide enough information for classification and 51 were al-

ready symptomatic at entry. Because the age of onset of symp-

toms for these 51 students was unknown at study entry, they

were excluded from longitudinal analyses.

Univariate Analysis

Over the 4-year interval, 4% (36 of 825) developed a partial

syndrome eating disorder as previously denned. Baseline scores

on the predictor variables of those who developed a partial syn-

drome and those who did not are compared in Table 1. It can

be seen that there is statistically significant differentiation on the

majority of the study variables.

Multivariate Testing

All the factors in Table 1 were considered for inclusion in

a stepwise Cox proportional hazards model. The proportional

hazards model seeks the best combination of variables to pre-

dict hazard of onset. To control for false-positive results, we

used a p < .01 criterion to enter each variable. For this analysis,

which requires complete data on all predictor variables, 734 of

the 825 available girls provided data.

The results of the analysis indicated that girls with high scores

on the measure of weight concerns were more likely to develop

a partial syndrome eating disorder over time. Only the measure

of weight concerns met the standard for statistical significance

in the multivariate analysis, x2( 1, Af = 734) = 19.10,p < .001.

Multivariate Description

Separate survival curves using the Kaplan-Meier method

were then generated for each quartile of the measure of weight

concerns. For this analysis, which requires complete data only

on significant predictor variables, 792 of the 825 available girls

provided data.

Curves for three of the four quartiles are shown in Figure

1. As expected from the aforementioned statistical significance

reported (Table 1), the groups show different survival, log-rank

X 2 (3, N = 792) = 30.85,p < .001. Those scoring in the highest

quartile on the measure of weight concerns (n = 195) had the

highest incidence (10%) of syndrome development by age 17,

whereas no girls in the lowest quartile (« = 193) developed par-

tial syndrome by age 17.

Discussion

The demonstration of a linkage between weight concerns and

subsequent development of a partial syndrome eating disorder

in this study is interesting for several reasons. First, as noted,

although social learning models implicate weight and shape

concerns as a risk factor for the development of eating disorders,

there have been few prospective demonstrations of such a link-

age. Previous studies, including our own work, that have exam-

ined relationships between weight and shape concerns and dis-

ordered eating have relied on paper-and-pencil measures rather

than clinical interviews to assess symptomatology (Attie &

Brooks-Gunn, 1989; Killen et at, 1994; Striegel-Moore et al.,

1989). However, there is general consensus that interviews are

needed to assess several of the key features of eating disorders

required for diagnosis (Fairbum & Beglin, 1990). Second, this

study extends the findings from our previous work to a new

sample of older adolescents. Thus, it appears that weight con-

cerns remain an important factor in the development of disor-

dered eating during the course of adolescent growth and devel-

WEIGHT CONCERNS AND EATING DISORDERS 939

Table 1
Baseline Scores for Girls Who Developed a Partial Syndrome Eating Disorder Over the
4- Year Interval and for Girls Who Remained Asymptomatic

Variable

Asymptomatic
<« = 789)

Partial syndrome
<n = 36)

M SD M SD

Weight concerns 33.03 24.22 58.65 24.80 <.001
Eating Disorder Inventory

Drive for Thinness 3.98 5.07 8.72 5.62 <.001
Bulimia 0.68 1.83 2.25 3.24 <.01
Body Dissatisfaction 9.94 7.24 14.27 7.56 <.01
Perfectionism 4.20 4.05 5.44 3.87 <.10
Maturity Fears 4.18 3.50 4.38 3.81 .73
Ineffectiveness 3.25 3.88 4.88 4.20 <.05
Interoceptive Awareness 2.28 3.20 5.19 4.53 <.001
Interpersonal Distrust 3.43 3.68 2.83 3.09 .33

Temperament
Distress 2.40 0.69 2.79 0.80 <.01
Fear 2.44 0.63 2.89 0.85 <.01
Anger 2.84 0.72 2.79 0.63 .67
Activity 3.03 0.66 3.27 0.69 .05
Sociability 3.70 0.63 3.96 0.71 .05

Bodymassindex 21.92 4.18 23.12 4.15 .13
Restraint (CD) 6.99 3.48 10.50 3.29 <.OOI
Restraint (WF) 3.46 2.75 5.91 3.21 <.001
Alcohol consumption

(30-day prevalence) 39% 56% <.OOI

Note. The maximum possible ns are shown. The ns change for each variable because of missing data (range
for partial syndrome, 28-36; range for asymptomatics, 731-789). CD = Concern for Dieting subscale; WF
= Weight Fluctuation subscale.

opment. Third, because of the apparent rise in incidence of eat-
ing disorders and their clinical importance, both clinicians and
researchers have called for public education and early interven-
tion to prevent the development of disordered eating behaviors
and to promote healthful weight regulation practices among
children and adolescents {Shisslak & Crago, 1987; Striegel-
Moore et al., 1986; Wadden & Stunkard, 1985). Therefore, our
results may be useful to the extent that a focus on weight and

"Weight Concern
Upper Quortile

AGE (Yr)

Figure I . Onset of partial syndrome eating disorder. Yr – year.

shape concerns represents a rational basis for the choice of a
prevention intervention target.

Few of the girls classified as symptomatic in this study met
the full set of criteria for bulimia nervosa from the DSM (4th
ed.; DSM-IV; American Psychiatric Association, 1994).
Rather, most presented with a range of symptoms reflective of
partial syndrome (Shisslak et al., 1995) or the DSM eating dis-
order not otherwise specified category. However, such groups
are deserving of serious research attention. First, several pro-
spective studies suggest that many who suffer from partial syn-
drome disorders progress to full syndrome disorders over time
(Striegel-Moore et al., 1989). Second, partial syndrome eating
disorders are associated with considerable psychological distur-
bance (Killen et al., 1987; Shisslak et al., 1995). For example,
even among the comparatively young female adolescents in this
study, the 30-day drinking prevalence was significantly higher
for those girls who became symptomatic. Third, our focus on
partial syndrome is important, given research indicating that
women suffering from partial syndrome eating disorders seek
evaluation and treatment (Shisslak, et al., 1995). For example,
the percentage of cases diagnosed as partial syndrome was re-
ported to be 40% to 46% at two eating disbrder treatment clinics
in the United States (Herzog, Hopkins, & Burns, 1993; Wil-
liamson, Cleaves, & Savin, 1992). Similarly, in studies of clini-
cal samples of adolescents, 35-50% of girls referred for treat-
ment of eating disorders are classified as partial syndrome
(Bunnel, Shenker, Nussbaum, Jacobsen, & Cooper, 1990; Shis-
slak etal., 1995).

940 KILLEN ET AL.

Girls who developed a partial syndrome eating disorder over

the study interval appeared more temperamentally prone to

emotionality on the basis of their scores on the distress and fear

components of the EAS; yet, when examined from a multivari-

ate perspective, only weight concerns appeared to mediate onset

in this study. This finding does not mean, however, that person-

ality has no influence on the development of eating disorders.

For example, weight and shape concerns appear more specifi-

cally linked and proximal to eating disorders than global mea-

sures of temperament. However, temperament may influence

eating disorders through weight and shape concerns and other

more proximal variables in ways that are not yet understood.

References

Agras, W. S., & Kirkley, B. G. (1986). Bulimia: Theories of etiology. In
K. D. Brownell & J. P. Foreyt (Eds.), Handbook of eating disorders:

Physiology, psychology and treatment of obesity, anorexia, and bu-
limia (pp. 367-378). New York: Basic Books.

American Psychiatric Association, (1987). Diagnostic and statistical

manual of mental disorders (3rd ed., rev.). Washington, DC: Author.
American Psychiatric Association. (1994). Diagnostic and statistical

manual of mental disorder (4th ed.). Vteshington, DC: Author.

Attie, I., & Brooks-Gunn. J. (1989). Development of eating problems
in adolescent girls: A longitudinal study. Developmental Psychology,
25. 70-79.

Bunnell, D. W., Shenker, I. R., Nussbaum, M. P., Jacobsen, M. S., &
Cooper, P. (1990). Subclinical versus formal eating disorders:

Differentiating psychological features. International Journal of Eat-

ing Disorders, 9, 357-362.
Buss, A., & Plomin, R. (1984). Temperament: Early developing person-

ality traits. Hillsdale. HI: Erlbaum.
Casper, R. C., Eckert, E. D., Halmi, K. A., Goldberg, S. C, & Davis,

J. M. (1980). Bulimia: Its incidence and clinical importance in pa-

tients with anorexia. Archives of General Psychiatry, 37,1030-1035.
Cooper, Z., Cooper, P. J., & Fairburn, C. G. (1989). The validity of

the eating disorder examination and its subscales, British Journal of
Psychiatry, 154,807-812.

Fairburn, C. G., & Beglin, S. J. (1990). Studies of the epidemiology of
bulimia ranasai. American Journal of Psychiatry, 4, 402-408,

Garfinkel, P. E., Moldofsky, H., & Gamer, D. M. (1980). The heteroge-
neity of anorexia nervosa: Bulimia as a distinct subgroup. Archives of

General Psychiatry, 37,1036-1040.
Garner, D. M., Olmsted, M. P., & Polivy, J. (1983). The development

and validation of a multidimensional eating disorder inventory for
anorexia nervosa and bulimia. International Journal of Eating Dis-

orders, 2, 15-34.
Graber, J. A., Brooks-Gunn, J., Paikoff, R. L., & Warren, M. (1994).

Prediction of eating problems: An 8-year study of adolescent girls.

Developmental Psychology, 30, 823-834.

Graham, J. W., Flay, B. R., & Johnson, C. A. (1984). Reliability of self-
report measures of drug use in prevention research: Evaluation of the
Project SMART questionnaire via the test-retest reliability matrix.
Journal of Drug Education, 14, 175-193.

Herman, C. P., Polivy, J., Pliner, P., & Threlkeld, J. (1978). Distracti-
bility in dieters and nondieters: An alternative view of "externality."

Journal of Personality & Social Psychology, 5,536-548.
Herzog, D. B., & Copeland, P. M. (1985). Eating disorders. New En-

gland Journal of Medicine, 313, 295-303.

Herzog, D. B., Hopkins, J. D., & Burns, C. D. (1993). A follow-up study

of 33 subdiagnostic eating disordered women. International Journal
of Eating Disorders. 14, 261-267.

Killen, i. D., Taylco; C. B., Hayward, C, Wilson, D. M., Haydel, K. F.,

Robinson, T. N., Litt, I., Simmonds, B. A., Varady, A., & Kraemer,
H. (1994). The pursuit of thinness and onset of eating disorder symp-
toms in a community sample of adolescent girls: A three year pro-

spective analysis. International Journal of Eating Disorders, 16, 227-
238.

Killen, J. D., Taylor, C. B., Telch, M. J., Robinson, T. N., Maron, D. J., &

Sayloj; K. E. (1987). Depressive symptoms and substance use among
adolescent binge eaters and purgers: A denned population study.
American Journal of Public Health, 77, 1539-1541.

Kraemer, H. C., Berkowitz, R. I., & Hammer, L. D. (1990). Method-

ological difficulties in studies of obesity. I: Measurement issues. An-
nals of Behavioral Medicine, 12, 112-118.

Shisslak, C. M., &Crago, M. (1987). Primary prevention of eating dis-
orders. Journal of Consulting and Clinical Psychology, 55,660-667.

Shisslak, C. M., Crago, M., & Estes, L. S. (1995). The spectrum of
eating disturbances. International Journal of Eating Disorders. 18,
209-219.

Striegel-Moore, R. H., Silberetein, L. R., Frensch, P., & Rodin, J.

(1989). A prospective study of disordered eating among college stu-

dents. International Journal of Eating Disorders, 8, 499-509.
Striegel-Moore, R. H., Silberstein, L. R., & Rodin, J. (1986). Toward

an understanding of risk factors for bulimia. American Psychologist,
41, 246-263.

Strober, M. (1984). Stressful life events associated with bulimia in an-

orexia nervosa. International Journal of Eating Disorders, 3,1-16.
Strober, M. (1986). Anorexia nervosa: History and psychological con-

cepts. In K. D. Brownell & J. P. Foreyt (Eds.), Handbook of eating
disorders: Physiology, psychology and treatment of obesity, anorexia
and bulimia (pp. 231 -246). New York: Basic Books.

Wadden, T. A., & Stunkard, A. J. (1985). Social and psychological con-
sequences of obesity. Annals of Internal Medicine, 103, 1062-1067.

Williamson, D. A., Gleaves, D. H,, & Savin, S. S. (1992). Empirical

classification of eating disorder not otherwise specified: Support for
DSM-IV changes. Journal of Psychopathology and Behavioral As-
sessment. 14, 201-216.

Received July …

RESEARCH AND PRACTICE

Eating When There is Not Enough to Eat: Eating Behaviors
and Perceptions of Food Among Food-Insecure Youths

Rachel Widome, PhD, MHS, Dianne Neumark-Sztainer, PhD, RD, MPH, Peter J. Hannan, MStat, Jess Haines, PhD, MHSc, RD, and Mary Story, PhD, RD

Food insecurity, or not having access to enough
food for an active, healthy life because of a lack
of resources, is a continuing problem in the
United States.' The US Department of Agricul-
ture monitors the extent and severity of food
insecurity in US households through the food
security section of the annual, nationally repre-
sentative Current Population Survey. According
to this survey, in 2006,10.9% of households
expedehced food insecurity at some point during
the year.' Households with children tended to he
more affected by food insecurity and were nearly
twice as likely to report food insecurity during at
least part of the past year as were households
with no children under the age of 18 years
(15.6% versus 8.5%, respectively).' Young chil-
dren are often protected from hunger even in
households that have very low food security;
however, adolescents may he more vulnerable.'

Growing up in a food-insecure household
places burdens on youths. Even after control-
ling for family income, adolescents living in
food-insecure households have lower psycho-
sodal functioning^ and a greater risk of having
suiddal symptoms'' than do their food-secure
counterparts. Because of the increasingly preva-
lent childhood obesity epidemic in the United
States,'*'̂ the effect of food security on both
weight outcomes and predictors of obesity is of
spedal interest Youths who are radal/ethnic
minorities, low income, or both are at greater risk
for overweight^® Several studies have found
evidence for a paradoxical assodation between
household food insecurity and overwei^t status
in both children**"" and adults.'^"'^ Other studies
found no relation between food insecurity and
weight in children,'^'® and several studies found
a negative assodation.^°'^' These mixed results
may he because households that are character-
ized as food insecure likely fall at various points
on a spectrum of food insuffidency and have
differing coping strategies.̂ ^ Two possible inter-
connected mechanisms relating to eating habits
that might explain why individuals who are food

more than those who are food

Objectives. We explored differences in adolescents' eating habits, percep-
tions, and dietary intakes by food security status. '•

Methods. As part of Project EAT (Eating Among Teens), we surveyed 4746
multiethnic middle and high school students in 31 pririiarily urban schools in the
Minneapolis-St. Paul, Minnesota, area during the 1998-1999 academic year.
Participants completed in-class surveys. We used multiple regression analysis to
characterize associations between behaviors, perceptions, nutritional intake,
and food security status.

Results. Compared with food-secure youths, food-irisecure youths were more
likely to perceive that eating healthfully was inconvenient and that healthy food
did not taste good. Additionally, food-insecure youths reported eating more fast
food but fewer family meals and breakfasts per week than did youths who were
food secure. Food-insecure and food-secure youths perceived similar benefits
from eating healthfully (P=.75). Compared with those who were food secure,
food-insecure youths had higher fat intakes (P95thpercentüe).'”‘

Age, grade level, gender, and race/ethnidty
were measured by self-report Race/ethnidty
was assessed with the question, “Do you think of
yourself as White, Black/Afiican American,
Hispanic or Latino, Asian American, Hawaiian/
Pacific Islander, or American Indian?” Because of
small numbers, we grouped youths who indi-
cated Hawauan/Padfic Islander into an “Other/
Multiple” category that also induded youths who
indicated more than 1 race/ethnidty.

Data Analysis

We report the demographic breakdowns
for each category of the 2 food security items.
We used multiple linear regression to calcu-
late mean values and their assodated 9 5 %

May 2009, Vol 99, No. 5 | American Journal of Public Health Widome et a/. | Peer Reviewed | Research and Practice | 823

RESEARCH AND PRACTICE

confidence intervals (CIs) to characterize the
associations between behaviors, perceptions,
and nutritional intakes and food security status.
Scales of perceived benefits and barriers to
healthy I eating and food availability were stan-
dardized so that the mean for the whole sample
was equal to zero and the stEindard deviation
was equal to 1. All regression models were
adjusted for race/ethnidty, grade level, and
gender.i Percentages of youths meeting each
Healthy People 2010 goal were reported for
each of, the food security categories ofthe 2
food security questions. We used the Mantei-
Haenszel y^ trend test (1 degree of freedom)
when examining the hunger frequency item.
We used the x^ test (3 degrees of freedom) to
test forlsignificant differences between home
food adequacy categories because this mea-
sure is not strictly ordinal. SAS version 9.1.3
was used for all analyses (SAS Institute Inc,
Cary, NC).

RESULTS

In response to the hunger frequency item,
8.4% of adolescents reported being hungry at
least once in the past year because their family
could not afford food (Table 1). For home food
adequacy, 4.4% of the adolescents reported
that often or sometimes they do not have
enough to eat. Both food security items were
significantly correlated with ethnicity, public
assistance, and eligibility for free or reduced
price lunch (results not shown).

The associations between perceived benefits
and barriers to healthy eating and food security
as assessed by the hunger frequency and home
food adequacy items are shown in Table 2.
Youths who reported a hunger frequency of
“almost every month” in the past year were
more likely than youths in the rest ofthe sample
to report both inconvenience and food prefer-
ence as barriers to healthy eating. However,

how these yoiiths scored on the benefits of
hecilthy eating scale did not differ significantly
from the youths who reported no hunger in
the past year.’ Youths who reported any fre-
quency of hunger were significantly less likely
to report high availability of both unhealthy
and healthy foods in their households. Adoles-
cents who reported that their households
“always have I enough to eat and the kinds of
foods we want” were significantly less likely
than the rest of the sample to indicate that
inconvenience and food preference were bar-
riers to eating healthy. Youths who reported
any home food inadequacy had a lesser
availability of; both unhealthy and healthy
foods in their, households.

The associations between food security and
selected eating patterns are shown in Table 3.
The overall P values were significant for the
assodations between breakfast and eating
family meals and both of thé food seairity

TABLE 1-Description of Project EAT Adolescents, by Food Security Status: Minneapolis-St. Paul, MN, 1998-1999

Overall

Race/Ethnicity

White

Black

Hispanic

Asian

Native American

Other/Multiple

Gender

Male

Female

Grade ievel

Middle school

High school

Public assistance

Yes

No

Free lunch

Yes

No

Row Total,

No.

4589

2243

816

264

871

158

180

2297

2292

1544

2997

490

3534

1149

2208

Almost Every

Month, % (No.)

or%

1.2 (53)

0.5

1.8

0.0

2.1

2.5

1.7

1.3

1.0

1.4

1.0

3.7

0.6

2.0

0.6

Hunger Frequency^

Some Months,

% (No.) or %

2.8 (128)

1.0

3.8

2.7

5.4

3.8

6.1

3.1

2.4

3.8

2.3

6.9

1.5

4.5

1.0

One or

Two Months,

% (No.) or %

4.4 (200)

3.5

4.3

3.8

6.1

7.0

6.1

4.5

4.2

4.6

4.2

10.0

3.4

7.1

2.4

Zero Months,

% (No.) or %

91.7 (4208)

95.0

90.1

93.6

86.5

86.7

86.1

91.0

92.4

90.3

92.56

79.4

94.5

86.3

96.1

Row Total,

No.

4615

2246

829

266

875

161

181

2314

2301

1561

3005

491

3548

1155

2216

Often

Inadequate,

% (No.) or %

1.3 (60)

0.5

1.9

2.6

1.9

2.5

2.2

1.2

1.4

1.9

1.0

2.0

1.0

1.4

0.5

Home

Sometimes

Inadequate,

% (No.) or %

3.1 (145)

2.3

4.1

1.9

4.5

3.7

3.3

3.4

2.9

3.7

2.7

5.3

2.5

4.7

1.85

Food Adequacy’

Adequate but not

Always the Kinds of Food

Wanted, % (No.) or %

33.8 (1560)

37.7

25.5

27.4

36.3

26.7

27.6

32.5

35.1

30.3

35.8

37.3

33.6

36.7

32.76

Adequate and the

Kinds of Foods We Want,

% (No.) or %

61.8 (2850)

59.5

68.5

68.0

57.3

67.1

66.9

62.8

60.7

64.2

60.5

55.4

62.9

57.2

64.89

HoXi. The number of adoiescents in each demographic category varies siightly for each food security outcome because of missing responses.

“Assessed by the question, “How often during the iast 12 months have you been hungry because your family couldn’t afford more food?”

‘Assessed by the question, “Which of the following best describes the food eaten in your household in the last 12 months?”

8 2 4 I Research and Practice | Peer Reviewed | Widome et al. American Journal of Public Health | May 2 0 0 9 . Vol 9 9 , No. 5

RESEARCH AND PRACTICE

TABLE 2-Adjusted Standardized Means of Perceived Benefits and Barriers to Heaithy Eating and Food
Availabiüty Scaies, by Food Security Category: Project EAT, iWinneapoiis-St. Paui, MN, 1998-1999

Perceived barrier

Convenience

Food preference

Perceived benefits

of heaithy eating

Heaithy food

avaiiabie in home

Unheaithy food

avaiiabie in home

Aimost Eveiy

Month,

Mean (95% Ci)

0.51

(0.23, 0.79)

0.45

(0.17, 0.73)

0.00

(-0.27, 0.28)

-0.79

(-1.06, -0.53)

-0.38

(-0.65, -0.10)

Some iHonths,
Mean (95% Ci)

0.37

(0.19, 0.55)

0.05

(-0.12, 0.23)

0.00

(-0.18, 0.18)

-0.57

(-0.74, -0.40)

-0.26

(-0.43, -0.08)

Hunger Frequency^

One or Two
Months,

Mean (95% Ci)

0.07
(-0.07, 0.20)

0.01

(-0.13, 0.15)

-0.05

(-0.19, 0.09)

-0.45

(-0.58, -0.32)

-0.22

(-0.36, -0.09)

Zero Months,
Mean (95% Ci)

-0.02

(-Ö.05, 0.01)

-0.02

(-0.05, 0.02)

0.01

(-0.03, 0.04)

0.06

(0.03, 0.09)

0.03

(0.00, 0.05)

P, for Trend

<.OO1

.008

.745

<.OO1

<.OO1

Often

Inadequate,
Mean (95% Ci)

0.24

(-0.03, 0.51)

0.14
(-0.13, 0.41)

-0.09

(-0.35, 0.18)

-0.39

(-0.64, -0.14)

-0.27

(-0.53, -0.01)

Home

Sometimes
Inadequate,

Mean (95% CI)

0.38

(0.21, 0.54)

0.09

(-0.08, 0.26)

-0.07

(-0.24, 0.10)

-0.73

(-0.89, -0.57)

-0.32

(-0.48, -0.16)

Food Adequacy''

Adequate but
not Always the
Kinds of Food

Wanted,
Mean (95% CI)

0.15

(0.10, 0.20)

0.11

(0.06, 0.16)

-0.02

(-0.07, 0.03)

-0.21

(-0.26, -0.17)

-0.11

(-0.16, -0.06)

Adequate and

the Kinds of
Foods we Want,
Mean (95% Ci)

-0.11

(-0.14, -0.07)

-0.08

(-0.12, -0.05)

0.02

(-0.02, 0.06)

0.18

(0.14, 0.21)

0.09

(0.05, 0.12)

<.OO1

<.OO1

.380

<.OO1

<.OO1

«oie. EAT »Eating Among Teens; Ci-confidence intervai. Scaies have been standardized so that for whole sample, mean-0 and SO = 1. AH estimates were adjusted for ethnicity grade ievei (middie
schooi vs high school), and gender.
^Assessed by the question, "How often during the last 12 months have you been hungry because your famiiy couldn't afford more food?"
Assessed by the question, "Which of the foilowing best describes the food eaten in your househoid in the iast 12 months?"

'^ype 3 sum of squares.

measures. Fully food-secure youths ate family
meals and breakfast more often than did the
other groups. Though none of the overall P
values for fast food were significant, youths
who reported a home food inadequacy of

"often" ate an average of approximately 2.15
(95% CI = 1.74, 2.56) fast-food meals per
week compared with 1.73 fast food meals
eaten by youths who reported no hunger in
the past year (Table 3). Youths who reported

hunger frequency during some months (2.03
[95% CI=1.75, 2.31]) ate slightly more fast-
food meals than did those who were hungry
every month (1.70 meals) or zero months
(1.72 meals).

TABLE 3-Associations of Food Security and Eating: Project EAT, Minneapoiis-St. Paui, MN, 1998-1999

Almost Eveiy Month,

Mean (95% CI)

Fast Food 1.70

(1.27, 2.14)

Family Meals 2.85

(2.13, 3.58)

Breai<fast 3.32

(2.60, 4.03)

Hunger Frequency"

Some Months,
Mean (95% Ci)

2.03

(1.75, 2.31)

3.37

(2.90, 3.84)

3.35

(2.85, 3.81)

One or Two

Months,
Mean (95% Ci)

1.84

(1.62, 2.07)

3.31

(2.94, 3.68)

3.54

(3.18, 3.90)

Zero Months,
Mean (95% Ci)

1.72

(1.67, 1.77)

4.19

(4.11, 4.28)

3.91

(3.83, 3.99)

P, for

Trend

.088

<.OO1

.001

Often

inadequate.
Mean (95% CI)

2.15

(1.74, 2.56)

3.29

(2.60, 3.98)

3.39

(2.72, 4.06)

Home

Sometimes
inadequate.

Mean (95% Ci)

1.74

(1.48, 2.01)

2.76

(2.32, 3.19)

3.44

(3.02, 3.87)

Food Adequacy'

Adequate but

not Always the

Kinds of Food
Wanted,

Mean (95% CI)

1.74

(1.66, 1.82)

3.54

(3.40, 3.67)

3.64

(3.51, 3.77)

Adequate and

Foods we
Want,

Mean (95% CI)

1.73

(1.67, 1.79)

4.53

(4.43, 4.62)

4.03

(3.93, 4.12)

pc

.259

<.OO1

95th percentile

Fat intake

< 3 0 % of calories from fat

1300 mg

Fruit, ivegetable, and grain intake

> 2 servings fruit

> 3 servings vegetables

> 3 servings vegetables”

> 6 servings of grain”

Sodium intakes2400 mg

Healthy
People 2010

Target, %

5

75

75

75

50

Almost Every
Month, %

17.4

39.5

44.2

37.2

51.2

27.3

25.0

11.1

48.8

Hunger Frequency^

Some
Months, %

24.3

43.0

41.0

29.0

41.2

21.2

14.4

5.7

62.0

One or Two
Months, %

16.5

42.9

38.0

31.5

39.8

15.3

12.5

3.8

57.1

Zero
Months, %

14.2

53.1

43.9

36.5

46.1

14.2

9.4

4.6

57.0

P, for
Trend

.010

<.OO1

.331

.146

.400

.003

<.OO1

.153

.974

Often
Inadequate,

%

20.4

43.8

45.8

25.0

40.4

20.4

18.4

4.0

56.3

Home! Food Adequacy'

Sometimes
Inadequate,

%

14.3

46.0

38.7

28.2

44.5

16.1

12.7

4.8

60.5

Adequate
but not

Always the
Kinds of Food

Wanted, %

14.9

50.5

41.5

32.1

39.8

12.0

7.9

3.7
62.2

Adequate
and the
Kinds of
Foods we
Want, %

14.4

53.7

44.9

38.9

49.4

15.8

10.6

5.1
54.0

pC

' .670

' .064

.134

; <.ooi

<.OO1

, .008

' .006

.236

; <.OÓ1

Noté. EAT=.Eating Among Teens; BMI-body mass index.
"Assessed by the question, "How often during the last 12 months have you been hungiy because your family couldn't afford more food?"
''Assessed by the question, "Which of the following best describes the food eaten in your household in the last 12 months?"
"With 3 degrees of freedom.
"Of which at least 1 is a deep yellow or green vegetable.
"Of which 3 are whole grain.

Food-insecure adolescents were less likely to
meet the Healthy People 2010 goal for per-
centage of calories from fat (Table 4). For the
hunger; frequency outcome item, food-insecure
youths;were less likely to meet the goal of
less than 30% of calories from fat Despite this,
they appeared to be more likely to meet
goals related to vegetable intake than were
their food-secure peers, but percentages were
well below the Healthy People 2010 targets for
all groups. Additionally, youths who reported
no huhger in the past year were least likely to
have a|BMI greater than or equal to the 95th
percentile. For the home food adequacy out-
come, food-insecure youths were less likely to
meet the caldum goal, fruit goal, and possibly
the goal of less than 30% of calories from fat.
There were significEuit differences between
groups for home food adequacy for the sodium
intake goal. Youths reporting that they have
enough food in their house but not always the
kinds they want were most likely to meet the
sodiurri intake goal. Youths who reported "often
inadequate" food in their homes were most
likely to meet vegetable goals.

DISCUSSION

We found that food-insecure youths had
several known eating-related risk factors for
overweight. Food-insecure youths consumed a
greater percentage of calories from fat and ate
fewer family meals and breakfasts. Our evi-
dence suggested that these youths may also
eat more fast-food meals. They also had less
food available in the home (both healthy and
unhealthy foods) and perceived greater bar-
riers to eating healthfully than did their food-
secure counterparts. However, encouragingly,
they did not perceive fewer benefits from
eating healthfully and appeared to eat signif-
icantly more vegetables than did their food-
secure peers. The group with the largest per-
centage of youths with a BMl greater than the
95th percentile was the group reporting a
hunger frequency of "some months but not
every month." Previoiis research has shown
that adolescents from low-income households
and those who are radal/ethnic minorities are
at greater risk for overweight,®"® and the
impact of food insecurity on eating behavior

may be 1 mechanism behind this observed
assodation.

Although ño food security groups came close
to the Healthy People 2010 target of 50% of
adolescents eating 3 servings of vegetables with
at least 1 serving being a deep-yellow or green
vegetable, it is interesting that youths who
reported that they were hungry nearly every
month or often had inadequate food available
in the home ,were more likely to achieve this
goal than were the food-secure youths. This
may be because of cultural factors or because
these youths! had access to more vegetables
through assistance programs such as j free or
reduced-price school breakfast and lunch, food
shelves, or nieals served at shelters. Future
resecirch should explore this dietary strength.

As might be expected, youths reporting food
insecurity also reported less home availability
of healthy food, as reported previously in an
analysis examining correlates of fruit and
vegetable intake among Project EAT pardd-
pants."*^ But I food-insecure youths did not seem
to have a greater absolute amount of'Unhealthy
food in theirihomes. However, the proportion of

826 I Research and Practice | Peer Reviewed | Widome et al. American Journal of Public Health | May 2009, Vol 99, No. 5

RESEARCH AND PRACTICE

healthy to unhealthy food in food-insecou-e
households appears to be less favorable, and this
may influence adolescents' eating choices.

Adolescents who reported that they "often"
did not have enough to eat or that they expe-
rienced hunger "some months" also reported
eating more fast food than did those who
were food secure. The overall P value across
the food security categories, however, was
not significant. In line with the idea that
food-insecure families may choose more en-
ergy-dense foods,̂ ^"^^ it follows that fast food
may be eaten more often by food-insecure ado-
lescents than by youths who come from food-
secure families. Past research has shown that
frequent consumption of fast food is assodated
with reduced availability of heal% food in the
home,*®'̂ ^ which could further impede healAy
eating. Youths who stated that their families
could not afford food "almost every month"
reported similar fast food use as those who said
they had not been hungry. It is possible that
households where money for bu3nng food is
most severely and consistently limited might not
be able to afford fast food, whereas households
where the situation is less dire may be more …

O R I G I N A L A R T I C L E

The relationship between social media use and disordered
eating in young adolescents

Simon M. Wilksch PhD1 | Anne O'Shea PhD1 | Pheobe Ho BSc2 |

Sue Byrne PhD2 | Tracey D. Wade PhD1

1School of Psychology, Flinders University,

Adelaide, South Australia, Australia

2School of Psychology, University of Western

Australia, Crawley, Western Australia,

Australia

Correspondence

Simon M. Wilksch, School of Psychology,

Flinders University, GPO Box 2100, Adelaide

5001, South Australia, Australia.

Email: [email protected]

Funding information

Australian Rotary Health, Grant/Award

Number: Mental Health Grant

Abstract

Background: The relationship between social media (SM) use and disordered eating

(DE) has not been adequately explored in young adolescents.

Methods: Data from 996 Grade 7 and 8 adolescents (n = 534 girls; M age = 13.08)

was investigated. DE cognitions (Eating Disorder Examination-Questionnaire [EDE-

Q]), DE behaviors (Project Eating Among Teens), and SM use measures related to

Facebook, Instagram, Snapchat, and Tumblr were completed.

Results: DE behaviors were reported by 51.7% of girls and 45.0% of boys, with strict

exercise and meal skipping the most common. A total of 75.4% of girls and 69.9% of

boys had at least one SM account where Instagram was the most common, used by

68.1% of girls and 61.7% of boys. Global EDE-Q scores were significantly higher for

girls and boys with each type of SM account, except for Facebook and Instagram for

girls. A greater number of SM accounts was associated with higher DE scores for

both cognitions and behaviors. Girls with Snapchat and Tumblr accounts and boys

with Snapchat, Facebook and Instagram were significantly more likely to have both

DE behaviors and over-evaluation of shape and weight in the clinical range. Greater

daily time spent using Instagram was associated with significantly higher Global EDE-

Q scores and DE behaviors for girls, while this pattern was also found for Snapchat

usage and DE behaviors for girls.

Conclusions: A clear pattern of association was found between SM usage and DE

cognitions and behaviors with this exploratory study confirming that these relation-

ships occur at younger-age than previously investigated.

K E Y W O R D S

eating disorders, prevention, risk factors, social media

1 | INTRODUCTION

The relationship between media usage, body image, and eating disor-

der risk has been studied for decades. Initially, magazines and televi-

sion were the primary forms of media examined, while more recently

online media and particularly social media (SM) has been explored

(Holland & Tiggemann, 2016). However, while there has been a prolif-

eration of research investigating the relationship between SM and

body image-related constructs (e.g., body dissatisfaction and objectifi-

cation: Holland & Tiggemann, 2016), the relationship between SM use

and disordered eating (DE), particularly DE behaviors (e.g., skipping

meals, binge eating, and compensatory behaviors) has received much

less attention. Further, the majority of this research has been
Trial registry name: Australian New Zealand Clinical Trials Registry; URL: http://www.anzctr.org.au;

Registration identification number: ACTRN12619000851167

Received: 1 August 2019 Revised: 28 October 2019 Accepted: 28 October 2019

DOI: 10.1002/eat.23198

96 © 2019 Wiley Periodicals, Inc. Int J Eat Disord. 2020;53:96–106.wileyonlinelibrary.com/journal/eat

https://orcid.org/0000-0002-2041-7503

https://orcid.org/0000-0003-4402-770X

mailto:[email protected]

http://www.anzctr.org.au

http://wileyonlinelibrary.com/journal/eat

conducted with young-adult women (e.g., Hummel & Smith, 2015; Smith,

Hames, & Joiner, 2013). Thus, the relationship between SM and DE in

young-adolescent girls and boys has not been adequately investigated.

This needs to be addressed given that 13 years is the minimum required

age for accessing many SM accounts (e.g., Facebook, Instagram,

Snaphcat, and Tumblr), and early adolescence is a time of increased DE

risk (Gowers & Shore, 2001). Further, given that media literacy is the

leading approach to ED risk reduction in young-adolescence (Wilksch

et al., 2015), if SM is associated with increased DE risk, content targeting

SM use could readily be incorporated in such programs.

DE consists of both cognitions (measured in the current study

with the Eating Disorder Examination-Questionnaire global [EDE-Q

Global]: Fairburn & Beglin, 1994) and behaviors (Haines, Neumark-

Sztainer, Eisenberg, & Hannan, 2006). A recent Australian study found

25.3% of adolescents aged 13 and 35.4% aged 14 were engaging in

DE behaviors (N = 202: Sparti, Santomauro, Cruwys, Burgess, & Har-

ris, 2019) as assessed by the Youth Risk Behavior Surveillance System

(Kolbe, Kann, & Collins, 1993). Another Australian study investigating

DE behaviors (EDE-Q) in adolescents aged 12–13 (N = 433: Bentley,

Gratwick-Sarll, Harrison, & Mond, 2015; Mond et al., 2014) found:

15.2% of girls and 11.2% of boys to be engaging in objective binge-

eating episodes; 3.0% of girls and 2.% of boys to be engaging in self-

induced vomiting; and, 18.6% of girls and 20.4% of boys to be engag-

ing in compulsive exercise. The proportion of girls and boys with clini-

cal levels (≥4) of over-evaluation of shape or weight (Fairburn &

Beglin, 1994) suggestive of current or future DE (Gowers & Shore,

2001) was 17.3 and 4.6%, respectively. Mean levels of EDE-Q Global

scores were 1.11 (SD = 1.27) for girls and 0.55 (SD = 0.81) for boys

(Bentley et al., 2015). This suggests that DE is already common by

early-adolescence, consistent with international findings (Croll,

Neumark-Sztainer, Story, & Ireland, 2002).

In 2017 Australians aged 14–17 spent an average of 3.3 hr daily

on SM, compared to 2.6 hr by adults (Australian Psychological Society,

2017). Facebook, Instagram, and Snapchat were most commonly used,

by 80.8, 66.0, and 64.7% of adolescents, respectively. This study also

identified that 60% of parents reported never monitoring their child's

SM use and 15% of adolescent Facebook users reported being con-

tacted daily by strangers. The nature of SM uptake and preferred plat-

forms is rapidly changing but the current study chose to focus on the

investigation of the relationship between young-adolescent SM use

of four specific sites—Facebook, Instagram, Snapchat, and Tumblr—

and DE cognitions and behaviors. These platforms were informed by

previous body image research with young-adolescents (Slater,

Varsani & Diedrichs, 2017; Tiggemann & Slater, 2013).

In the context of eating disorder risk, SM is thought to impact the

well-established risk factors of media internalization and peer influences

(Mabe, Forney, & Keel, 2014). A review of the relationship between SM

and body image and DE outcomes by Holland and Tiggemann (2016)

found that, of the 20 studies included, only four included a measure of

DE as an outcome variable with three of these being with samples of

undergraduate adults focusing on Facebook usage (Hummel & Smith,

2015; Mabe et al., 2014; Smith et al., 2013). The fourth study (Ferguson,

Muñoz, Garza, & Galindo, 2014) explored the relationship between SM

use in 237 Hispanic female adolescents (M = 14.11 years) from Texas,

United States and found no relationship with either television exposure

to the thin ideal or SM use and Eating Attitudes Test scores.

To the best of our knowledge, of the few studies to examine the

relationship between SM use and DE, all apart from one have focused

exclusively on cognitions, while Smith et al. (2013) also included one

specific behavior, namely overeating. Maladaptive Facebook usage

(defined as using Facebook for social comparisons) at baseline was

found to be positively associated with episodes of over-eating at

4-week follow-up (Smith et al., 2013). One of the few studies to use

the EDE-Q to measure DE explored racial differences in young-adult

women (N = 922). Findings indicated that the frequency of Facebook

usage was not related to DE, but users who sought reassurance from

their Facebook followers (e.g., expecting followers to comment on

their posts) had higher DE scores than those who did not (Howard,

Heron, MacIntyre, Myers, & Everhart, 2017). There has been some

exploration of brief measures of DE (SCOFF: Solmi, Hatch, Hotopf,

Treasure, & Micali, 2015) and SM use in young-adults (Sidani, Shensa,

Hoffman, Hanmer, & Primack, 2016) where a significant positive rela-

tionship was found between amount of SM use and eating concerns.

Taken collectively, there is some evidence of a relationship between

DE and SM usage in young-adult women. Thus the primary aim of this

research was to explore rates of DE (cognition and behaviors), SM

usage (Facebook, Instagram, Snapchat, and Tumblr), and the relation-

ship between these variables in young-adolescent girls and boys.

Given this was an exploratory study with younger participants than

previous studies, it was decided to keep measurement of SM use to

fundamental features (e.g., whether the user had an account, time

spent on it, types of pictures posted) than the more nuanced features

that have been measured with young-adult samples (e.g., types of

accounts followed such as fitspiration or thinspiration accounts,

whether the participant uses airbrushing techniques prior to posting

an image on their account).

2 | METHODS

2.1 | Participants

Participants were 996 Grade 7 and Grade 8 girls (n = 534; 53.6%) and

boys (n = 462) from classes across five schools in two Australian

states (South Australia n = 710; Western Australia n = 286:

M age = 13.08 years; SD = .60). All five schools were private schools,

where four were coeducational (n = 894; 89.7%) and one was girls-

only (n = 102; 10.3%). Socioeconomic status of participating schools

was obtained from the Australian government's Index of Community

Socio-Educational Advantage (ICSEA) whereby 1,000 represents the

mean, with a standard deviation of 100 (Australian Curriculum,

Assessment and Reporting Authority, 2011). The mean ICSEA rating

was 1,153 (range = 1,106–1,177), indicating above average socioeco-

nomic advantage. Mean BMI data were around the 50th percentile

(Onis et al., 2007) expected for age for both girls (M = 18.94, SD = 3.44)

and boys (M = 19.37, SD = 3.34). Data regarding participant ethnicity

WILKSCH ET AL. 97

was not collected but the participants were primarily Caucasian.

School recruitment and measure completion occurred between

February and May 2018.

2.2 | Procedure

Measure completion occurred in the context of baseline data collec-

tion for an eating disorder risk reduction randomized-controlled trial.

Following parental consent for assessment completion, students com-

pleted questionnaires online in class time under the supervision of

usual class teacher and a research assistant. Approval for this research

was received from the Flinders University Social and Behavioural

Research Ethics Committee and Principals of participating schools.

2.3 | Measures

2.3.1 | DE–Cognitions

The EDE-Q was used to provide a continuous measure of DE cogni-

tions. This self-report version of the EDE interview is widely used in

both risk factor and intervention research in the field (Bentley et al.,

2015; Mond et al., 2014; Wilksch et al., 2015; Wilksch & Wade,

2009). The 22 items that form four subscales (Shape Concern, Weight

Concern, Restraint, and Eating Concern) and a total Global score were

used. Mean item scores on these items range from 0 to 6, with higher

scores indicating higher levels of DE. The internal consistency of the

global EDE-Q in this sample was α = .96 and α = .94 for girls and boys,

respectively, consistent with reliability findings in other Australian

adolescent samples (Mond et al., 2014). The validity of the EDE-Q

with adolescent girls and boys has been confirmed, though less differ-

entiation between the Shape Concern and Weight Concern subscales

was found than in adult samples (White, Haycraft, Goodwin, & Meyer,

2014). It was decided to not measure the behavioral items given find-

ings that the validity of behavioral items is lower for the questionnaire

than interview formats and the young-adolescent age of the sample

(Berg, Peterson, Frazier, & Crow, 2012).

2.3.2 | DE–Behaviors

The Project EAT (Eating Among Teens) questionnaire was used to

measure DE behaviors (Haines et al., 2006). Participants were asked

to respond with No (0) or Yes (1) to the stem question: “Have you

done any of the following things in the last 12 months in order to lose

weight or keep from gaining weight?” with respect to the following

behaviors: skipped meals; ate very little food; a strict eating plan; made

myself vomit (throw up); and, a strict exercise program. Frequency of

behaviors was not assessed. An additional item measured binge eating

“In the past year, have you ever eaten so much food in a short period

of time that you would be embarrassed if others saw you (binge-eat-

ing)?” where a response of yes resulted in an additional question

addressing loss of control (“During the times when you ate this way,

did you feel you couldn't stop eating or control what or how much

you were eating?”). To be counted as a case with binge eating, both

questions required a Yes response. This measure has been used in

numerous longitudinal risk factor studies and has been found to be

valid and reliable in young-adolescent samples (Neumark-Sztainer

et al., 2002). It was developed through focus group discussions with

youth, social cognitive theory as a theoretical framework, in-depth lit-

erature review of similar measures, input from various pediatric men-

tal health experts, leading to extensive pilot testing, further revisions

and the final measure (Ackard, Neumark-Sztainer, Story, & Perry,

2003; Neumark-Sztainer et al., 2002). It has also been used previously

in school-based eating disorder risk reduction trials (Wilksch et al.,

2015, 2017).

Given that three DE behaviors were of a food restriction nature

we examined each of these three behaviors in a simultaneous linear

regression with EDE-Q Global as the dependent variable. For both

girls (ate very little food β = .28, meal skipping β = .29, and strict meal

plan β = .29) and boys (ate very little food β = .19, meal skipping

β = .29, and strict meal plan β = .27), each of the three items predicted

unique variance in EDE-Q Global scores (p < .001) and were therefore

retained separately in subsequent analyses.

An additional item of DE was assessed by requiring (a) the pres-

ence of at least one DE behavior (i.e., Project EAT items) and (b) an

over-evaluation of shape or weight score in the clinical range (≥4) on

the two relevant items from EDE-Q (Has your weight/shape influenced

how you think about [judge] yourself as a person?). Over-evaluation of

shape and weight is suggestive of current or future DE (Gowers &

Shore, 2001; Wilksch & Wade, 2010). The use of this combined item

was to provide an indicator of clinical significance where the com-

bined presence of both DE cognitions and behavior, has been used as

an indicator of DE in risk factor and intervention research (Wade,

Wilksch, & Lee, 2012).

2.3.3 | SM use

SM use was assessed using items from previous body image research

in adolescents (Slater et al., 2017; Tiggemann & Slater, 2013). Items

included No (0) or Yes (1) responses to the following questions

“Please indicate which of the following accounts you have?”

(Facebook; Instagram; Snapchat; Tumblr); “Do you have a parent who

is one of your friends/followers for this online account?”(No, Yes), “Is

your profile set to a public mode on this account?” (No, Yes) and

“How much time do you spend on this account on a typical day?”

(0 = no time; 1 = 30 min; 2 = 1 hr; 3 = 2 hr; 4 = 3 hr; 5 = 4 hr; 6 = 5 hr;

7 = 6+ hr daily). Participants were then asked a series of questions

about photo posting: “What do your posted pictures mainly consist

of?” with No (0) or Yes (1) responses for types of photos including:

“Selfies; Pictures of yourself or friends taken by someone else; Food;

Possessions/items; Scenery and places; Animals; Other people (family,

friends, celebrities), and Memes/quotes.” Twitter use was not mea-

sured, as image-based SM platform use is more common in

98 WILKSCH ET AL.

adolescents with one Australian study finding 84.4% of Grade 7 and

8 girls and boys never having used Twitter (Paxton, 2019).

2.4 | Statistical analyses

SPSS (version 25) was used to conduct analyses. EDE-Q Global and

subscale scores were standardized and compared between girls and

boys using logistic regressions. Logistic regressions were used to com-

pare girls and boys on their frequency of individual DE behaviors and

also to compare girls and boys on SM account usage including: having

an account; if it was publicly viewable; if a parent was a follower; time

spent using the account; whether pictures were posted and if so, what

type of pictures were posted. For each of the above analyses, gender

was the predictor variable.

All subsequent analyses were conducted separately by gender.

Logistic regressions were used to examine if standardized EDE-Q

global scores were related to each type of SM account (predictor vari-

able). ANOVAs were used to investigate whether a cumulative rela-

tionship was found between total number of SM accounts (0, 1,

2, 3–4) and mean EDE-Q Global scores (including Bonferroni-adjusted

post-hoc testing), while logistic regressions were used to explore this

relationship between number of SM accounts (predictor variable) and

DE behaviors. Logistic regressions were used to investigate the rela-

tionship between SM account type (predictor variable) and individual

DE behaviors. Linear regressions explored the relationship between

time spent using SM (with time scored in line with the distribution of

responses: 0 = 0–30 min; 1 = 1 hr; 2 = 2+ hr) and DE cognitions (EDE-

Q Global), while logistic regressions were used for the relationship

between time spent using SM (predictor variable) and DE behaviors.

Finally, types of images posted were simultaneously entered to

explore their relationship with Global EDE-Q (linear regression) and

DE behaviors (logistic regressions).

3 | RESULTS

3.1 | DE cognitions and behaviors

Table 1 reports mean scores for girls and boys for the respective EDE-

Q scales and frequencies of DE behaviors. Girls scored significantly

higher on each EDE-Q scale. Of the DE behaviors, skipping meals, eat-

ing very little food, and binge eating were reported by significantly

more girls than boys. Over half of girls (51.7%) reported at least one

DE behavior compared to 45.0% of boys. The presence of a DE

behavior and over-evaluation of shape and weight in the clinical range

was three times more common in girls (14.0%) than boys (5.2%).

3.2 | SM use

SM use data is reported in Table 2. A significantly higher proportion

of girls than boys reported having Instagram and Tumblr accounts,

while more boys than girls reported having Facebook. Of those with

an account, boys were significantly more likely to have it set to Public

for Instagram and Snapchat. Over half of both girls and boys did not

have a parent as a follower on Snapchat. A significantly higher propor-

tion of boys than girls did not have a parent follower for Facebook,

Instagram, and Snapchat. Girls reported being more likely to post pic-

tures of themselves or friends on both Instagram and Snapchat, and

pictures of other people (e.g., family, celebrities) on both Instagram

and Snapchat. Girls were also much more likely than boys to post

TABLE 1 Disordered eating by girls and boys

Girls Boys Group difference

(n = 532) (n = 461) OR (95% CI)

EDE-Q scores M (SD) M (SD)

Weight concerns (0–6) 1.79 (1.56) 1.22 (1.19) 0.60 (0.53–0.70)

Shape concern (0–6) 2.06 (1.66) 1.31 (1.33) 0.59 (0.52–0.68)

Eating concern (0–6) 0.94 (1.20) 0.65 (0.98) 0.78 (0.66–0.87)

Restraint (0–6) 0.86 (1.20) 0.63 (1.04) 0.80 (0.71–0.92)

EDE-Q Global (0–6) 1.52 (1.30) 1.00 (1.02) 0.62 (0.54–0.72)

Project EAT behaviors N (%) N (%) OR (95% CI)

Skipped meals 162 (30.6) 83 (18.1) 1.99 (1.47–2.69)

Ate very little food 137 (25.8) 63 (13.7) 2.19 (1.58–3.05)

Strict meal plan 91 (17.2) 63 (13.7) 1.31 (0.92–1.86)

Vomit 13 (2.5) 17 (3.7) 0.66 (0.32–1.36)

Strict exercise 172 (32.5) 144 (31.4) 1.05 (0.80–1.37)

Binge eating (LOC) 67 (12.6) 27 (5.9) 2.32 (1.46–3.70)

OE SW ≥ 4 + DE behavior 75 (14.0) 24 (5.2) 2.99 (1.86–4.83)

Abbreviations: CI, confidence intervals; M, mean; OE SW ≥ 4 + DE behavior, over-evaluation of shape and weight mean item score of 4 or above and the

presence of at least one disordered eating behavior; OR, odds ratios for logistic regressions with a significant difference are bolded; SD, standard deviation.

WILKSCH ET AL. 99

T
A
B
L
E
2

F
re
q
u
e
n
ci
e
s
o
f
so
ci
al
m
e
d
ia
ac
co

u
n
ts
,d

e
ta
ils

o
f
ac
co

u
n
t
se
tt
in
g
s,
an

d
ty
p
e
s
o
f
p
h
o
to
s
p
o
st
e
d

A
cc
o
u
n
t
ty
p
e
N
(%

)
A
cc
o
u
n
t
d
e
ta
ils

(N
%
)

T
y
p
e
s
o
f
p
h
o
to
s
N
(%

)

H
av

e
ac
co

u
n
t

P
u
b
lic

N
o
p
ar
e
n
t

T
im

e
u
si
n
g

P
o
st

p
ic
tu
re
s

S
e
lf
ie
s

S
e
lf
b
y
o
th
e
rs

F
ri
e
n
d
s,
ce

le
b
s

F
o
o
d
s

A
n
im

al
s

S
ce

n
e
ry

M
e
m
e
s

P
o
ss
e
ss
io
n
s

F
a
ce
b
o
o
k

G
ir
ls

6
1
(1
2
.1
)

7
(1
1
.7
)

9
(1
5
.0
)

0
.8
2
(1
.0
5
)

2
7
(3
0
.0
)

4
(1
4
.8
)

1
6
(5
9
.3
)

1
2
(4
4
.4
)

6
(2
2
.2
)

1
2
(4
4
.4
)

1
6
(5
9
.3
)

9
(3
3
.3
)

4
(1
4
.8
)

B
o
y
s

1
0
5
(2
4
.1
)

5
(2
2
.1
)

3
3
(3
2
.4
)

0
.9
8
(1
.0
9
)

4
6
(4
4
.7
)

4
(9
.1
)

2
6
(5
9
.1
)

2
1
(4
7
.7
)

9
(2
0
.0
)

1
5
(3
3
.3
)

2
6
(5
6
.5
)

1
6
(3
6
.5
)

1
1
(2
5
.0
)

O
R
(9
5
%

C
I)

0
.4
3
(0
.3
1

0
.6
1
)

0
.3
7
(0
.1
6

0
.8
4
)

In
st
a
gr
a
m

G
ir
ls

3
4
8
(6
8
.1
)

2
3
(6
.6
)

9
2
(2
6
.6
)

2
.0
2
(1
.2
4
)
3
2
5
(9
3
.9
)

6
8
(2
1
.1
)
2
7
0
(8
3
.1
)

1
9
1
(5
9
.1
)

7
3
(2
2
.6
)

1
5
1
(4
6
.7
)

2
3
7
(7
3
.1
)

4
4
(1
3
.6
)

5
4
(1
6
.7
)

B
o
y
s

2
7
4
(6
1
.7
)

3
5
(1
2
.8
)

1
0
9
(4
0
.1
)

1
.8
9
(1
.4
5
)
2
2
4
(8
2
.1
)

3
3
(1
4
.9
)
1
3
8
(6
2
.2
)

9
5
(4
2
.8
)

4
0
(1
7
.9
)

9
4
(4
2
.2
)

1
4
9
(6
7
.1
)

7
1
(3
1
.8
)

8
5
(3
8
.3
)

O
R
(9
5
%

C
I)

1
.3
3
(1
.0
1

1
.7
3
)2

.0
7
(1
.1
9

3
.5
9
)0

.5
4
(0
.3
9

0
.7
6
)

3
.3
9
(1
.9
8

5
.8
0
)

2
.9
9
(2
.0
1

4
.4
5
)1

.9
3
(1
.3
7

2
.7
3
)

0
.3
4
(0
.2
2

0
.5
2
)0

.3
2
(0
.2
2

0
.4
8
)

S
n
a
p
ch
a
t

G
ir
ls

3
0
1
(5
8
.9
)

1
7
(5
.7
)

1
5
8
(5
2
.7
)

2
.4
7
(1
.6
3
)
2
7
1
(9
0
.3
)

1
4
4
(5
3
.5
)
1
6
7
(6
2
.1
)

1
7
9
(6
6
.3
)

1
6
9
(6
2
.8
)

1
8
9
(7
0
.8
)

2
1
8
(8
1
.0
)

6
3
(2
3
.6
)

1
2
0
(4
4
.9
)

B
o
y
s

2
4
2
(5
4
.4
)

4
5
(1
8
.9
)

1
6
0
(6
7
.2
)

2
.2
2
(1
.7
2
)
2
1
1
(8
8
.7
)

9
8
(4
6
.7
)

9
8
(4
6
.7
)

9
9
(4
7
.1
)

6
3
(3
0
.0
)

1
0
4
(4
9
.5
)

1
2
2
(5
8
.4
)

7
7
(3
6
.7
)

9
7
(4
6
.2
)

O
R
(9
5
%

C
I)

3
.8
8
(2
.1
6

6
.9
8
)0

.5
4
(0
.3
8

0
.7
7
)

1
.8
7
(1
.3
0

2
.7
0
)2

.2
1
(1
.5
2

3
.1
9
)3

.9
4
(2
.6
8

5
.8
0
)2

.4
7
(1
.6
9

3
.6
0
)3

.0
5
(2
.0
2

4
.6
0
)0

.5
3
(0
.3
6

0
.7
9
)

T
u
m
b
lr

G
ir
ls

4
2
(8
.3
)

1
3
(3
1
.0
)

3
6
(8
5
.7
)

0
.8
1
(1
.3
5
)

9
(2
1
.4
)

1
(1
1
.1
)

2
(2
2
.2
)

4
(4
4
.4
)

1
(1
1
.1
)

4
(4
4
.4
)

3
(3
3
.3
)

7
(7
7
.8
)

4
(4
4
.4
)

B
o
y
s

1
6
(3
.7
)

5
(3
3
.3
)

1
3
(8
6
.7
)

1
.0
0
(.4

1
)

4
(2
6
.7
)

2
(5
0
.0
)

2
(5
0
.0
)

1
(2
5
.0
)

2
(5
0
.0
)

1
(2
5
.0
)

2
(5
0
.0
)

3
(7
5
.0
)

1
(2
5
.0
)

O
R
(9
5
%

C
I)

2
.3
6
(1
.3
1

4
.2
7
)

A
b
b
re
v
ia
ti
o
n
s:
C
I,
co

n
fi
d
e
n
ce

in
te
rv
al
s;
O
R
,o

d
d
s
ra
ti
o
s
fo
r
lo
g
is
ti
c
re
g
re
ss
io
n
s
w
it
h
a
si
g
n
if
ic
an

t
d
if
fe
re
n
ce

b
e
tw

e
e
n
g
ir
ls
an

d
b
o
y
s
ar
e
re
p
o
rt
e
d
;
B
e
tw

e
e
n
-g
ro
u
p
co

m
p
ar
is
o
n
s
th
at

ar
e
si
g
n
if
ic
an

tl
y
d
if
fe
re
n
t
ar
e

b
o
ld
e
d
(p

<
.0
5
).

100 WILKSCH ET AL.

pictures of food on Snapchat. Conversely, boys were more likely to

post pictures of Memes (Instagram and Snapchat) and possessions

(Instagram) than girls.

The modal number of followers was 0–50 for each account other

than Instagram, where 100–500 was the modal number for both girls

and boys. Girls had the following numbers of accounts: 0 = 24.7%;

1 = 19.7%; 2 = 42.3%; 3 = 11.7%; and, 4 = 1.6%. The commensurate

numbers for boys were: 0 = 30.2%; 1 = 20.9%; 2 = 28.5%; 3 = 18.6%;

and, 4 = 1.9%.

3.3 | Relationship between SM accounts and DE
cognitions

Table 3 presents DE cognitions (Global EDE-Q) by SM account type.

Snapchat and Tumblr were associated with significantly higher levels

of DE for girls, while all SM accounts were associated with higher DE

for boys. Bonferroni-adjusted post-hoc analyses were conducted to

explore if EDE-Q subscales (Shape & Weight concern, Eating Concern,

Dietary Restraint) uniquely contributed to this relationship. The only

significant predictor was Shape and Weight Concern scores as a pre-

dictor of Snapchat use by girls (OR = 1.66, 95% CI [1.26–2.18]).

The relationship between total number of SM accounts and DE

was investigated using one-way ANOVA (see Table 3). DE increased

as did the number of SM accounts participants had for both girls (F

[3,498] = 4.01, p < .001) and boys (F[3,426] = 5.77, p < .001). For girls,

those with no SM had significantly lower DE than those with 2 or 3–4

accounts. For boys, those with 3–4 SM accounts had significantly

higher DE scores than those with zero through to two accounts.

3.4 | Relationship between SM accounts and DE
behaviors

Table 4 presents the proportion of participants engaging in DE behav-

iors based on SM account type. For girls, Snapchat use was associated

with increased likelihood of eating little food, meal skipping, and fol-

lowing a strict meal plan. All SM platforms were associated with

increased likelihood of strict exercise in girls. Tumblr was the only

platform associated with increased risk of binge eating, with this

found for both girls and boys. For boys, all platforms were associated

with increased risk of meal skipping, while Tumblr use was also associ-

ated with eating little food. Girls with Snapchat or Tumblr accounts,

and boys with Facebook, Instagram, or Snapchat accounts were signif-

icantly more likely to have both a DE behavior and presence of over-

evaluation of shape and weight in the clinical range.

The relationship between total number of SM accounts and DE

behaviors was investigated using logistic regressions (see Table 3). Fre-

quency of DE behaviors increased as did the number of SM accounts

participants had for both girls and boys. For girls, those with no SM had

significantly lower likelihood of DE behaviors than those with 2 or 3–4

accounts. For boys, those with 3–4 SM accounts had significantly

higher likelihood of DE scores than those with no SM accounts.

3.5 | Relationship between time spent using SM
and DE

Additional usage analyses were completed for Instagram and Snapchat

only given that these account types were reported far more commonly

TABLE 3 Disordered eating cognitions by social media account type and disordered eating cognitions and behaviors by total social media
accounts for girls and boys

No
Girls
Yes

Group difference
No

Boys
Yes Group difference

M (SD) M (SD) OR (95% CI) M (SD) M (SD) OR (95% CI)

Global EDE-Q

Facebook 1.50 (1.30) 1.77 (1.26) 1.21 (0.94–1.56) 0.91 (0.92) 1.35 (1.25) 1.40 (1.28–1.82)

Instagram 1.43 (1.28) 1.58 (1.31) 1.12 (0.93–1.36) 0.86 (0.80) 1.09 (1.12) 1.28 (1.04–1.58)

Snapchat 1.29 (1.23) 1.70 (1.33) 1.39 (1.15–1.68) 0.89 (0.88) 1.11 (1.14) 1.24 (1.02–1.50)

Tumblr 1.49 (1.27) 2.02 (1.48) 1.43 (1.08–1.90) 1.00 (1.02) 1.57 (1.15) 1.53 (1.02–2.27)

Total accounts 0
n = 123

1
n = 99

2
n = 213

3–4
n = 67

Sig group contrasts 0
n = 129

1
n = 90

2
n = 123

3–4
n = 87

Sig group contrasts

M
(SD)

M
(SD)

M
(SD)

M
(SD)

ES (d) M
(SD)

M
(SD)

M
(SD)

M
(SD)

ES (d)

Global EDE-Q 1.22
(1.15)

1.49
(1.38)

1.65
(1.31)

1.80
(1.31)

0 < 2 (−0.35)
0 < 3–4 (−0.49)

0.84
(0.78)

0.93
(0.97)

0.99
(1.03)

1.40
(1.29)

0 < 3–4 (−0.76)
1 < 3–4 (−0.58)
2 < 3–4 (−0.41)

N
(%)

N
(%)

N
(%)

N
(%)

OR (95% CI) N
(%)

N
(%)

N
(%)

N
(%)

OR (95% CI)

DE behavior 50
(40.3)

46
(46.9)

120
(56.3)

49
(73.1)

0 < 2
3.20 (1.45–7.14)
0 < 3–4
3.14 (1.21–8.11)

51
(39.5)

40
(44.9)

55
(44.7)

42 (48.3) 0 < 3–4
5.00 (1.56–16.07)

Abbreviations: CI, confidence intervals; EDE-Q, Eating Disorder Examination–Questionnaire; ES, effect size for significant between-group contrasts
between number of social media accounts (p < .05); M, mean; OR, odds ratios for logistic regressions with a significant difference between girls and boys;

SD, standard deviation.

WILKSCH ET AL. 101

than Facebook or Tumblr. Linear …

error: Content is protected !!