7
JOURNAL OF ADOLESCENT HEALTH CARE 1987;8:266-272 Use of the Eating Attitudes Test an Disorder Inventory in Adolescents RICKEY L. WILLIAM$ M.D. The Eat@ Attitudes Test and the Eating DieorderIn- ventorywere developed to measureabnormal eating at- titudes mnd behavior6in patients with eating disorders. Thie report describes each of these self-reportquestion- naires and outlines their functions, limitations, and pre- vioue use with adolescent subjects. KRY WORDS: Eating Attitudes Test Eating Disorder Inventory Eating disorders Anorexia newosa Bulimia Appetite disorders The eliagnch of an eating disorder, including an- orexia nervosa and bulimia,is being made more fre- quently, especially in tenage and young adult women (l-7). Physicians who care for adolescents to evaluatepabents with a suspected . In addition to the traditional history and physical examination, some physicians also use self-report questionnaires such as the Eating Atti- tudes Test and the Eating Disorder Inventory as part of their assessment. Patients with eating disorders have abnormal eat- ing attitudes and behaviors. For example, patients with anorexia nervosa still exhibit an intense desire tolose weight even when they are emaciated. The patient may follow a ritualistic diet that avoids car- bohydrates and red meats, eat the same foods day ham fk iIk!prtnunt of Pediufrics, University of Arizona Htwlth Center, &st.m, Atina. ms reprint requests to: Rickey L. Wilhms, M.D.. Department l!iOl North cmnpbelt Avenue, Tucson, AZ 85724. ubpt ucvxptedApril 29, 2986. 266 after day, avoid eating with others, and be terrified that they will be unable to control their food intake once they start eating. Patients with bulimia fre- quently binge on high-calorie, easily digested food, then engage in purging behaviors with self-induced vomiting. The Eating Attitudes Test and Eating Dis- order Inventory were developed to help measure some of these attitudes and behaviors. The Eating Attitudes Test The Eating Attitudes Test (EAT) was developed by Garner and GarEnkel(8). It is a IO-item self-admin- istered questionnaire that originally was adminis- tered to patients with anorexia nervosa. Each of the 40 forced-choice Likert-scale items can be answered “always,” “very often,” “often,” “sometimes,” “rarely,” or “never.” A score of 3 is earned for an extreme response in the “anorexic” direction, with adjacent alternatives given a score of 2 or 1. For example (Table l), subject A receives a score of 3 on item 1, 2 on item 2, 1 on item 3, and 0 on items 4, 5, and 6; and subject B receives a score of 3 on item 6, 2 on item 5, 1 on item 4, and 0 on items 3, 2, and 1. The sum of the scores on all 40 items gives the total score for the EAT. Gamer and Garfinkel (8) applied the EAT to 65 female anorexia nervosa patients in various stages of therapy and 93 control subjects. The mean age of the subjects was 22 years. The mean total score on the EAT was 58.9 for patients with anorexia nervosa and 15.6 for normal controls. When a total score of 32 was used as a minimum cut-off point for anorexia nervosa, all patients with anorexia and only 7% of the normal controls scored 1 32. As patients with anorexia nervosa improved clinically, their total scores decreased into the normal range. Internal re- 0 Society for Adolescent Medicine, 1987 Pubfished by Elsevier Science Publishing Co., Inc., 52 Vanderbilt Ave., New York, NY 10017 0191-/$3.50

Use of the eating attitudes test and eating disorder inventory in adolescents

Embed Size (px)

Citation preview

Page 1: Use of the eating attitudes test and eating disorder inventory in adolescents

JOURNAL OF ADOLESCENT HEALTH CARE 1987;8:266-272

Use of the Eating Attitudes Test an Disorder Inventory in Adolescents

RICKEY L. WILLIAM$ M.D.

The Eat@ Attitudes Test and the Eating Dieorder In- ventory were developed to measure abnormal eating at- titudes mnd behavior6 in patients with eating disorders. Thie report describes each of these self-report question- naires and outlines their functions, limitations, and pre- vioue use with adolescent subjects.

KRY WORDS:

Eating Attitudes Test Eating Disorder Inventory Eating disorders Anorexia newosa Bulimia Appetite disorders

The eliagnch of an eating disorder, including an- orexia nervosa and bulimia,is being made more fre- quently, especially in tenage and young adult women (l-7). Physicians who care for adolescents

to evaluate pabents with a suspected . In addition to the traditional history

and physical examination, some physicians also use self-report questionnaires such as the Eating Atti- tudes Test and the Eating Disorder Inventory as part of their assessment.

Patients with eating disorders have abnormal eat- ing attitudes and behaviors. For example, patients with anorexia nervosa still exhibit an intense desire tolose weight even when they are emaciated. The patient may follow a ritualistic diet that avoids car- bohydrates and red meats, eat the same foods day

ham fk iIk!prtnunt of Pediufrics, University of Arizona Htwlth Center, &st.m, Atina.

ms reprint requests to: Rickey L. Wilhms, M.D.. Department l!iOl North cmnpbelt Avenue, Tucson, AZ 85724.

ubpt ucvxpted April 29, 2986.

266

after day, avoid eating with others, and be terrified that they will be unable to control their food intake once they start eating. Patients with bulimia fre- quently binge on high-calorie, easily digested food, then engage in purging behaviors with self-induced vomiting. The Eating Attitudes Test and Eating Dis- order Inventory were developed to help measure some of these attitudes and behaviors.

The Eating Attitudes Test The Eating Attitudes Test (EAT) was developed by Garner and GarEnkel(8). It is a IO-item self-admin- istered questionnaire that originally was adminis- tered to patients with anorexia nervosa. Each of the 40 forced-choice Likert-scale items can be answered “always,” “very often,” “often,” “sometimes,” “rarely,” or “never.” A score of 3 is earned for an extreme response in the “anorexic” direction, with adjacent alternatives given a score of 2 or 1. For example (Table l), subject A receives a score of 3 on item 1, 2 on item 2, 1 on item 3, and 0 on items 4, 5, and 6; and subject B receives a score of 3 on item 6, 2 on item 5, 1 on item 4, and 0 on items 3, 2, and 1. The sum of the scores on all 40 items gives the total score for the EAT.

Gamer and Garfinkel (8) applied the EAT to 65 female anorexia nervosa patients in various stages of therapy and 93 control subjects. The mean age of the subjects was 22 years. The mean total score on the EAT was 58.9 for patients with anorexia nervosa and 15.6 for normal controls. When a total score of 32 was used as a minimum cut-off point for anorexia nervosa, all patients with anorexia and only 7% of the normal controls scored 1 32. As patients with anorexia nervosa improved clinically, their total scores decreased into the normal range. Internal re-

0 Society for Adolescent Medicine, 1987 Pubfished by Elsevier Science Publishing Co., Inc., 52 Vanderbilt Ave., New York, NY 10017 0191-/$3.50

Page 2: Use of the eating attitudes test and eating disorder inventory in adolescents

May 1987 EATING-DISORDER QUESTIONNAIRES 267

Table 1. Selected Items from the Eating Attitudes Test, with Responses by Two Hypothetical Subjects, A and B

EAT item

1. Engage in dieting behavior

2. Eat diet foods

3. Feel uncomfortable after eating sweets

4. Have the impulse to vomit after meals

5. Vomit after I have eaten

6. Have gone on eating binges where I feel that I may not be able to stop

Response

Always Very often Often Sometimes Rarely Never

A B

A B

A B

B A

B A

B A

liability (Cronbach’s alpha) was reported as 0.79 for the anorexia nervosa subjects and 0.94 for the pooled sample of anorexics and normal controls.

In a subsequent study using the EAT as a screen- ing device to determine whether individuals whose profession focuses increased attention on a slim body shape are at risk for anorexia nervosa, a normal control group compared to ballet and fashion mod- eling students had significantly iower mean total scores (8). The EAT helped to identify 7% of the ballet and fashion modeling students who were sub- sequently diagnosed by clinical interview as having anorexia nervosa.

Other investigators have used the EAT to screen for eating disorders. Button and Whitehouse (10) administered the test to 578 students (446 women and 132 men) at a technical college. The 28 students with total scores 2 32 were invited for an interview, as were 28 randomly selected students with scores < 32. The interviews were directed at eliciting de- tailed information of clinical relevance to anorexia nervosa. Although no cases of anorexia nervosa were found, high-scoring students tended to report a lower minimum weight (lowest since puberty) than did the low-scoring students. In addition, 39% of the high-scoring students but none of the low- scoring students reported self-induced vomiting.

Szmukler (3) found that 65 (5%) of 1331 South London upper-class schoolgirls 14-18 years of age scored L 30 on the EAT. Subsequent clinical inter- views disclosed that of those 65, 8 individuals had anorexia nervosa, 5 had bulimia, 25 had a partial syndrome of anorexia nervosa (that is, they showed most of the typical psychopathological features of anorexia nervosa), and 27 were normal dieters. Three cases of anorexia nervosa were not detected: one girl with known anorexia nervosa was excused

from the survey, one girl did not hand in her ques- tionnaire, and one girl who was later referred to the author for treatment of anorexia nervosa admitted to lying on the questionnaire. In addition, some of the girls who scored < 30 were considered to show a partial syndrome of anorexij nervosa. There was no correlation between the EAT scores and age.

Lowe and coworkers (11) administered the test to screen for eating disorders in 1514 New Zealand school girls (mean age 15 years). Fourteen percent scored > 30.

Are all 40 items on the EAT necessary? To answer this question, Garner and coworkers (12) used factor analysis to show that 14 of the 40 original items on the EAT could be deleted and still maintain a high correlation (I’ = 0.98) with the original test. Stan- dardized reliability coefficients (Cronbach’s alpha) for the EAT-26 total score were reported as 0.90 and 0.83, respectively, for the anorexia nervosa and fe- male comparison groups.

Mann and coworkers (13) administered the 26- item EAT version to 262 15-year-old South London schoolgirls and found that 6.9% scored in the an- orexic range of B 21. Compared with their peers, the girls who scored in the anorexic range reported missing more meals during the day and having more weight instability. The authors concluded that the 26-item EAT is an efficient screening instrument for abnormal eating attitudes and behaviors.

Because two individuals can achieve the same to- tal EAT score by answering every item differently (Table l), the patterns of relationships between items have been assessed to determine whether items can be grouped together in a small number of subscales representing an underlying construct. For example (Table l), items 1,2, and 3 could represent a dieting subscale, and items 4, 5, and 6 a bulimia

Page 3: Use of the eating attitudes test and eating disorder inventory in adolescents

JOURNAL OF ADOLESCENT HEALTH CARE Vol. 8, No. 3 26s WILLIAMS

subscale. Garner and coworkers (12) used factor analysis to define three EAT sobscales related to di- eting, bulimia and food preoccupation, and oral con- trol, They found that although the bulimic and restricter subsamples of their anorexia nervosa pop- ulation did not differ in total test scores, there were significant differences in subscale scores between the two groups. Standardized reliability coefficients, for the three subscale scores ranged from 0.83 to 0.90 for their anorexia nervosa group and from 0.46 to 0.86 for their female comparison group.

Subsequent investigators using adolescent sub- jects have confirmed that these subscales are useful in these patient groupings. Wells and coworkers (14) administered the 40sitem EAT to 749 teenage New Zealand schoolgirls. The mean EAT-40 score was 12.1, with 4.5% of the girls scoring Z 30, The mean EAT-26 score was 66. Factor analysis revealed that their Factor 1 contained 12 and 13 items of Garner and coworkers’ dieting factor, with the other sub- scales also being similar to those reported by Garner et al. (12)

Williams and coworkers (1 S) administered the 40- item EAT to 72 Tucson, Arizona, junior and senior high school women (mean age 15 years). The mean EAT-40 score for subjects classified as normal was 17.6, and the mean EAT-26 score was 10.7. F‘ictor analysis revealed results essentially identical tc the subscales of Garner and coworkers (12) (Table 2).

In an attempt to evaluate the utility of the EAT alone as a screening device for the detection of an- orexia nervosa and bulimia in a female college pop- ulation, Garter and Moss (16) administered the EAT- 40 to 162 white females (mean age 19 years) and found that 21.6% of the subjects scored B 30. Only

) of56 selected subjects who underwent a struc-

tured interview could be classified as anorexic, and only 4 (7%) of the 56 subjects could be classified as bulimic. The authors reported the test-retest reli- ability coefficient of the EAT as 0.84.

The Eating Disorder inventory The Eating Disorder Inventory (EDI), developed by Garner and coworkers (17), is a 64-item, self-report, multiscale measure designed to assess the psycho- logic and behavioral traits common to anorexia ner- vosa and bulimia. It consists of eight subscales: 1) Drive for Thinness, 2) Bulimia, 3) Body Dissatisfac- tion, 4) Ineffectiveness, 5) Perfectionism, 6) Inter- personal Distrust, 7) Interoceptive Awareness, and 8) Maturity Fears. These subscales were derived from a large pool of items generated by clinicians familiar with the research and treatment of patients with anorexia nervosa. To be included in a subscale, an item had to be answered significantly differently by control patients compared to those with anorexia nervosa and had to be more highly correlated to the subscale to which it was intended to belong than to any other subscale. As a measure of internal con- sistency, the authors required coefficients of reli- ability (standardized Cronbach’s alpha) for each subscale to be above 0.80 for the anorexia nervosa samples.

The format of the ED1 is similar to the EAT, Scor- ing is the same. In contrast to the EAT, the ED1 focuses more on the specific cognitive and behav- ioral dimensions to help differentiate patient subgroups, or those with serious psychopathokgy, from “extreme dieters.”

Garner and coworkers applied the ED1 to subjects with “restricter” or “bulimia” subtypes of anorexia

Fable 2. Eating Attitudes Test Scores Reported in Previous Studies

Bulimia and Qral

Investi@ors Mean age Dieting Food Prcoccu- Control

Population II (yr) EAT-40 EAT-26 subscale pation subscale subscale

Garner and Carfmkel Anorexia nervosa 33 22 - - - (7)

58.9(13.3)” - Control 59 22

Carnat et al. (It) 15.6( 9.3)

Anorexia nervosa 160 22 52.9(23.0) Control

36.1(17-O) 19.9(10.9) 8.0 (5.2) 140 20

8.3(5.8)

Wells et al. (13) 15.4(11.0)

Unselected 9.9( 9.2) 7.1( 7.2) LO(2.1)

749 Range 1.9(2-l)

12.2(-) - - - Schoolgirls

6.6(-) 12-18

Williams et al. (14) Normals 54 15 Dieters

17.6( 8.3) 10.7( 6.6) 7.0( 5.4) 9

0.6(1.5) 15

3.1(3.0 18.7( 7.1)

Suspected bulimics 10.9( 6.6) 8.7( 5.5) 0.5(1.1)

8 16 1.7(1.5)

Szfiwkter (3) 25.5( 9.3)

Schoolgirls 17.0( 7.0) 12.0( 4.5)

1331 16 1.6(2.1) 3.4(2.5)

Mann et al. (12) 10.7( 9.3)

Schoolgirls 262 15 9.6( 6.4)

“Tote number in parentheses is the standard deviation.

Page 4: Use of the eating attitudes test and eating disorder inventory in adolescents

May 1987 EATING-DISORDER QUESTIONNAIRES 269

Table 3. E,nting Disorder Inventory Scores Reported in Previous Studies

Investigdtors

Mean Drive for age

Population n Body

(yr) Thinness Bulimia Dissatisfaction

113 21.8 15.4(5.3)”

48 21.0 65 22.4

577 19.9

Garner e/i al. Anorexia (16) nervosa

Restricters Bulimics

Female comparison

Recovered anorexia

Male comparison

Gp/rner et al. Anorexia (18) nervosa

Weight preoccupied

Not weight preoccupied

Williams et al. Normals

(14) Dieters Suspected

bulimics Bulimic

5.0(5.3)

2.7(3.8) 14.2(6.9) 10.8(5.6) 17.4(8.1) 2.0(3.4) 10.2(7.7)

17 23.9 3.6(5.4) 0.3(0.6) 6.3(6.2)

166 20.3 1.6(3.1) l.O(l.8) 3.9(5.0)

50 21.9 15.1(4.2) 8.3(5.1) 18.5(7.8)

35 19.8 16.1(3.4) 6.2(4.1) 15.8(6.3)

134 20.2 O.Cq3.5) O.li(4.3) 3.9(6.5)

54 15.4 4.8(5.2) 1.3(2.2) 9.7(7.3) 9 15.3 7.7(5.6) 2.1(3.4) 14.8(8.2) 8 15.9 8.9(4.9) 2.4(2.2) 12.6(8.1)

1 17 18.0(-) 17.0(-) ll.O(-)

“The number in parentheses is the standard deviation.

nervosa, bulimia, obesity, former obesity, and nor- mal controls. Subscale scores differentiate between these groupings, with little overlap of subjects with an eating disorder and normals (Table 3).

In a subsequent study, the ED1 was administered to 42 women with bulimia who had not met weight- loss criteria for anorexia nervosa (18). Scores were compared with those obtained from 42 restricting and 42 bulimic anorectic patients. Normal-weight bulimic patients scored significantly higher than re- stricting anorectic patients on the Drive for Thinness and the Bulimia subscales. Normal-weight bulimic and bulimic anorectic patients scored similarly on all subscales.

More recently, Garner and coworkers (19) have administered the ED1 to 237 female college students and 66 female ballet students. Women from these samples were then selected for inclusion in a weight- preoccupied group (N = 35) and a not weight-preoc- cupied group (N = 134) on the basis of their scores on the Drive for Thinness subscale. Cluster analysis of the weight-preoccupied women was performed using ED1 subscale scores. A 2-cluster solution iden- tified one group of 11 women characterized by ele- vated scores on all of the subscales, and a second

group of 24 women with elevated scores only on the Drive for Thinness, Body Dissatisfaction, and Per- fectionism subscales. Clinical interviews of a sub- sample of these women suggest that the first group had significant pyschopathology, whereas the sec- ond group were probably normal dieters. The au- thors conclude that “although there are some highly weight-preoccupied females who display psycho- pathology quite similar to anorexia nervosa, others only superficially resemble patients suffering from serious eating disorders.”

Williams and coworkers (15) administered the ED1 to 72 randomly selected young women in a junior or senior high school in Tucson, Arizona. Subjects were classified by clinical interview as normal, di- eter, suspected bulimia, and bulimia categories. T., &e results of their ED1 subscale scores are shown in Table 3. Normal adolescents had similar scores to the normal subjects in previous studies (17, 19).

Cooper and coworkers (20) questioned whether all eight subscales of the ED1 are specific for patients with eating disorders. The authors administered the EDI, EAT, and the General Health (GHQ) to 27 female psychiatric outpatients (age range 17-39 years) and found that only the Drive for

Page 5: Use of the eating attitudes test and eating disorder inventory in adolescents

270 WILLIAMS JOURNAL OF ADOLESCENT HEALTH CARE Vol. 8, No. 3

Table 3. (continued)

Ineffectiveness Perfectionism Interpersonal

Distrust Interoceptive Awareness

h4aturity Fears

14.4(8.0) 10.0(5.0) 7.7(5.2) 12.5(11.7) 6.0(10.5)

2.0(3.6)

3.1(3.9)

l.iq3.1)

13.7(7.6)

6.3(6.2) 3.5(4*0)

1*1(6.3) 2.0(4. I)

3,1(4.2) 3.7(3.8) 4.5(3*3) ¶3.0(--)

5.3(3.8) 4.2(3.1) 7,2(5.6) 13.0(-)

5.2(3.8)

6.5(3.6)

6.2(3,9)

6.9(4.9)

2.2(2.9)

1.9(2.3)

3.1(3.1)

9.5(5.9)

8.9(4.7)

5kq4.9)

3.5(4.2) 3.3(2.5) 3.6(3.6) 7.0(-)

2.9(11.3)

2.1( 3.5)

1.4( 5.0)

12.0( 6.1)

7.8( 4.9)

l.l( 5.0)

4.4( 5.4) 4.3( 5.6) 5.2( 2.0)

23.O(-)

2.5( 7.9)

l.l( 1.3)

2.7( 5.7)

5.3( 5.7)

4.7( 4.6)

2.2( 4.6)

3.8( 3.2) 3.8( 3.4) 3.8( 4.2) 4.0(-)

Thinness, Bulimia, and Body Dissatisfaction sub- scales were significantly correlated with the EAT to- tal score. The Ineffectiveness, Interpersonal Distrust, and Interoceptive Awareness subscale scores were significantly correlated with the GHQ, leading the authors to speculate that some of the

DI subscales assess dimensions of psychological distur common to patients with a variety of

psych al disorders.

As with all self-report questionnaires, the EAT and EDI are vulnerable to distortion by inaccurate subject reporting. Vanderdeycken and Vanderlinden (21) found in administering the EAT-40 to 40 patients with anorexia nervosa (mean age 23) that 13 patients scored Q 30, presumably because they were denying their illness.

Bven tho the EAT is reported to have a sen- sitivity of 9 and a specificity of 88% (9, 22), Wil- liams and coworkers (22) cautioned against using

as a screening instrument for anorexia se it is an uncommon illness. They

state that the positive predictive value of the EAT is only 0.19, i.e. of 100 high scorers on the test, only 19 wilI actually have anorexia nervosa.

Positive predictive value is the probability of dis-

ease in a patient with a positive (abnormal) test result (23). This value is determined by the sensitivity and specificity of the test and the prevalence of the dis- ease in the population tested. Positive results even for a very specific test, when applied to subjects with a low likelihood of disease, will be largely false pos- itives. Because the prevalence of anorexia nervosa in a high school population is less than l%, single screening tests should always yield a substantial number of false positives. Even a test with 99% spec- ificity and 99% sensitivity has a positive predictive value of only 0.50 for a disease with a prevalence in the population of 1%.

Several assumptions are made when the EAT and ED1 are scored as described by the original authors. The magnitude of the interval between the quasi- dimensional scores on individual items is not nec- essarily equal, yet a rating of “always” is assumed to be three times as important as a rating of “often.” Each item is weighted equally in the overall score. It is reasonable to suspect that some items may ac- tually be more valuable than others in classifying subjects as normal or abnormal.

To determine the importance of individual items in a multiple-item test, discriminant analysis is em- ployed. Discriminant analysis is a family of statistical methods used to predict in which of several discrete groups (e.g., normal, bulimic) an individual subject

Page 6: Use of the eating attitudes test and eating disorder inventory in adolescents

May 1987

should be classified (24). Williams and coworkers (15) performed a classical discriminant analysis of EAT and ED1 items to determine whether a combi- nation of items could predict the classification of junior and senior high school students (n = 71, mean age 15 years) into one of three categories (nor- mal, dieter, suspected bulimic) as predicted from a structured clinical interview. A preliminary equation was derived that correctly classified 86% of the sub- jects into the categories as determined from the in- terviews. A second discriminant analysis was performed, and the subjects were classified as nor- mal (normal, dieter) or abnormal (suspected bulimic, bulimic). In this analysis, 96% of the subjects were correctly classified. Classification of abnormals using the discriminant score from this analysis had a sen- sitivity of 67%, a specificity of 966, and a positive predictive value of 100%. Further research will be necessary to validate and improve upon these find- ings before such a discriminant analysis technique can be recommended as a screening device for sub- jects with eating disorders.

Finally, the original versions of the EAT and the ED1 were developed for college students; therefore some of the words used are not likely to be under- stood by younger patients or those with limited reading ability. Consequently, minor modifications in item wording may be necessary when adminis- tering these tests to adolescents.

Discussion How can practitioners benefit from usin;; self-report questionnaires such as the EAT and EDI in assessing adolescents with suspected eating disorders? The EAT and ED1 can be used to help confirm one’s clinical impression derived from a thorough history and physical examination. However, the benefits gained by using one or both of these tests must be weighed against the possible threat to the doctor- patient relationship caused by impersonal question- naires. The EAT and ED1 can be used to help monitor a patient’s progress as studies of current and re- covered anorexics (8,17) suggest that scores decrease with clinical improvement.

Which EAT should the clinician use? Although the EAT-40 total score is frequently reported in studies of patients with eating disorders, one must be cau- tious in interpreting the result when reported alone. Recent studies have shown that the 26-item version of the EAT has a high correlation with the 40-item version. In addition, three subscales related to di- eting, bulimia and food preoccupirtion, and oral con-

EATING-DISORDER QUESTIONNAIRES 271

trol can be derived from the 26-item test and may be useful in discriminating normals dram bulimics or anorectics. The additional information gained from determining EAT subscale scores suggests that the clinician may benefit from using these subscale scores as a supplement to the EAT-26 and EAT-40 total score.

Selected items from these self-report question- naires could be part of a screening program for eat- ing disorders in adolescents, if properly developed. Although the clinical interview remains the “gold standard” for diagnosing eating disorders, the time and expense involved in conducting clinical inter- views, combined with the relatively low occurrence of clinically significant eating disorders in the ado- lescent population, may make screening by clinical interview prohibitively expensive. Preliminary evi- dence (15) suggests that development of an effective self-report screening device for eating disorders in young women is feasible.

References 1.

2.

3.

4.

5.

Willi J, Grossman S. Epidemiology of anorexia nervosa in a defined region of Switzerland. Am J Psychiatry 1983;140:564- 7.

Crisp AH, Palmer RL, Kalucy RS. How common is anorexia nervosa? A prevalence study. Br J Psychiatry 1976;128:549- 54. SzmukIer GI. Weight and food preoccupation in a population of English schoolgirls. In Understanding anorexia nervosa and bulimia, Report of the Fourth Ross Conference on Med- ical Research. Columbus, OH: Koss Laboratories, 1983:21-B. Pyle RL, Mitchell JR, Eckert ED, et al. The incidence of bulimia in freshman college students. Int J Eating Disord 1982;2:75- 85. Halmi KA, Falk JR, Schwartz E. Binge-eating and vomiting: A survey of a college population. Psycho1 Med 1981;11:697- 706.

6. Johnson CL, Lewis C, Love S, et al. A descriptive survey of dieting and bulimic behavior in a female high school popu- lation. in Understanding anorexia nervosa and bulimia, Re- port of the Fourth Ross Conference on Medical Research. Columbus, OH: Ross Laboratories, 1983~14-20.

7. Killen JD, Taylor CB, Telch MJ, et al. Self-induced vomiting and laxative and diuretic use among teenagers: Precursors of the binge-purge syndrome? JAMA1986;255:1447-9.

8. Gamer DM, Garfinkel PE. The Eating Attitudes Test: An in- dex of the symptoms of anorexia nervosa. Psycho1 Med 1979;9:273-9.

9. Gamer DM, Garfinkel PE. Socio-cultural factors in the de- velopment of anorexia nervosa. Psycho1 Med 1980;10:647-56.

10. Button EJ, Whitehouse A. Subclinical anorexia nervosa. Psy chol Med 1981;11:509-16.

11. Lowe NC, Miles SW, Richards CG. Bating attitudes in an adolescent schoolgirl population. N Z Med J 1985;98:330-1.

12. Garner DM, Olmstead Ml’, Bohr Y, Garfinkel PE. The Eating Attitudes Test: Psychometric features and clinical correlates. Psycho1 Med 1982;12:871-8.

Page 7: Use of the eating attitudes test and eating disorder inventory in adolescents

272 WILLIAMS JOURNAL OF ADOLESCENT HEALTH CARE Vol. 8, No. 3

13.

14.

15,

16.

17,

10.

Mann AH, WaLeling A, Wood K, et al. Screening for abnor- ma1 eating attitudes and psychiatric morbidity in an unse- lected population of 15 year-old schoolgirls. Psychol Med 1983; 13573-80. Welts JE, Coope PA, Gabb DC, Pears RK. The factor structure of the Eating Attitudes Test with adolescent schoolgirls. Psy- rho1 Med 1985;15:141-6. Williams RL, Schaefer CA, Shisslak CM, Gronwaldt VH, Comerci CD. Eating attitudes and behaviors in adolescent women: Discrimination of normals, dieters, and suspected bulimics using the Eating Attitudes Test and Eating Disorder Inventory. lnt J Eating Disord 1986;5:879-94. Carter PI, Moss RA. Screening for anorexia and bulimia ner- vosa in a college population: problems and limitations. Addict Behav 1 ;9:417-9.

Garner DM, Olmstead MP, Polivy J. Development and val- idation of a multidimensional Eating Disorder Inventory for anorexia nervosa and bulimia. lnt J Eating Disord 1982;2:15- 34.

Gamer DM, Garfinkel PE, Q’ghaughnessy M. Clinical and psychometric comparison between bulimia in anorexia ner-

vosa and bulimia in normal-weight women. In Understand- ing anorexia nervosa and bulimia, Report of the Fourth Ross Conference on Medical Research. Columbus, OH: Ross Lab- oratories, 1983;6-13.

19. Gamer DM, Olmstead MP, Polivy J, Garfinkel PE. Compar ison between weight-preoccupied women and anorexia ner- vosa. Psychosom Med 1984;46:255-46.

20. Cooper Z, Cooper PJ, Fairbum CG. The specificity of the Eating Disorder Inventory. Br J Clin Psycho1 1985;24:129-30.

21.

22.

23.

24.

Vanderdeycken W, Vanderlinden J. Denial of illness and the use of self-reporting measures in anorexia nervosa patients. Int J Eating Disorcl 1983;2:101-7.

Williams P, Hand D, Tamopolsky A. The problem of screen- ing for uncommon disorders==-A comment on the Eating At- titudes Test. Psycho1 Med 1982;12:431-4.

Fletcher RH, Fletcher SW, Wagner EH. Clinical epidemiol- ogy--The essentials. Baltimore, Williams and Wilkins, 1982;52-6.

Brown GW. Discriminant analysis. Am J Dis Child 19&4;138:395-400.