Screening High School Students for Eating

Disorders: Results of a National Initiative
ORIGINAL RESEARCH
Suggested citation for this article: Austin SB, Ziyadeh NJ, Forman S, Prokop LA, Keliher A, Jacobs D. Screening high school students for eating disorders: results of a national initiative. Prev Chronic Dis 2008;5(4). http://www.cdc.gov/pcd/issues/2008/oct/07_0164.htm. Accessed [date].
PEER REVIEWED
Abstract
Introduction
Early identification and treatment of disordered eating and weight control behaviors may prevent progression and reduce the risk of chronic health consequences.
Methods
The National Eating Disorders Screening Program coordinated the first-ever nationwide eating disorders screening initiative for high schools in the United States in 2000. Students completed a self-report screening questionnaire that included the Eating Attitudes Test (EAT-26) and items on vomiting or exercising to control weight, binge eating, and history of treatment for eating disorders. Multivariate regression analyses examined sex and racial/ethnic differences.
Results
Almost 15% of girls and 4% of boys scored at or above the threshold of 20 on the EAT-26, which indicated a possible eating disorder. Among girls, we observed few significant differences between ethnic groups in eating disorder symptoms, whereas among boys, more African American, American Indian, Asian/Pacific Islander, and Latino boys reported symptoms than did white boys. Overall, 25% of girls and 11% of boys reported disordered eating and weight control symptoms severe enough to warrant clinical evaluation. Of these symptomatic students, few reported that they had ever received treatment.
Conclusion
Population screening for eating disorders in high schools may identify at-risk students who would benefit from early intervention, which could prevent acute and long-term complications of disordered eating and weight control behaviors.
Introduction
The acute and chronic medical and psychiatric consequences of eating disorders are well documented. Anorexia and bulimia nervosa are associated with comorbid medical conditions such as osteoporosis and complications of the gastrointestinal, cardiovascular, and endocrine systems (1-3). Binge eating disorder has been linked with psychiatric comorbidity and severe obesity (4). Compared with the general population, people with anorexia or bulimia nervosa are at increased risk of suicide (5,6). More prevalent forms of disordered weight control behaviors, such as vomiting and abuse of laxatives, are also associated with a range of negative health outcomes, such as esophagitis, gastric rupture, and impairment of digestive functioning (7-9). Additionally, disordered eating behaviors may be causally related to overweight and obesity (10-12).
In the United States, the lifetime prevalence of anorexia nervosa, bulimia nervosa, and binge eating disorder are estimated to be 0.9%, 1.5%, and 3.5%, respectively, among women and 0.3%, 0.5%, and 2.0%, respectively, among men (4). Men may make up 10%-25% of the population with anorexia nervosa or bulimia nervosa (4,13) and
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services, the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
www.cdc.gov/pcd/issues/2008/oct/07_0164.htm • Centers for Disease Control and Prevention
S. Bryn Austin, ScD, Najat J. Ziyadeh, MPH, Sara Forman, MD, Lisa A. Prokop, BA, Anne Keliher, MMHS, Douglas Jacobs, MD
VOLUME 5: NO. 4
OCTOBER 2008
Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2008/oct/07_0164.htm
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only
and does not imply endorsement by any of the groups named above.
nearly half of cases of binge eating (4). Median age of onset
for the 3 disorders is estimated to be 18-21 years (4).
Disordered weight control behaviors and symptoms that
do not necessarily meet psychiatric criteria for an eating
disorder diagnosis (7) are estimated to be as much as 20
times more common in community samples (14) than
are those behaviors and symptoms that meet diagnostic
criteria. In 2005 the Youth Risk Behavior Surveillance
System (YRBSS) found that 6.2% of girls and 2.8% of boys
reported vomiting or taking laxatives in the past month
to lose or maintain weight (15). The Minnesota Student
Survey of more than 81,000 high school students found
that, among girls, in the past year, 8.8% vomited to control
their weight, 1.9% used laxatives for weight control, and
25.6% reported binge eating; comparable estimates from
this study for the 3 behaviors in boys were 1.6%, 1.7%,
and 12.5%, respectively (16). The proportion of high school
youth who report these behaviors that have been treated
for their eating disorder symptoms is unknown.
Some studies have found a higher prevalence of disordered
eating behaviors and attitudes among white girls
than among girls of color, particularly African American
girls, although others have reported varying results (16-
23). The 2005 YRBSS documented a similar proportion of
white and Latina high school girls who reported vomiting
or using laxatives in the past month to control weight
(6.7% and 6.8%, respectively); these behaviors were least
commonly reported by African American girls (4%) (15).
In the Minnesota Student Survey, compared with white
girls, Latina and Asian girls, but not African American or
American Indian girls, reported higher rates of disordered
eating behaviors (16). In another school-based study, however,
vomiting and laxative use to control weight were more
common in African American than in white girls (24).
In research with boys, findings have been more consistent
in documenting equal or higher risk in boys of color
relative to white boys. In the 2005 YRBSS, Latino boys
reported the highest rate (3.9%) of vomiting and laxative
abuse in the past month, while a similar proportion of
white (2.3%) and African American (2.8%) boys reported
these disordered weight control behaviors in the past
month (15). In the Minnesota Student Survey, compared
with white boys, Latino, Asian, and American Indian boys,
but not African American boys, reported higher rates of
disordered eating and weight control behaviors (16). The
Commonwealth Fund survey of more than 6,700 US youth
in grades 5 through 12 found that African American and
Latino boys reported higher rates of ever having binged
and purged than did white boys (21).
Early identification and intervention for a range of mental
health problems may reduce risk of progression of the
illness, relapse, and comorbid conditions (25). A shorter
period between symptom onset and start of treatment
may improve prognosis for recovery from anorexia (26)
and bulimia nervosa (27). Early detection through schoolbased
screening can shorten the period between symptom
onset and accessing care and help adolescents begin treatment
at younger ages. Support staff in schools may be
ideally situated to help identify at-risk youth of both sexes
and all races/ethnicities, make referrals for clinical evaluation
and treatment, and offer in-school support (28).
Working with staff in schools across the country in the
winter of 2000, the National Eating Disorders Screening
Program (NEDSP) coordinated, to our knowledge, the
first-ever nationwide eating disorders screening initiative
for high schools in the United States. The program was
designed to promote early detection and treatment-seeking
in adolescents with untreated eating disorder symptoms.
NEDSP’s parent organization, the national nonprofit
organization Screening for Mental Health (http://www.
mentalhealthscreening.org), has coordinated a number
of national, broad-scale screening initiatives in schools,
workplaces, communities, and the military that address
depression, bipolar disorder, anxiety disorders, and alcohol
abuse; it has also screened for eating disorders on
college campuses (29). For the present analysis, our aims
were to evaluate the screening program’s ability to reach
symptomatic youth who had not yet accessed treatment
and to examine sex and racial/ethnic differences in symptoms
and treatment history.
Methods
The NEDSP Program
NEDSP staff sent out registration information about
the program by direct mail and e-mail to individual
membership lists of national professional organizations
for school psychologists, nurses, and counselors to invite
high schools across the country to enroll in the program.
Representatives from high schools then contacted NEDSP
to enroll. All participating high schools were provided with
VOLUME 5: NO. 4
OCTOBER 2008
www.cdc.gov/pcd/issues/2008/oct/07_0164.htm • Centers for Disease Control and Prevention
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only
and does not imply endorsement by any of the groups named above.
a questionnaire to screen for student eating disorders; educational
materials for use in classrooms or assemblies; and
technical assistance to help staff implement the screening,
handle student requests to discuss eating disorders, and
make appropriate referrals for evaluation and treatment.
NEDSP educational materials included a video and discussion
guide, participatory classroom curriculum, and
activity guide. All materials were designed to help motivate
students to seek help with eating disorder symptoms.
Care was taken to design materials that did not glamorize
eating disorders or provide unnecessary details about
disordered weight control methods. Educational content
addressed healthy diet and activity, signs and symptoms
to watch out for in friends and family, availability and efficacy
of treatment, and the need to seek help for symptoms.
In addition, materials offered students guidance on how
to talk with a friend or family member who may have an
eating disorder (30).
Screening questionnaire
High schools administered the anonymous, self-report
eating disorders screening questionnaire to students in
classrooms and assemblies. The survey included the
Eating Attitudes Test (EAT-26), a validated eating disorders
screening instrument (31). Possible scores on the
EAT-26 range from 0 to 78. A score of 20 or above indicates
that a person may have an eating disorder and should be
evaluated further by a mental health professional. The
student screening questionnaire also included items that
assessed how often in the past 3 months students had
vomited to control their weight, engaged in eating binges,
or exercised to lose or control their weight. Each of these
behavioral questions was followed by 7 response options:
never, less than once per month, 1-3 times per month, once
per week, 2-6 times per week, once per day, and more than
once per day. The item on vomiting was adapted from the
YRBSS (32). The questionnaire included an item on past
treatment for eating disorders and items on sex, age, race/
ethnicity, height, and weight.
Participants and sampling procedure
A total of 270 public, private, and parochial high schools
signed up to participate in the screening program, and
152 schools from 34 states completed the screening and
educational components of NEDSP. Ninety-eight schools
returned more than 35,000 student screening forms for
analysis. Because of cost constraints on data entry, a subset
of student screening forms were randomly selected for
analysis by using a 2-stage, clustered-sampling method.
First, 33 schools were randomly sampled from the 98 that
returned screening forms, then a random sample of forms
was selected from these schools; the number of forms
selected from a school was proportional to the number
received from that school. Because of a change in protocol
at the data entry site, 8 of the 33 schools had all of their
surveys entered rather than a proportional random sample;
therefore, weighting was used in analyses to adjust for
the oversampling of student surveys from these 8 schools.
This 2-stage selection procedure resulted in a sample of
5,740 screening forms.
Variables and data analysis
Total EAT-26 scores were computed by adding individual
item scores. For students who were missing 1 or 2 items
on the EAT-26 (n = 272), total scores were scaled to values
within the full possible range of the instrument. Students
who were missing 3 or more items were excluded from
analysis. A binary term for EAT-26 score was created on
the basis of the recommended cutoff of a score of 20 as an
indication of a possible eating disorder. Binary terms were
created for each of the items on disordered behavior in the
past 3 months: any report of vomiting to control weight,
binge eating once a week or more, and exercising to lose or
control weight more often than once per day.
Multivariate linear and logistic regression models were
used to test sex and racial/ethnic group differences in
mean EAT-26 scores and frequencies of reporting disordered
eating and weight control behaviors and ever having
been treated for an eating disorder. Multivariate models
examining sex differences controlled for age and race/ethnicity,
and models examining racial/ethnic group differences
controlled for age and were stratified by sex.
In secondary analyses to explore whether symptom type
and severity may explain sex differences in treatment
history, we tested 4 multivariate models that estimated
the odds of ever having been treated for an eating disorder,
comparing girls with boys within each symptom-type
subgroup of students, controlling for symptom severity.
Thirty-nine students who did not respond to the eating
disorders treatment history item were excluded. All models
controlled for sex, age, race/ethnicity, and extreme thinness,
which was included because it is a widely recognized
sign and symptom of eating disorders (33) and is readily
VOLUME 5: NO. 4
OCTOBER 2008
Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2008/oct/07_0164.htm
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only
and does not imply endorsement by any of the groups named above.
observable. Extreme thinness was classified according to
the World Health Organization definition of grade 1 thinness
as a body mass index (BMI) less than 18.5 kg/m2 in
adults aged 18 years or older (34), then coded by using
age- and sex-specific BMI values for ages younger than 18
years to correspond with the adult cutoff (35). Subsample
restrictions and additional covariates included in each
model were as follows: model 1, restricted to the subgroup
of students (n = 518) with an EAT-26 score of 20 or higher,
controlled for total EAT-26 score; model 2, restricted to
the subgroup of students (n = 435) who reported vomiting
to control their weight in the past 3 months, controlled for
vomiting frequency; model 3, restricted to the subgroup
of students (n = 366) who reported binge eating once per
week or more in the past 3 months, controlled for binge
frequency; model 4, restricted to the subgroup of students
(n = 155) who reported exercising more than once per day
to lose or control weight in the past 3 months, which is the
highest severity level assessed for this item.
For all models, generalized estimating equation methods
were used to account for the clustered study design
by using SAS PROC GENMOD (SAS Institute Inc, Cary,
North Carolina) (36). After 173 surveys were excluded
because of missing data, the analytic sample included
5,567 students. Compared with students included in analyses,
those excluded were less likely to describe themselves
as white (P = .002) and more likely to have not reported
a race/ethnicity (P < .001); we found no differences in age, sex, EAT-26 score, disordered eating or weight control behaviors, or past eating disorder treatment (P > .05).
Analysis of NEDSP data was approved by the institutional
review board at Children’s Hospital Boston.
Results
The sample included 58% (3,252) girls and 42% (2,315)
boys; 3% (189) were African American, 2% (93) American
Indian, 2% (134) Asian/Pacific Islander, 5% (303) Latino,
83% (4,629) white, and 4% (219) reported no ethnicity. The
mean age was 15.9 (standard deviation 1.0) years. Girls
were 3 to 5 times more likely than boys to score at or above
the threshold on the EAT-26, to report vomiting to control
their weight in the past 3 months, and to have ever been
treated for an eating disorder (Table 1).
Among girls, few significant differences were found in
eating disorder symptoms across racial/ethnic groups
(Tables 2A and 2B). Compared with white girls, Latina
girls were less likely and American Indian girls were more
likely to score 20 or more on the EAT-26, and African
American and American Indian girls were more likely
to report exercising more than once per day to control
their weight. In contrast, among boys, African American,
American Indian, Asian/Pacific Islander, and Latino boys
were consistently more symptomatic than were white boys
across the range of disordered eating and weight control
symptoms and behaviors.
Within symptom subgroups defined by EAT-26 score
and binge eating, girls were roughly 3.5 times more likely
to report that they had been treated for an eating disorder
than were boys with comparable symptom severity (Table
3). Within the symptom subgroup defined by exercising
once a day or more often to control weight, girls were
almost 8 times more likely than boys to report having
been treated for an eating disorder. In contrast, within the
subgroup that reported vomiting, no sex difference was
observed in the odds of having ever received treatment. In
most models, extreme thinness was positively associated
with the odds of having been treated for an eating disorder,
but age and race/ethnicity were not, controlling for sex
and symptom type and severity (data not shown).
Discussion
NEDSP, to our knowledge the first national screening
program for eating disorders held in high schools across
the United States, found that almost 1 in 4 girls and 1
in 10 boys reported at least 1 disordered eating or weight
control symptom serious enough to warrant further evaluation
by a health professional. Applying these findings
to the roughly 35,000 students who completed screening
questionnaires, we estimate that close to 7,000 students
with potential eating disorder symptoms were identified
in participating schools. Furthermore, a large proportion
of symptomatic students had never been treated for an
eating disorder. Depending on the symptom type, the proportion
of symptomatic students who had never received
treatment was 83% to 95% of boys and 83% to 86% of girls.
These results support 2 conclusions: 1) national screening
for eating disorders in high schools reached a large number
of students who were likely to have symptoms of disordered
eating and weight control and 2) most symptomatic
high school students were untreated. Coupled with evidence
that early detection and intervention may improve
VOLUME 5: NO. 4
OCTOBER 2008
www.cdc.gov/pcd/issues/2008/oct/07_0164.htm • Centers for Disease Control and Prevention
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only
and does not imply endorsement by any of the groups named above.
treatment outcomes (26,37,38), these findings underscore
the suitability of population screening (39) in high schools
as a strategy to identify youth in need of clinical evaluation
for eating disorders.
Findings relating to racial/ethnic group patterns are
also informative. Among girls, we observed few differences
in eating disorder symptoms across ethnic groups,
a finding consistent with the national YRBSS (15) and
Commonwealth Fund survey (21). Our findings differ
from those reported in a meta-analysis of 35 studies,
which included predominantly female participants aged 9
to 73 years; results of the meta-analysis document higher
risk in white than in African American girls and women
(22). That said, a second meta-analysis suggests that differences
between African American and white girls and
women may manifest for some symptom types (drive for
thinness) but not for others (bulimia and binge eating)
(23). Among boys, in almost every symptom category,
whites reported lower rates than did any other ethnic
group. These findings are consistent with results of the
national YRBSS (15) and Commonwealth Fund survey
(21) and the Minnesota Student Survey (16). Our study
expands on the literature on adolescent boys by providing
racial/ethnic group comparisons of eating disorder treatment
history and of symptom severity with a validated
eating disorder screening tool. NEDSP results indicate
that school-based screening for eating disorders is appropriate
and needed, both in schools that are racially/ethnically
diverse and in those that are not diverse. In
fact, these findings suggest that distributing screening
resources differentially by sex or race/ethnicity would be
inappropriate and possibly unethical; boys and girls of all
races/ethnicities should be targeted.
Girls who participated in NEDSP were more than 4
times more likely than boys to have been treated for an
eating disorder. Boys have a lower prevalence of disordered
eating and weight control symptoms than do girls, and this
difference may partly account for our finding. However,
results of subgroup analyses of girls and boys with similar
symptoms, controlling for symptom severity, age, race/
ethnicity, and extreme thinness, suggest that differences
in prevalence alone may not explain sex disparities in
treatment history. As shown in Table 3, symptomatic boys
were far less likely than girls to have accessed treatment.
Our findings may overestimate the sex disparity in treatment
history, possibly because of residual confounding
as a result of incomplete control for sex differences in
symptom severity. Alternatively, symptomatic boys may
be overlooked by clinicians, school personnel, parents, and
others because the prevalence of eating disorders in boys
is commonly underestimated (40).
Limitations
This study has several limitations. The sampling methods
used to select screening forms for inclusion in the analysis
allow our results to be generalizable to the larger pool
of more than 35,000 screening forms received at NEDSP
headquarters. However, students who participated in
the screening program may not be representative of high
school students as a whole in the United States. Schools
that enrolled in the program and returned screening forms
to NEDSP headquarters may have had greater resources,
or their staff may have been more concerned about eating
disorders than the staff of other schools. Despite these
limits in generalizability, our estimate of the prevalence of
eating disorder symptoms is comparable to that of a similarly
designed screening study conducted with adolescent
girls and young women aged 12 to 21 years who received
routine care from a US military health care facility. In
that sample, 21% of girls scored 20 or more on the EAT-26
or reported disordered eating or weight control behaviors
(41); in NEDSP, 24.8% of girls met these criteria.
When controlling for symptom type and severity, we did
not find race/ethnicity to be associated with the odds of
ever having been treated for an eating disorder; however,
because subgroup sizes were small, we may not have
had sufficient power to detect disparities in treatment
history. Prior research suggests that stereotypes may
cause eating disorder symptoms to be underrecognized in
African American and Latina girls (42). Furthermore, in
a national eating disorder screening study carried out on
college campuses in the United States, among those with
eating disorder symptoms, nonwhite participants were
less likely than white participants with comparable symptom
severity to be asked about their symptoms by doctors
and mental health professionals and less likely to receive a
recommendation for further clinical evaluation (29).
The frequency and severity thresholds for each of the
disordered eating and weight control symptoms were
chosen to be clinically meaningful as a screening tool
and were not designed to be diagnostic thresholds. Some
students may have been incorrectly identified by the
screening tool as having an eating disorder, and some
VOLUME 5: NO. 4
OCTOBER 2008
Centers for Disease Control and Prevention • www.cdc.gov/pcd/issues/2008/oct/07_0164.htm
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the US Department of Health and Human Services,
the Public Health Service, the Centers for Disease Control and Prevention, or the authors’ affiliated institutions. Use of trade names is for identification only
and does not imply endorsement by any of the groups named above.
may not have needed treatment. Despite the limitations
of the thresholds selected for NEDSP, evidence suggests
these cutoffs are meaningful. With a cutoff of 20 or higher,
the sensitivity and specificity of EAT-26 are moderately
high to high for detecting eating disorder cases that meet
Diagnostic and Statistical Manual of Mental Disorders,
Fourth Edition criteria, and when used as a dimensional
measure, EAT-26 score is positively associated with severity
of symptoms in women with subsyndromal eating
disorders (43). In addition, a study in a nonpsychiatric
sample of adults recommends a threshold of 11 on the
EAT-26 to detect subsyndromal and EDNOS (eating disorder
not otherwise specified) cases (44). In other work with
the NEDSP sample, students who reported the disordered
eating or weight control behaviors scored on average in the
subthreshold (a score of 10 to 19) to threshold (a score of
20) range on the EAT-26; mean EAT-26 scores increased
fairly linearly with increasing behavioral frequency for
vomiting and binge eating (J. Haines, written communication).
In addition, in female NEDSP participants, vomiting
for weight control in the previous 3 months, even when
infrequent, was associated with disruption of regular menstrual
cycles (45).
Our analyses were based on self-report data, which are
subject to bias resulting from cognitive and situational
factors (46). Nevertheless, a validation study that used
self-report to assess vomiting and laxative use for weight
control in adolescent girls found high sensitivity (0.93)
and specificity (0.86), although comparable estimates for
a measure of binge eating were lower (sensitivity 0.53,
specificity 0.78) (47).
Conclusions
In June 2007, the US Senate directed the Centers for
Disease Control and Prevention to intensify efforts to
investigate the problem of eating disorders and their
health implications for the US population (48). On the
basis of results from the NEDSP screening initiative,
we have identified a need for population screening and
public health intervention in US high schools, since only
a small fraction of students who self-identified as engaging
in disordered eating and weight control behaviors had
ever received treatment. For many of these adolescents,
beginning treatment during high school or earlier would
improve treatment effectiveness and mitigate acute and
chronic complications of disordered eating and weight
control behaviors, such as impaired growth and digestive
functioning, osteoporosis, and obesity (4,26,37,38). NEDSP
is a useful addition to the screening tools available to public
health and school health practitioners to address this
critical public health problem.
Acknowledgments
NEDS was funded by the McKnight Foundation. Dr
Austin and Dr Forman were supported by the Leadership
Education in Adolescent Health Project grant T71 MC
00009-16 from the Maternal and Child Health Bureau,
Health Resources and Services Administration, Department
of Health and Human Services. We acknowledge Jillian
Barber, Nancy Conlon, James DiCanzio, Cheryl D’Souza,
S. Jean Emans, Linda Garcia, Barbara Kopans, Joelle
Riezes, David Wypij, Anne Zachary, the Massachusetts
Eating Disorder Association, and the NEDSP advisory
board for their contributions. We also thank the thousands
of students, faculty, and staff from high schools across the
country who made the screening program possible.
Author Information
Corresponding Author: S. Bryn Austin, ScD, Division of
Adolescent Medicine, Children’s Hospital, 300 Longwood
Ave, Boston, MA 02115. Telephone: 617-355-8194. Email:
bryn.austin@childrens.harvard.edu. Dr Austin is
also affiliated with the Department of Society, Human
Development, and Health, Harvard School of Public
Health, Boston, Massachusetts.
Author Affiliations: Najat J. Ziyadeh, Sara Forman, Lisa
A. Prokop, Division of Adolescent Medicine, Children’s
Hospital, Boston, Massachusetts; Anne Keliher, Douglas
Jacobs, National Eating Disorders Screening Program,
Screening for Mental Health, Wellesley, Massachusetts.