See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/232555737

A revision of the Bulimia Test: The Bulit-R

Article in Psychological Assessment · March 1991

DOI: 10.1037/1040-3590.3.1.119

CITATIONS READS

535 4,629

4 authors, including:

Marcia Smith Pasqualini

Avila University
41 PUBLICATIONS 2,537 CITATIONS

SEE PROFILE

All content following this page was uploaded by Marcia Smith Pasqualini on 05 July 2020.

The user has requested enhancement of the downloaded file.

 Psychological Assessment: Copyright 1991 by the American Psychological Association, Inc.
A Journal of Consul! ing and Clinical Psychology 1040-3 590/91/$3.00
1991. Vol.3, No. 1,119-124

A Revision of the Bulimia Test: The BULIT-R

Mark H. Thelen and Janet Farmer

University of Missouri-Columbia

Stephen Wonderlich Marcia Smith

University of North Dakota Washington University

The Bulimia Test was revised (BULIT-R) to accommodate the DSM-III-R criteria of bulimia
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

nervosa. A 28-item, self-report, multiple-choice scale was developed by comparing responses of

This document is copyrighted by the American Psychological Association or one of its allied publishers.

clinically identified female bulimics with those of female college students. Cross-validation was
performed on independent samples of bulimic and college control subjects. The BULIT-R was a
good predictor of group membership. The scale was then completed by female college students, and
a stratified sample of these subjects was retested and interviewed approximately 2 months later. The
results of retesting and diagnostic judgments based on interviews showed that the BULIT-R was a
reliable and valid predictor of bulimia nervosa in a nonclinical population. The BULIT-R corre-
lated highly with 2 other measures of bulimia, indicating a high degree of construct validity.

Bulimia was first recognized as a distinct psychiatric dis- order to prevent weight gain. These criteria contain several sig-
order only a few years ago (American Psychiatric Association, nificant changes from the earlier criteria, a few of which will be
1980); however, researchers and practitioners have come to real- mentioned here. The DSM-1II criteria (American Psychiatric
ize that bulimia is a significant health problem. Research on the Association, 1980) included a reference to depression and did
prevalence of bulimia in the general population is relatively not allow for a diagnosis of bulimia if the person met the crite-
scarce. One study reported that 1.9% of the women who at- ria for anorexia nervosa. Both of these criteria were dropped
tended a family-planning clinic in Oxford were diagnosed buli- from DSM-IH-R. Two of the significant criteria present in
mic (Cooper & Fairburn, 1983). Estimates of prevalence rates DSM-III-R, but not in DSM-III, were a specified minimum
among college women have been higher than the estimates of frequency of binge episodes and a persistent overconcern with
prevalence in the general population. Among female college body weight and shape.
students, prevalence estimates have ranged from 2% to over Several studies used assessments that approximated the
18% (Thelen, McLaughlin-Mann, Pruitt, & Smith, 1987). Even DSM-lll-R criteria for bulimia nervosa. Schotte and Stunkard
if only the most conservative estimates of prevalence are ac- (1987) used a 15-item self-report measure of DSM-IH-R crite-
cepted, it is clear that bulimia is a serious health problem that ria in a study, but no estimates of reliability or validity were
warrants further research. reported. Similarly, Drewnowski, Yee, and Krahn (1987) used a
Recently, the criteria for bulimia nervosa were revised in the questionnaire based on the DSM-III-R criteria for bulimia
revised third edition of the Diagnostic and Statistical Manual of and, in a 1988 study (Drewnowski, Hopkins, & Kessler, 1988),
Mental Disorders (DSM-IH-R; American Psychiatric Associa- used this questionnaire adapted for a telephone interview.
tion, 1987) and the term bulimia nervosa replaced the earlier Again, however, no reliability and validity estimates were pre-
label of bulimia. The new criteria are as follows: (a) persistent sented. To date, it appears that there is no self-report measure
overconcern with body shape and weight; (b) recurrent episodes with established reliability and validity that assesses bulimia
of binge eating (rapid consumption of a large amount of food in nervosa as specified in the DSM-III-R criteria.
a discrete period of time); (c) a minimum average of two binge- The Bulimia Test (BULIT) is a self-report questionnaire
eating episodes a week for at least 3 months; (d) a feeling of lack measure of bulimia that is based on the DSM-III criteria for
of control over eating behavior during the eating binges; and (e) bulimia (Smith & Thelen, 1984). The BULIT has demonstrated
regularly engaging in either self-induced vomiting, use of laxa-
validity with both clinically identified bulimic populations and
tives or diuretics, strict dieting or fasting, or vigorous exercise in
with nonclinical college female populations. Test-retest reliabil-
ity was established with the latter group. The purpose of this
study was to revise the BULIT in an effort to devise a reliable
We are grateful to the following people for their assistance in collect-
and valid measure of bulimia nervosa that accommodates the
ing data from subjects for this study: Barbara Bauer, Lauri Humphries,
DSM-III-R criteria.
Melanie Katzman, Retta Lusky, Linda Pendleton, Kenneth Rockwell,
and Donald Williamson. A self-report instrument based on DSM-III-R would be a
Correspondence concerning this article should be addressed to cost-effective means to identify individuals with bulimia ner-
Mark H. Thelen, Department of Psychology 210 McAlesterHall, Uni- vosa, especially in nonclinical populations that have a relatively
versity of Missouri-Columbia, Columbia, Missouri 65211. low prevalence of bulimia (Thelen et al., 1987). Such a scale

119
 120 THELEN, FARMER, WONDERLICH, AND SMITH

could also be a useful adjunct to the diagnostic interview in original 36 items of the BULIT (Smith & Thelen, 1984) and ending with
clinical settings. a final version of 36 items that incorporated the DSM-IH-R criteria.
To develop the BULIT-R, items were written and revised on All items were presented in a 5-point, forced-choice, Likert format, in
the basis of the responses of several groups of bulimic and con- which responses were mutually exclusive and exhaustive. In all stages,
subjects1 responses were scored by giving up to 5 points for items an-
trol subjects. Data from these same groups were used to deter-
swered in the extreme bulimic direction, and down to I point for items
mine the validity and reliability of the measure. The scale was
answered in the extreme normal direction. Eor several items, the 5-
then cross-validated using independent samples of bulimic and
point Likert response scale was reversed to prevent a response bias
control subjects. Finally, the BULIT-R was administered to a based on order of presentation (scoring was reversed for these items).
nonclinical sample to determine its effectiveness as a screening Specific details about item selection and analysis during Stage 1 are
instrument and its reliability over time. outlined in the Results section.
Stage 2 consisted of cross-validation of the 36-item BULIT-R with a
replication sample of bulimic (ft = 23) and control (n = 157) women. In
Method Stage 3. the BULIT-R was administered to 1,739 college women. A
subset (n = 819) of these subjects was also given the Binge Scale (Haw-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Subjects
kins & Clement, 1980). Finally, Stage 4 involved further assessment of
This document is copyrighted by the American Psychological Association or one of its allied publishers.

The subjects at all stages, except Stage 4, constituted independent selected (n = 161) Stage 3 subjects. At this testing, each subject partici-
samples (i.e., a given subject was not tested at more than one stage). pated in a structured diagnostic interview consisting of questions
Stage IA. Subjects in this stage were 24 female bulimics and 116 about the subjects' eating attitudes and behaviors that pertain to the
female controls. Bulimic subjects were referred by therapists in several DSM-JI1-R criteria for bulimia nervosa. Although all of the inter-
Midwestern and Southern states. Each bulimic subject had sought help viewers followed a structured and predetermined interview procedure
for her condition at an eating-disorder clinic and currently met the similar to the one used in the development of the BULIT (Smith &
DSM-IH-R criteria for bulimia nervosa, as judged by the clinician Thelen, 1984), the diagnosis was not based on a standardized instru-
who was treating her. The control subjects were college undergraduates ment. The interviewer and an independent rater, who listened to au-
of varying body weights from psychology classes in two Midwestern diotapesof the interviews, judged whether or not the subject met each
universities. College women were used as the control group because of the DSM-III-R criteria for bulimia nervosa. Subjects who were
most bulimics are female and of college age, and a disproportionate rated as meeting all of the criteria were judged as bulimic If = .68). A
percentage are of either the middle or upper social class (Hsu, 1989). third independent rater was used in cases of disagreement. Graduate
Although the control groups in Stage I and Stage 2 were not screened students and research assistants served as interviewers and raters and
for bulimia nervosa, the base rate was considered to be sufficiently low were unaware of the subjects' BULIT-R scores. They were trained by
to allow meaningful comparisons with the clinical bulimia nervosa us, and we led discussions and role plays of the structured interview
group (Thelen etal., 1987). and reviewed the first several diagnostic judgments with the raters and
Stage IB. Subjects in this stage were 25 female bulimics and 204 interviewers. The interviewers also had clinical experience. All of the
female control subjects. As in Stage 1 A, the bulimics met the DSM-III- subjects were given the BULIT-R at this stage, and a portion of these
R criteria for bulimia nervosa, as judged by their therapists, and the subjects were tested again approximately 2 months after Stage 3 in
controls were college women in psychology classes. order to assess the test-retest reliability of the BULIT-R. These sub-
Stage 1C. This stage contained 21 female bulimics (asdiagnosed by jects were also given the BULIT (Smith & Thelen, 1984) to determine
their therapists) and 100 female control subjects from psychology the relation between the BULIT and the BULIT-R.
classes.
Stage 2. Test validation in Stage 2 (replication sample) was per-
formed on independent groups of bulimic and control women who Results
were selected in the same manner as described for Stage 1 subjects.
In instances in which all of Ihe individual items were ana-
Twenty-three female bulimic and 157 female control subjects were
lyzed with multiple t tests and correlations, the alpha level was
tested.
adjusted using the Bonferroni procedure. With the Bonferroni
Stage 3. A large number (n = 1739) of women from psychology
classes participated in this stage of the study. adjustment, the overall alpha was held at the .05 level. When 28
Stage 4. One hundred sixty-one of the subjects from Stage 3 partici- items from the BULIT-R were analyzed, a Bonferroni correc-
pated in the last stage of the study. tion of p < .002 was used. For comparisons based on 36 items
from the BULIT-R, the correction was p< .001.

Procedure
Stage 1. Test Construction
Bulimic subjects were individually administered the BULIT-R if
their therapist determined that they met the DSM-IH-R criteria tor a Stage IA. The 36 items from the BULIT and 22 new items
diagnosis of bulimia nervosa. The therapists did not use a standard- pertaining to new criteria in DSM-IH-R constituted the first
ized diagnostic instrument, and there was no assessment of reliability draft of the BULIT-R. Twenty-six items were retained from the
of the diagnosis. However, these therapists were practitioners experi-
original pool of 58 items because they discriminated best be-
enced with eating-disordered clients and were informed about the
tween the bulimic (n = 24) and control (« = 116) groups (I tests
clients' bulimic symptomatology because they had just sought treat-
showed that all itemsdiscriminated at p < .001) and covered the
ment. Control subjects completed Ihe BU LIT-R as part o/a battery of
tests given to college-level psychology classes. These subjects received various DSM-fIJ-R criteria for bulimia. Because 8 of Ihe items
extra credit for their participation. pertaining to specific means of radical weight control (two
Validation of the BULIT-R was accomplished in fourstages. During items each concerning laxatives, diuretics, fasting, and exercise)
Stage 1, the BULIT-R was constructed. Test construction took place did not discriminate well between the groups, they were not
over three phases (Stages IA, IB, and 1C), beginning with a base of the scored but were retained to provide information about
 BULIMIA TEST-REVISED (BULIT-R) 121

these behaviors. Finally, the wording on 1 item was changed. the expected direction ((s ranged from 3.84 to 17.68; ps < .001).
Thus, the next version of the BULIT-R included 34 items, with An overall validity coefficient was obtained by correlating BU-
8 items unscored. LIT-R scores with group membership (/•= .67, p < .0001). The
Stage IB. Although t tests revealed that each of the 26 scored validity of the individual items was similarly assessed by corre-
items discriminated well (p < .001) between the bulimic (n = lating item scores with group membership. For 27 items, the
25) and control (« = 204) groups, the total scores did not pro- point-biserial correlations ranged from .26 to .80 (ps < .001;
vide satisfactory predictive ability. Items pertaining to concern mean r of items = .52). Analysis with Cronbach's coefficient
with body shape and weight appeared to be a problem because alpha indicated high internal consistency Ift = .97) within the
many normals indicated concerns about body shape and scale.
weight. Of the 34 scale items, 26 items were retained without Using the cutoff of 104 on the BULIT-R that was established
change. The wording of 8 items was changed, 5 of which per- at Stage 1C, 4 of the 23 bulimic subjects and 7 of the 157 control
tained to concern about body weight and shape, and 1 each that subjects were misclassified. It is likely that several of the 7 con-
pertained to exercise, control over eating, and binge eating. In trol subjects who scored above the cutoff on the BULIT-R
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

addition, 10 new items were written: Seven of these pertained to would be judged as bulimic in an interview. A summary of the
This document is copyrighted by the American Psychological Association or one of its allied publishers.

concerns about body weight and shape, and each of the other 3 predictive ability of the BULIT-R is presented in Table 1.
items referred to all of the radical means of weight control speci-
fied in the DSM-III-R criteria. At this point the scale con- Stage 3. Administration of Scale to a Nondinical
tained 44 items, including the 8 nonscored items. Population

Because the BULIT-R was developed to identify bulimic

Stage 1C
individuals in nonclinical as well as in clinical populations, the
On the basis of analyses of the bulimic (n = 21) and control BULIT-R was administered to 1,739 college women. Although
(n = 100) subjects' responses to each of the 36 scored items, 8 the percentage of college women who are bulimic is low, it is
items were deleted. Six of the 8 items pertained to body shape generally believed that the prevalence of bulimia is higher
and weight and did not discriminate between the groups as well among college women than among other nonclinical popula-
as other items assessing these features. One item concerning tions (Thelen et al, 1987). Therefore, college women were con-
binge eating discriminated poorly. Finally, a vomiting item was sidered to be a suitable nonclinical population with which to
dropped because it correlated highly (.95) with another similar assess the validity and reliability of the BULIT-R. The subjects
item. The final version of the BULIT-R contained 28 scored were classified as bulimic if their total BULIT-R score was 104
items covering all of the DSM-1I1-R criteria and 8 unscored or above, and normal if their score was less than 104. Fifty-three
items referring to specific weight-control behaviors.' subjects (3%) met the scale criterion for bulimia.
BULIT-R scores were computed by summing responses to
the 28 scored items. BULIT-R scores showed a significant dif- Stage 4. Validation of Scale in a Nonclinical Population
ference between bulimic (M= 117.95) and normal (M = 57.50)
Attempts were made to contact all of the 53 subjects in Stage
groups, 1(46) = 16.41, p < .0001. Each of the 28 items discrimi-
3 who met the scale criterion for bulimia and all of the 20
nated well between bulimic and control groups ((s ranged from
subjects who scored just below the cutoff(BULIT-R score 99-
4.64 to 14.01; all ps<.OOI). A validity coefficient was obtained
103). Thirty-nine of the bulimics and 17 of those with BULIT-
by correlating total BULIT-R scores with group membership
R scores between 99 and 103 were tested at this stage. The
(/•= .74, p < .0001, point-biserial correlation). The statistic is
bulimics who were retested at Stage 4 (n = 39) did not differ on
sufficiently large to suggest that BULIT-R scores and diagnosis
the BULIT-R at Stage 3 from those who were not retested (n =
on the basis of a clinical interview are good correlates. The
14; BULIT-R means = 111.56 and 111.21, respectively; ([51] =
validity of individual items was established by correlating item
.14, p = .89). In addition, a randomly chosen stratified sample
scores with group membership. For the 28 scored items, the
of 105 subjects who scored less than 99 was retested. An approxi-
point-biserial correlation ranged from .39 to .79 (ps < .001;
mately equal number of subjects was randomly chosen from
mean r of items = .58).
each of four ranges: those with BULIT-R scores between 81-
A cutoff score of 104 was chosen because it resulted in the
98, 63-80, 45-62, and 28-44.
smallest number of errors in classification. Five subjects were
Reliability. A reliability coefficient was computed by com-
misclassified: Two bulimics had BULIT-R scores less than 104,
paring subjects' (n = 98) total scores from the first testing (Stage
and 3 controls had BULIT-R scores equal to or greater than
3) with their scores from Stage 4 testing 2 months later. Overall
104.
test-retest reliability was found to be .95 (p < .0001) using the
Pearson product-moment correlation.
Stage 2. Cross-validation Predictive ability. Total BULIT-R scores from Stage 4 test-
ing showed a significant difference between the subjects that
The BULIT-R was administered to independent samples of
the raters judged to be bulimic (M = 104.54) and the subjects
bulimic (« = 23) and control (n = 157) subjects to determine the
judged to be normal (M= 60.71), ((100) = 13.02, p < .0001. All
test's predictive ability. BULIT-R scores again showed a signifi-
cant difference between bulimic (M= 118.08)andcontrol(M=
1
59.62) groups, ((41) = 17.08, p < .0001. All of the 28 items A copy of the BULIT-R, along with scoring information, may be
showed highly significant differences between group means in obtained from Mark H. Thelen.
 122 THELEN, FARMER, WONDERLICH, AND SMITH

Table 1 based on an initial varimax rotation (SAS program). A promax

Summary of Predictive Ability for the Bulimia matrix (oblique) rotation was used because the factors were theo-
Test-Revised (BULIT-R) retically expected to correlate with each other. Stage 2 analyses
included 180 subjects: 23 bulimics and 157 control subjects.
Measure Stage 2 Stage 4
Stage 4 analyses included 161 subjects: 37 bulimics and 124
BULIT-R control subjects (based on judges' ratings). The results sug-
£104 gested that items targeted for each criterion area clustered to-
Bulimic* 19 23 gether (see Table 2 for highest factor loadings).2
Control 7 5 In Stage 2, five factors emerged, using a criterion of an eigen-
<104
Bulimic" 4 14
value of 1.0 that accounted for 46.8% of the total variance. Fac-
Control 150 119 tor labels were created on the basis of the primary items loading
Predictive ability on each factor. Factor 1 included items pertaining to binging
Sensitivity .83 .62 and control, and Factor 2 contained items pertaining to general
False-negative rate .17 .38
use of radical weight-loss measures and body image. Nonscored
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Specificity .96 .96

This document is copyrighted by the American Psychological Association or one of its allied publishers.

False-positive rate .04 .04 items that specifically targeted fasting were included in Factor
Positive predictive value .73 .82 2. Factor 3 comprised the four nonscored items related to laxa-
Negative predictive value .97 .89 tive and diuretic use. Factor 4 contained items related to vomit-
b
ing, and Factor 5 clustered the nonscored items related to exer-
" Referred by therapist. Based on raters' judgments.
cise. All items that were targeted to the same criterion area
showed positive communal loadings. Because of the small sam-
of the 28 items also showed highly significant differences be- ple size, these findings are only suggestive.
tween group means(ts ranged from 3.67 to 13.79; ps < .001). An The overall factor structure from BULIT-R responses with
overall validity coefficient was obtained by correlating total Stage 4 data appeared to be similar to Stage 2 data, with the
BULIT-R retest scores with group membership based on rater principal exception of questions regarding body image. Again,
judgment (r - .62, p < .0001). The validity of individual items five factors emerged, accounting for 41.6% of the total variance.
was assessed by correlating item scores with rater judgment. Factor 1 comprised items concerning binging, control, and
The predictive ability of each individual item is supported by body image. Questions pertaining to body image had consis-
the point-biserial correlations (rs ranged from .35 to .63; mean tently smaller loadings (.49 to .64) than other items. Factor 2
r= .49). All 28 items were significant at the .001 level of confi- included items related to extreme weight-loss measures and
dence. fasting. Exercise again emerged as a separate factor, whereas
On the basis of the BULIT-R scores at Stage 4, 28 of the 161 laxative use was factored with vomiting rather than diuretic use
subjects were classified as bulimic (see Table 1). Fourteen buli- as in Stage 2.
mic and 5 normal subjects (as judged by the raters) were mis- Two items concerning perceived body shape after eating did
classified by the BULIT-R. As Table 1 shows, the specificity, not cluster with their criterion area. Item 32 clustered with the
positive predictive value, and negative predictive value were .82 Binge factor in Stage 2, and Item 35 clustered with Radical
or higher, and the sensitivity was .62. These figures, however, Measures at Stage 4.
may represent an underestimate of the scale's true ability be-
cause the sampling method used to select subjects for Stage 4 BULIT-R Compared With Other Eating Disorder Scales
resulted in an overrepresentation of subjects with scores close to
The Binge Scale (Hawkins & Clement, 1980), a measure of
the cutoff score. Seventeen of the subjects interviewed fell just
binging behavior, was completed by 819 nonclinical subjects at
below the cutoff at Stage 3 (BULIT-R score 99-103). Also, the
Stage 3. A Pearson product-moment correlation of .85 (p <
stratified sampling procedure resulted in an overrepresentation
.0001) between the BULIT-R and the Binge Scale suggested
of subjects with relatively high BULIT-R scores. These subjects
that the two scales are based on similar constructs.
are probably more difficult to classify than subjects with ex-
To determine the similarity between the BULIT-R and the
treme scores; an exceptionally rigorous test of the BULIT-R's
BULIT, the BULIT was given to 98 subjects at Stage 4. A Pear-
predictive ability was thus produced.
son product-moment correlation coefficient of .99 (p < .0001)
Researchers using the BULIT-R for screening purposes may
showed that these two measures were very similar. To examine
wish to reduce the number of false negatives by using a cutoff
the relation between the BULIT and the DSM-1II-R criteria, a
lower than 104. For example, using a cutoff of 85, only one of
point-biserial correlation was computed between group mem-
the 37 subjects who were judged as bulimic on the basis of the
bership (on the basis of interview ratings) and the BULIT. A
interview fell below this cutoff. Ten percent of the Stage 3 sam-
correlation of .57 (p < .0001) suggested that the BULIT pro-
ple scored 85 or higher on the BULIT-R.
vides a reasonably good assessment of the DSM-IH-R criteria.
The relation between the BULIT and DSM-II1-R criteria (as
Factor Analyses determined by interview judgments) was further explored by
Principal-factor analyses of all 36 BULIT-R items were per-
formed separately on Stage 2 and Stage 4 data, using iterative 2
Tables showing the loading of each item on each factor for the Stage
estimates of item communalities and promax matrix rotation 2 and Stage 4 data may be obtained from Mark H. Thelen.
 BULIMIA TEST-REVISED (BULIT-R) 123

Table 2
Highest Factor Loadings for BULIT-R Items at Stage 2 and Stage 4

Factor Item

Stage 2 data
1. Binging/Control I (.46), 2 (.73), 3 (.70), 8 (.86), 9 (.89), 10 (.70), 13 (.44), 16
(.87), 21 (.82). 22 (.84), 23 (.89), 28 (.63), 30 (.69), 32 (.56),
33 (.76), 34 (.73)
2. Radical Weight-Loss 4 (.69), 5 (.60), 7 (.76), 12 (.73), 14 (.73), 17 (.47), 19 (.70), 24
Measures/Body Image (.78), 25 (.66), 29 (.88), 35 (.58)
3. Laxatives/Diuretics 6 (.70), 27 (.84), 31 (.73), 36 (.86)
4. Vomiting 15 (.78), 18 (.38), 26 (.91)
5. Exercise I1 (.90), 20 (.80)
Stage 4 data
1. Binging/Control/Body Image
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

I (.66), 2 (.85), 3 (.81), 4 (.64), 7 (.57), 8 (.78), 9 (.64), 10 (.56),

12 (.50), 13 (.80), 14 (.55), 16 (.73), 17 (.59), 21 (.65), 22
This document is copyrighted by the American Psychological Association or one of its allied publishers.

(.91), 23 (.90), 24 (.54), 28 (.79), 30 (.60), 32 (.60), 33 (.93),

34 (.85)
2. Radical Measures/Fasting 5 (.64), 18 (.69), 19 (.62), 25 (.68), 29 (.65), 35 (.47)
3. Exercise I1 (.68), 20 (.72)
4. Vomiting/Laxatives 6 (.65), 15 (.71), 26 (.74), 31 (.65)
5. Diuretics 27 (.87), 36 (.83)

Note. Factor loadings are in parentheses.

categorizing subjects who scored above the cutoff for bulimia The scale showed good predictive ability with a nonclinical
(BULIT score £ 102) as bulimic and those who scored below population of college women. Of the 28 subjects who scored
the cutoff as normal. This cutoff was suggested by Smith and above the cutoff on the BULIT-R, 23 were judged bulimic on
Thelen (1984). Using these procedures, 31 subjects scored the basis of the interview. Hence, a person who scores at or
above the cutoff for bulimia on the BULIT. On the basis of above the 104 cutoff on the BULIT-R is likely to be diagnosed
interview ratings, 16 of these subjects were judged to be bulimic as bulimic in an interview. Of the 37 who were judged to be
using DSM-III-R criteria, and 15 were judged to be normal. Of bulimic on the basis of an interview, 14 scored below the cutoff
the 20 subjects who were judged as bulimic by interview, only 4 on the BU LIT-R. We remind the reader that the Stage 4 popula-
scored below the cutoff on the BULIT. Sixty-three subjects were tion contained an overrepresentation of subjects who scored
identified as normal by both the BULIT and the interview high on the BULIT-R but below the cutoff. Use of the scale
judgments. with a general sample of college women would result in more
subjects with low BULIT-R scores, increasing the predictive
Discussion
ability of the BULIT-R.
The purpose of this project was to revise the BULIT (Smith & One important use of the BULIT-R is as an easily adminis-
Thelen, 1984) to accommodate the DSM-III-R criteria of bu- tered, cost-effective screening instrument to be followed by a
limia nervosa. The data showed that the BULIT-R is a reliable clinical interview. In effect, the scale would be used to identify
and valid measure by which to identify individuals who have subjects who are most likely to be diagnosed as bulimic on the
bulimia nervosa. Cross-validation with independent samples of basis of an interview. If researchers who are using the BULIT-R
control and bulimic subjects suggested that the scale has ade- for screening purposes wish to minimize the number of false
quate predictive ability. The BULIT-R was also validated with negatives, they should consider using a cutoff lower than 104.
a nonclinical population of college women. Scale scores were This procedure, followed by an interview, might be appropriate
predictive of diagnosis, as judged on the basis of independent in epidemiological research in which it is important to mini-
clinical interviews. mize false positives and false negatives.
The use of college women as nonbulimic control subjects in Although the Binge Scale and the BULIT are directed toward
Stages 1 and 2 placed some limitations on the data. Almost the DSM-III criteria for bulimia, it was considered that these
certainly some of these subjects were bulimic and would have measures should correlate with the BULIT-R. Indeed, the high
been diagnosed as bulimic in an interview, thus resulting in a correlation between the subjects' scores on the BULIT-R and
very conservative estimate of the predictive ability of the BU- these two scales further supports the validity of the BULIT-R.
LIT-R. Using a control group of women who have a non-eat- In view of the high correlation between the BULIT and the
ing-related psychological disorder would have certain advan- BULIT-R, we might ask if the BULIT-R adds any incremental
tages. Such a control group would permit a determination of validity to the BULIT. Using the recommended cutoffs for each
whether the BULIT-R measures non-eating-related psychopa- of these instruments, Stage 4 data suggests that the BULIT
thology instead of, or in addition to, symptoms specific to bu- identifies more false positives, and the BULIT-R identifies
limia. more false negatives. Using a different cutoff might change the
 124 THELEN, FARMER, WONDERLICH, AND SMITH

number of false positives and false negatives; however, these research should examine the correlation between the BULIT-
cutoffs were appropriately established using independent sam- R and measures that are unrelated to eating disorders. It is
ples. Certainly, the BULIT-R has more face validity than the hoped that this scale will be helpful to researchers who are
BULIT because the item content is directed toward the DSM- working in the area of eating disorders and to clinicians who
III-R criteria. wish to use a standardized measure to aid in their diagnosis of
Another observation about the BULIT warrants comment. bulimia nervosa.
Using interview-based judgments to determine diagnosis, most
References
of the subjects who were diagnosed as bulimic on the basis of
DSM-IH-R criteria scored above the cutoff on the BULIT. American Psychiatric Association. (1980). Diagnostic and statistical
However, only about one half of those who scored above the manual of mental disorders (3rd ed.). Washington, DC: Author.
cutoff on the BULIT were diagnosed as meeting the DSM-III- American Psychiatric Association. (1987). Diagnostic and statistical
R criteria for bulimia nervosa. It would appear that the BULIT, manual of mental disorders (3rd ed., rev). Washington, DC: Author.
which is based on the DSM-III criteria, identifies some people Cooper, Z., & Fairburn. C. G. (1983). Binge-eating and self-induced
vomiting in the community: A preliminary study. British Journal of
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

as bulimic who do not meet the DSM-IH-R criteria (high false-

Psychiatry, 142,139-144.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

positive rate). These data suggest that the DSM-H1-R criteria

Drewnowski, A., Hopkins, S. A., & Kessler, R. C. (1988). The preva-
are more stringent than the DSM-III criteria. lence of bulimia nervosa in the U.S. college student population.
Although the development of the BULIT-R was based on American Journal of Public Health. 78.1322-1335.
the premise of a single dimension that was affirmed by a high Drewnowski, A., Yee, D., & Krahn, D. D. (1987, August). Bulimia on
coefficient alpha, there are multiple criteria for a diagnosis of campus: Incidence and recovery fates. Presented at the American Psy-
bulimia. Therefore, factor analyses were performed on the data. chiatric Association Annual Meeting, Chicago, IL.
Factor analyses of the Stage 2 and Stage 4 data revealed some Fairburn, C. G., & Garner, D. M. (1988). Diagnostic criteria for an-
orexia nervosa and bulimia nervosa: The importance of attitudes
factors that are similar to the DSM-III-R criteria for bulimia
toward shape and weight. In D. M. Garner & P. E. Garfinkle (Ed.),
nervosa. At both stages, items pertaining to binging and control
Diagnostic issues in anorexia nervosa and bulimia nervosa (pp. 36-
clustered together, suggesting that criteria concerning control
55). New York: Brunner Mazel.
may be an unnecessary redundancy. Items concerning body Hawkins, R. C, & Clement, P. F. (1980). Development and construct
image clustered with extreme measures of weight control at validation of a self-report measure of binge eating tendencies. Addic-
Stage 2, but at Stage 4 body-image items clustered with the tive Behaviors, 5, 219-226.
Binging/Control factor. These data raise questions about over- Hsu, L. K. G. (1989). The gender gap in eating disorders: Why are the
concern with body shape and weight as a criterion of bulimia eating disorders more common among women? Clinical Psychology
nervosa. The items targeted to this criterion did not generate a Review. 9, 393-407.
separate factor, and they were not associated with the same Schotte, D. E., & Stunkard, A. J. (1987). Bulimia vs. bulimic behaviors
on a college campus. Journal of the American Medical Association,
factor at Stage 2 and at Stage 4. Difficulties with this criterion
258.1213-1215.
were apparent during the early stages of scale development
Smith, M. C, & Thelen, M. H. (1984). Development and validation of a
when we observed that many control subjects expressed consid-
test for bulimia. Journal o! Consulting and Clinical Psychology, 52,
erable concern with their body shape and weight. Although 863-872.
concerns about body weight and shape in the bulimia nervosa Thelen, M. H., McLaughlin-Mann, L., Pruitt, J., & Smith, M. (1987).
syndrome have been studied (Fairburn & Garner, 1988), more Bulimia: Prevalence and component factors in college women. Jour-
research is needed. nal of Psychosomatic Research, 31, 73-78.
Validation studies of the BULIT-R with other populations
(e.g., community samples, high school populations) would be Received December 13,1989
helpful to determine the generalizability of this instrument. To Revision received May 29,1990
determine the discriminant validity of the BULIT-R, future Accepted June 26,1990 •

View publication stats