THEVALIDATION OF
THE DISSOCIATIVE
EXPERIENCES SCALE
AGAINST THE CRITERION
OF THE SCID-D, USING
RECEIVER OPERATING
CHARACTERISTICS
(ROC) ANALYSIS
Nel Draijer, Ph.D.
Suzette Boon, Ph.D.
Drs. Draijer and Boon are psychologists in the Department
of Psychiatry, Free University of Amsterdam in Amsterdam,
The Netherlands.
For reprints write Nel Draijer, Ph.D, Department ofPsychiat-
ry, PCD, De Boelelaan 1117, 1081 HVAmsterdam, The Nether-
lands.
An earlier version of this paper was presented at the Ninth
International Conference on Multiple Personality/ Dissociative
States, Chicago, Illinois, October 1992.
Acknowledgements: This work was supported in part by a
grant from the Dutch Department of Welfare, Health and
Culture.
We wish to thank drs. L. Wouters, Department of Social and
Epidemiological Psychiatry, Municipal Health Service,
Amsterdam, who conducted the LABROC1-analysis.
ABSTRACT
Objective and method: The aim of this study is to analyze the util-
ity of the Dissociative Experience Scale (DES) as a screener for dis-
sociative disorders. The Structured Clinical Interview for DSM-III-
R DissociativeDisorders (SCID-D) was used as standard of comparison.
Forty-three patients with a dissociativedisorderand 36 control patients
with a range of psychiatric diagnoses participated in the study.
Results: The DES distinguishes dissociative disorder patients
from non-dissociative disorder patients very well (p<.0001); diag-
nostic utility of the DES based on Receiver Operating Characteristic
(ROC) analysis is excellent (.4 UC=. 96). The optimal cut-off score of
25 yields good to excellent sensitivity (93 %) and specificity (86% ).
The positive predictive value of the DES (>25) in random clinical
samples is limited (.26 - .54) due to the relatively low estimated
prevalence rate of dissociative disorders (5 - 15% respectively); the
negative predictive value is high (.99). The use of a confirmatory
interview such as the SLID-D is required to eliminate false positives.
INTRODUCTION
The Dissociative Experience Scale developed by Bernstein
& Putnam (1986) is the most widely used instrument for the
screening for dissociative symptomatology in clinical sam-
ples. Good reliability and validity have been reported at dif-
ferent centres (Bernstein & Putnam, 1986; Ross, Norton &
Anderson, 1988; Ensink & van Otterloo, 1989; Frischholz et
28
al., 1990).
Forthe screening of dissociative disorders in clinical sam-
ples different cut-off scores of the DES are suggested in the
literature (Chu & Dill, 1990; Carlson et al., 1990; Ross,
Anderson, Fleisner & Morton, 1991; Saxe, etal., 1993). Most
are based on median scores for certain diagnostic groups,
few are based on validation research. The 'golden standard'
for the assessment of the dissociative disorders in those stud-
ies was an independent clinical diagnosis according toDSM-
Illcriteria. No other criterion was available at the time.
To assess the prevalence of severe dissociative symp-
tomatology in a clinical sample (N=98), Chu and Dill (1990)
used a cut-off score of 31.3 based on the median for the 10
post traumatic stress disorder (PTSD) patients and a cut-off
score of 57.1 based on the 20 multiple personality disorder
(MPD) patients who participated in the original study by
Bernstein and Putnam (1986). Chu and Dill used the DES
without comparison with another clinical diagnostic inter-
view.
Carlson et al. (1993) analyzed the capacity of the DES
to distinguish between subjects with and without a clinical
diagnosis of MPD in a multicenter sample of 1051 subjects
with a range of psychiatric diagnoses. They concluded that
the DES performed quite well as a screening instrument to
identify subjects with MPD. Using discriminant analysis they
found a sensitivity of 76% (proportion of subjects with MPD
who were correctly classified) and a specificity of 76% (pro-
portion of subjects without MPD who were correctly classi-
fied). For clinical use they suggested a cut-off score of 30 to
identify patients likely to have MPD; this cut-off score result-
ed in their study in a sensitivity of 80% and a specificity of
80%: 31% of the subjects misclassified as having MPD had
another dissociative disorder and 30% had PTSD. Based on
an estimated prevalence-rate of MPD of 5% in random clin-
ical samples, they calculate the positive predictive value of
the DES: only 17% of the patients with a DES score of 30 or
more actually had MPD.
Ross et al. (1991), trying to estimate the prevalence of
MPD in a clinical population (N=299), used a cut-off score
of 20 with the motivation that DES scores beyond 20 are sug-
gestive of PTSD or a dissociative disorder; for their choice
of this cut-off point they refer to the original study of Bernstein
& Putnam (1986). Ross et al. found a prevalence rate of
patients with a DES score beyond 20 of 31% . Diagnostic inter-
views with the DDIS confirmed the presence of a dissocia-
tive disorder in '77.5% of these patients.
Saxe et al. (1993) chose a cut-off score of 25 for the same
DISSOCIATION, Vol. VI, No.1, March 1993
DRAIJER/BOON
purpose, because it is an intermediate to the scores of 30,
used by Chu and Dill (1990) and by Quimby and Putnam
(1991), and 20, used by Ross et al. (1991). Fifteen percent
of this clinical sample (N=110) scored above 25 on the DES.
Using the DDIS as diagnostic instrument they assessed a dis-
sociative disorder in 100% of those patients.
Neither Ross et al. (1991), nor Saxe et at. (1993) take
the possibility of false negatives - dissociative disorder patients
with a DES-score below the cut-off point - into account.
Steinberg, Rounsaville & Cicchetti (1991) were the first
to validate the DES as a screening instrument against a struc-
tured clinical interview (SCID-D) as a standard for systemati-
c comparison. They investigated its utility as a screening instru-
ment for the identification of patients at high risk for
dissociative disorders and examined several possible cut-off
scores. Their results indicate that a DES cut-off score of 15-
20 yields good to excellent sensitivity (90-95%) and speci-
ficity (93%) as a screening instrument in an outpatient pop-
ulation (N=36). For higher cut-off scores the sensitivity can
be much lower. Steinberg et al. conclude that high-risk patients
identified with the DES should be further evaluated with a
diagnostic instrument, such as the SCID-D or by in-depth
clinical follow-up.
Our study is to some extent a replication of this valida-
tion study of Steinberg et al. (1991): we use the SCID-D as
standard of comparison to determine the best possible cut-
off score of the DES. We use ROC-analysis to illustrate the
choice of optimal cut-off. The main difference is that we did
not use a normal comparison group, as we were interested
in the discriminant ability of the DES in clinical populations.
Sample sizes (79 versus 36) and characteristics (in- and out-
patients versus outpatients only) differ as well. And to
enhance understanding of the false positive en negative cases
we will give a clinical picture of the patients concerned. Finally
we will discuss the predictive value of the DES as a screener
for dissociative pathology.
METHOD
Instruments
1. The screening instrument: the Dissociative Experience
Scale. The DES is a 28 item self-report questionnaire
that is developed to quantify dissociative experiences
in both normal and clinical populations. The questions
are rated with slashes on 100-mm lines that indicate
where the subject falls on a continuum for each item.
The DES score ranges from 0 to 100 and represents the
mean of all item scores. The DES is not intended as a
diagnostic instrument for the assessment of the DSM-
III-R dissociative disorders, but has been used as a screen-
ing instrument for the identification of patients with a
dissociative disorder. Although a Dutch translation exist-
ed (Ensink & van Otterloo, 1989) we tested a new trans-
lation (Boon, Draijer & Van der Hart, 1988) that fol-
lowed the original more closely.
2. The Structured Clinical Interview forDSM-III-R Dissocia-
tive Disorders (SCID-D) (Steinberg, Rounsaville &
Cicchetti, 1990; Steinberg et al., 1991) is a diagnostic
instrument developed for the systematic assessment of
five dissociative symptom areas (amnesia, depersonal-
ization, derealization, identity confusion and identity
fragmentation) and for the assessment of the diagnoses
of the DSM-III-Rdissociative disorders. Severity ratings
of the 5 dissociative symptoms range form 1-4 (absent-
severe); the total SCID-D score range from 5-20. Good
to excellent reliability and validity have been reported
in the US as well as in The Netherlands (Steinberg et
al., 1990; Boon & Draijer, 1991; 1992; I993b).
Administration of the DES
To prevent bias, the DES-questionnaires were submitted one
week prior to the SCID-D interview. Patients with a dissocia-
tive disorder were asked to complete the DES by their treat-
ing clinician. Patients without a dissociative disorder were
given the DES by the independent psychiatrist, who had inter-
viewed them one week prior to the SCID-D interview. All
patients completed the DES by themselves and returned the
questionnaire at the SCID-D interview.
The SCID-D interview
All patients were interviewed with the SCID-D by the authors.
Interviews were videotaped or (in a few cases) audiotaped.
Informed consent, including consent to video- and audiota-
ping, was obtained from all patients.
Subjects
Two groups of psychiatric patients were compared on their
DES-scores: patients with and without a DSM-III-Rdissociative
disorder. Seventy-nine psychiatric patients-inpatients as well
as outpatients-participated in the study.
A. The dissociative disorder patients. This group consist-
ed of 43 patients with a dissociative disorder, assessed
by an independent clinician and confirmed by the authors
with the SCID-D: 20 patients with a diagnosis multiple
personality disorder (MPD) and 23 with a diagnosis dis-
sociative disorder not otherwise specified (DDNOS) . Two
patients were originally participating in the control group,
but a dissociative disorder (in both cases DDNOS) was
assigned based on the SCID-D interview.
B. The control group without a dissociative disorder. This
group consisted of 36 psychiatric patients (both inpa-
tients and outpatients), drawn from two university psy-
chiatric clinics. Clinical DSM-111--R diagnoses were
assigned on the basis of consensus within the treatment
teams, based on all available data. One week prior to
the SCID-D interview all control subjectswere interviewed
by an independent psychiatrist with the Present State
Examination (PSE) (Wing, Cooper & Sartorius, 1974)
and a selection of questions from the Structured
Interview for DSM-III-Rpersonality disorders (SIDP-R)
(Pfohl, Stangle, & Zimmerman, 1992). Dissociative dis-
orders in this group (n-2) were excluded with the SCID-
D by the authors.
29
DISSOCIAT10N, Vol.
VALIDATION OF THE DES
The control patients had a range of Axis I and II diag-
noses. On Axis I patients were diagnosed with: mood disor-
der; schizophrenia; delusional disorder; psychotic disorder;
eating disorder; somatoform disorder; obsessive compulsive
disorder; adjustment disorder; organic mental disorder; anx-
iety disorder. On Axis II patients were diagnosed with: bor-
derline personality disorder; histrionic personality disorder;
personality disorder not otherwise specified and dependent
personality.
Demographic characteristics
The two diagnostic groups did not differ in clinical set-
ting, nor marital status or employment. They slightly dif-
fered in age, dissociative disorder patients having a mean
age of 32.9 (SD=.3) versus controls having a mean age of
36.3 (SD=0.2) (t=1.67 df=78 p=.10).
Calculation of AUC and ROC curve.
We used I ABROCI-program for the calculation of the
Area Under Curve and the ROC curve. LABROC1 is a modi-
fied version by Metz et al. of the program RSCORE II
(Dorfman, 1982).
RESULTS
Demographic characteristics and DES-scores.
In the whole sample there was no significant relation of
DES-score with age, marital status and level of education,
nor with treatment setting. Neither was there a difference
between patients from different treatment settings (outpa-
tients versus inpatients) in the separate groups.
Reliability
Cronbachs alpha coefficient
(Cronbach, 1971) was used to estimate
the internal consistency of DES scores.
The alpha coefficient of the DES based
on 74 subjects with answers on all 28
questions was .96. The Dutch version
of the DES was found to be highly in to r-
nally consistent.
For its subscales (based on the fac-
toranalysis described by Carlson et al.
1991) Cronbach alpha was .90 for
amnesic dissociation (8 items) , .91 for
absorption and imaginative involve-
ment (9 items) and .88 for deper-
sonalization and derealization (6
items). The subscales are highly inter-
nally consistent as well.
Validity
The dissociative disorder patients
differed significantly from the non-dis-
sociative controls in the severity of the
dissociative experiences measured by
the DES (t=11.1 df=76 p=<.00001).
Dissociative disorder patients had a
30
mean DES score of 47.6 (SD=+16.3) and a median of 46.8
(range 11.6 - 81.3). Patients without a dissociative disorder
had a mean DES score of 12.0 (SD=1.4) and a median of
9.3 (range 0.0 - 38.6) . The mean DES score of the two groups
differed more than two standard deviations. That is more
than the slight age difference could account for. A graph-
ic representation of the frequency distributions of DES-scores
in both groups is presented in Figure 1.
Among the dissociative disorder group we found sig-
nificant differences on the DES-scores between MPD and
DDNOS patients; patients (n=20) had a meanDES score
of 56.8 (SD-3.4) and a median of 57.8; DDNOS patients
(n=23) had a mean DES score of39.7 (SD=4.5) and a medi-
an of 40.7. Those two groups did not differ on the severity
of amnesia, depersonalization, derealization and identity alter-
ation as measured by theSCID-D; they differed slightly on
identity confusion (t=2.11 df=42 p<.05), the MPD patients
reporting more confusion as to who they were. We will dis-
cuss the meaning of these results later.
In the whole sample the total DES-score correlated signifi-
cantly with the severity of the five dissociative symptoms,
assessed with the SCID-D: amnesia (r=.68), depersonaliza-
tion (r=.64), derealization (r=.58),identity confusion (r=.76)
and identity fragmentation (r=.78). Both total scores cor-
related strongly (r=.78). Correlations between the severity
of the five symptom areas assessed with theSCID-D and the
subscales of the DES (cf. Bernstein et al., 1991) reached from
.58 till .73; all three subscales correlated most strongly with
the severity of identity alteration, amnesic dissociation and
depersonalization/derealization even more so than with their
counterparts in the SCID-D. Although the two instruments
have different purposes - theDES being a screening instru-
ment and the SCID-D being a clinical diagnostic instrument
FIGURE 1
Dissociative disorders vs controls. Frequency distribution of DES.
Percentage
15
10
5
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95100
DES scores
Dissoc. die. (N-43) Pe. controls (N6)
DI55OCIATIO`, Vol. \1, No. 1, March 1993
DRAIJER/BOON
- these results support the convergent
validity of theDES with the SCID-D as
criterion.
Sensitivity
We used the SCID-D as the standard
of comparison or `truth standardto
analyze the different DES cut-off scores.
Table 1 shows the sensitivity and speci-
ficity values, false positive and false neg-
ative rates at each cut-off score.
Sensitivity at a certain cut-off score indi-
cates how likely a patient with a dis-
sociative disorder is to have a score
above this cut-off point; specificity at
a certain cut-off score indicates how
likely a patient without a dissociative
disorder is to have a score below this
cut-off point.
ROC analysis
Analysis of the receiver operating
characteristics (ROC) is-according to
Rey, Morris-Yates and Stanislaw
(1992)-the only tech-
nique currently available
that provides an overall
index of diagnostic accu-
racy that is not dependent
on prevalence (as is pos-
itive predictive value) or
on the cut-off score (as
are sensitivity and speci-
ficity) . ROC analysis orig-
inally used in radiologi-
cal and biomedical
research, is judged to be
highly promising in the
comparison of the qual-
ities of psychiatric screen-
ing tests (Mari 
Williams, 1985; Bridges 
Goldberg, 1986; Murphy
et al., 1987; Weinstein,
Berwick, Goldman,
Murphy Barsky, 1989;
Hsiao, Bartko Potter,
1989; Rey et al., 1992).
An index of discriminating
ability of a screening instru-
ment can be obtained
from ROC analysis. The
most useful index is the
area under the ROC curve
(Swets 1979; Swets,
Pickett, Whitehead 
Getty, 1979). This curve
FIGURE 2
Estimated binormal ROC curve. Dissociative Experience Scale.
Comparing 43 DD-pat. with 36 Ps. controls.
10 20 30 40 50 80 70 80 90 100
False Positive Rate
- DES ROC +- Random ROC
Standard of comparison: SCID-D
AUC = .96 SE = .02
FIGURE 3
Positive predictive value of DES at different estimated prevalence rates
for dissociative disorders.
Cut-off scores DES
Estimated Prevalence
- 5% 10% 15%
Based on DES against SCID-D
Positive Predictive Value
100
90
80
70
60
50
40
30
20
10
0
0 5 10
3 I I 1 I I I
15 20 25 30 35 40 45 50 55 60 65 70 75 80 85
31
DISSOCIATION, Vol. VI. No. I, March 1993
VALIDATION OF THE DES
is obtained by plotting sensitivity against false positive rate
for all possible cut-off points. The area under curve (AUG)
varies from 50% (for a test with no better discriminating
ability than chance) till 100% (for a test with perfect dis-
crimination between the patients with and without the dis-
order: all values fall along the upper and left boundaries).
Figure 2 is based on the sensitivity rates plotted against the
specificity for all possible cut-off points of the DES; it shows
the estimated binormaI ROC curve. This curve indicates that
the DES discriminates rather well between dissociative and
non-dissociative disorders: 95.6% (Standard Error = 2.3%)
of the trapezoid is under the curve (Hanley & McNeil 1982) .
This means that the DES has an extremely high discrimi-
nating ability. Following ROC-analysis the optimum cut-off
point (a compromise between high sensitivity and low false
positive rate) was at the point on the ROC curve which is the
greatest perpendicular distance from the diagonal (Mari &
Williams, 1985) . This results in a DES cut-off score of 25 yield-
ing optimal sensitivity (93%) and specificity (86%).
Predictive Value of the DES
The positive predictive value of a positive test result is
defined as the proportion of subjects correctly identified by
the test as having the disorder (true positives) to all subjects
TABLE 1
Sensitivity, specificity and predictive value of the Dissociative Experience Scale at a range of cut-off scores, discrim-
inating between 43 psychiatric patients with and 36 patients without DSM-III dissociative disorders, also positive
predictive values at estimated prevalence rates of 15%, 10%, and 5%.
Cut-off Sensitivity Specificity False Pos.Rate Positive Negative Predictive Value
Score (1spec)/
False Neg. Rate
(1 seas)
Predictive
Value
Predictive
Value
Prevalence: 15%, 10%, 5%
Positive
15%/10%5%
Neg.
15%
85 0.00 1.00 0.00/ 1.00 1.00 0.46 1.00/ 1.00/ 1.00 0.85
80 0.02 1.00 0.00 / 0.98 0.99 0.46 Imo/ Imo/ Loo 0.85
75 0.05 1.00 0.00 / 0.95 0.99 0.47 1.00/ 1.00/1.00 0.85
70 0.12 1.00 0.00 / 0.88 0.99 0.49 1.o0/1.00/1.00 0.86
60 0.21 1.00 0.00/ 0.79 0.99 0.51 ].oo/Imo/ Lao 0.87
65 0.21 1.00 0.00 / 0.79 0.99 0.51 1.00/ 1.00/1.00 0.87
55 0.28 1.00 0.00 / 0.72 0.99 0.54 1.00/ 1.o0/ 1.o0 0.89
50 0.47 1.00 0.00 / 0.53 0.99 0.61 1.00/1.00/1.00 0.91
45 0.56 1.00 0.00 / 0.44 0.99 0.65 1.00/ 1.00/ 1.00 0.92
40 0.65 1.00 0.00 / 0.35 0.99 0.71 1.00/ 1.00/ 1.00 0.94
35 0.81 0.89 0.11 / 0.19 0.90 0.80 0.57/0.45/0/28 0.98
30 0.88 0.89 0.11/ 0.12 0.90 0.86 0.57/0.45/0/28 0.98
25 0.93 0.86 0.14/ 0.07 0.89 0.91 0.54/0.42/0.26 0.99
20 0.95 0.78 0.22/ 0.05 0.84 0.93 0.42/0.32/0.19 0.99
15 0.95 0.75 0.25 / 0.05 0.82 0.93 0.40/0/30/0/17 0.99
10 1.00 0.50 0.50/ 0.00 0.70 1.00 0.26/0.18/0.10 1.00
5 1.00 0.31 0.69 / 0.00 0.63 1.00 0.20/0/14/0/07 1.000
0 1.00 0.00 1.00/ 0.00 0.54 1.00 0.15/0.10/0.05 1.00
Positive predictive value (corrected for prevalence) _ [sensitivity x prevalence]/ [(sensitivity x prevalence) + (1-specificity) x (I-
prevalence)]
Negative predictive value (corrected for prevalence) = [specificity. x (1-prevalence)] / [(specificity x (I-prevalence) + (1-sensitivity)
x prevalence]
32
DISSOCIATION, Vol.March
DRAIJER/BOON
identified by the test as having the disorder (true positives
+ false positives) (see Table 2). In our sample at a cut off
score of 25 the positive predictive value of a positive score
on the DES is 40 / 40 + 5 = .89 (89%). The negative predic-
tive value of a negative test result is defined as the propor-
tion of subjects correctly identified by the test as not having
the disorder (true negatives) to all subjects identified by the
test as not having the disorder (true negatives + false nega-
tives) . In our sample at a cut off score of 25 the negative pre-
dictive value of a negative score on the DES is 31 / 31 + 3 =
.91 (91%).
Predictive value, however, is affected by the prevalence
of the disorder in that particular population. Even when sen-
sitivity and specificity are high, the predictive power of a test
is low if the prevalence of the condition in that population
is low (Rey et al., 1992). We calculated different positive pre-
dictive values for different prevalence rates of dissociative
disorders among psychiatric patients: a relatively high esti-
mate of 15% (based on Saxe et al., 1993), a more conserva-
tive estimate of 5% (Carlson et al., 1993) and a value in
between. The positive predictive value of a DES cut-off score
of 25 would drop from 89% to 54% at prevalence rate 15%,
to 42% at prevalence rate 10% and to 26% at rate 5% (Table
1; Figure 3).
Implications for screening clinical use DES
For screening of dissociative pathology in a random psy-
chiatric population one needs to have a high sensitivity and
a high negative predictive value: as many cases as possible
that have the disorder need to be selected and the chance
that a negative test score really excludes the disorder illness
needs to be maximal. So in our sample one could prefer a
cut-off score of 20 for screening purposes, reaching a sen-
sitivity of 95% and a specificity of 78%; the 22% false posi-
tives at that rate need to be excluded by clinical assessment
or the use of a structured clinical interview, such as the SCID-
D.
In a random clinical sample, though, a cut-off score of
25 has an optimal negative predictive value (99% of the cases
with a DES below 25 can be expected not to have the disor-
der), but a limited positive predictive value. At an estimat-
ed prevalence rate of 15% only 54% of the positive scores
can be expected to have a dissociative disorder and at an
estimated prevalence of 10% only 42% (Table 1).
For clinical use, to identify patients likely to dissociate,
a score of 40 predicts a dissociative disorder in all cases: the
estimated positive predictive value in a random clinical sam-
ple assuming a prevalence of dissociative disorders of 15%
is 100%. But at this score one `missesmany dissociative dis-
order patients: in our sample 37% of all patients with a dis-
sociative disorder had a score below 40. For a detailed sum-
mary of results we refer to Table 1.
Qualitative analysis false negativesand false positives.
With 25 as optimal cut-off point, we found in our sam-
ple a false negative rate of 7% and a false positive rate of
14% (Table 1). To get a clinical understanding of deviant
DES-scores, we analyzed the 7% `false negativesand the 21%
`false positivesqualitatively.
Patients with disorderand a low DES score (< 25):
false negatives
The two patients with a dissociative disorder and a DES-
score below 25 (11.6 and 13.6 respectively), turned out to
be both cases with strong resis-
tance against acknowledging
the dissociative symptoms. One
women was able to report severe
dissociative symptoms in the
structured clinical interview
(SCID-D)--amnesia, deper-
sonalization, derealization,
identity-confusion, and
fragmentation-and reached
a total SCID-D score of 20 (which
is the highest possible). The
other women had showed diffi-
culty to admit the presence of
dissociative symptoms, but was
positive on all indirect ques-
tions of the SCID-D interview.
Both patients met criteria for
dissociative disorder not oth-
erwise specified. In both cases
the diagnosis of a dissociative
disorders was independently
confirmed over time. The clin-
ical pictures of the two patients
were as follows.
TABLE 2
Two x two contingency table obtained when using a cut-off score of
25 on the Dissociative Experience Scale
Diagnosis
Dissociative
Disorder (MPD/DDNOS)
No Dissociative
Disorder
Score on DES (N=43) (N=36)
> 25 (N=45) 40 (TP) 5 (FP)
< 25 (N=34) 3 (FN) 31 (TN)
Of all subjects 89,9% were correctly classified at this cut-off score.
Sensitivity = T P/(TP+FN) = .93 False Negative Rate = .07
Specificity = TN/(TN+FP) = .86 False Positive Rate = .14
Positive predictive value = TP/(TP+FP) = .89
Negative predictive value = TN/(TN+FN) = . 91
33
DISSOCI.ITIOS, Vol. V I. o. I, Mardi 1 93
VALIDATION OF THE DES
One 33-year old patient was clearly ambivalent and very con-
fused about herself. At the research interview she was ini-
tially minimizing and denying dissociative symptoms and
she showed signs of a continuous internal struggle. Moreover
she was recurrently dissociating during the interview.
Gradually she was able to give more information. She also
told the interviewer that she heard almost continuously voic-
es in her head that told her not to answer the questions.
Although at the research interview it became clear that she
probably suffered fromMPD, this diagnosis was not assigned
because it was not yet possible to get information on alter
personalities. At follow-upMPD was confirmed.
The second patient was 19 years old and had just fin-
ished highschool, left her family of origin and started uni-
versity in a different part of the country. She had had a his-
tory of (pseudo) seizures and had been treated for epilepsy
since age 16, although the epilepsy was not clearly corrobor-
ated by EEG findings. Since she had left home, there had
been a dramatic increase in seizures. Moreover, she had only
recently become aware that often after a seizure she would
change into a younger person with the same name as she
had, who was very confused and unaware of the current date
or the place were she was. The patient was totally amnesic
for these episodes but had heard in detail about these "younger
selves" from friends who were looking after her. They had
told her that they had met several younger persons with dif-
ferent ages - 12, 14 and 16. These younger persons seemed
to be unaware of the existence of each other and of the fact
that they were currently at a university in another part of the
country. Some were very anxious, others were preoccupied
that they had to go home and see the father. At the research
interview the patient was telling this in detail without any
emotion. She did report clear amnesic episodes, that would
always start with a pseudoseizure. She did not report indi-
rect indications for amnesic episodes such as finding things
she couldn account for etc. She reported occasional deper-
sonalization or derealization and denied identity confusion.
She did not report any Schneiderian symptoms and didn
dissociate during the interview. She reported vague, frag-
mented memories of sexual abuse by her father, starting at
age 12.
Although this patient definitely minimized some of her
symptoms, the low mean DES score was more in concurrence
with the way she presented at the research interview.
Control patients without a dissociative disorder, but with a high
DES score (>25): false positives
Among the control patients without a dissociative dis-
order, five patients had a meanDES score above the cut-off
point of 25 (28.6, 35.9, 37.5, 38.3 and 38.6). Four of these
patients had a DES score that fell in the range of scores of
patients with a dissociative disorder not otherwise specified
(DDNOS) or a post traumatic stress disorder.
It is of interest to note that these five patients did not
have comparable high scores on theSCID-D,in fact two patients
had the lowest possible total score of 5, which means
that, at the SCID-D interview, no dissociative symptoms were
reported. Two patients had reported recurrent episodes of
34
depersonalization and derealization at theSCID-D interview
(with atotalSCID-D score of 9 and one of 11) and one patient
(total SCID-D score of 7) reported depersonalization which
primarily seemed to be associated with the use of soft drugs
(marihuana). A further analysis of the five patients with a
high mean DES score showed the following:
The first patient (DES score 28.6; totalSCID-D score 11)
was 42 years old. She was in out-patient treatment and was
assigned a diagnosis of schizo-affective disorder with
histrionic and borderline traits. She had a long psychiatric
history (since age 22). There was drug and alcohol addic-
tion in the past. She was currently on depot neuroleptics.
At the SCID-D interview she reported recurrent deperson-
alization in the present and severe derealization associated
with psychotic episodes during which she was hospitalized.
She reported a history of physical abuse by her mother for
which she reported to the police as a teenager.
The second patient (DES score 38.6; totalSCID-D score
5) was 47 years old. She was assigneda DSM-III-Rdiagnosis
of somatoform pain disorder on axis I and histrionic per-
sonality disorder on axis II. The independent psychiatrist
who had assessed thePSE and the SIDP-R had commented
that the patient seemed to aggravate her symptoms. At the
SCID-D interview she did not report dissociative symptoms,
at the trauma interview she did report severe emotional neglect.
The third patient (DES score 35.9; totalSCID-D score 9)
was 23 years old. She had no current axis I diagnosis and a
histrionic personality disorder on axis II; she had reported
some depersonalization and derealization at the inter-
view; her most important complaints were panic attacks and
the inability to be alone. She also had had an anorectic episode
during adolescence. She did not report a history of physi-
cal or sexual abuse.
The fourth patient (DES score 38.3; totalSCID-D score
5) was 50 years old and inpatient at the time of the inter-
view. She was assigned the diagnosis schizophrenia, para-
noid type and personality disorder not otherwise specified
with histrionic and borderline traits. She had had a history
of psychotic episodes since age 30 and several psychiatric
admissions. The independent psychiatrist, who assessed the
PSE and the SIDP-R had commented that the patient was a
classical case of hysterical psychosis. She did not report dis-
sociative symptoms at the SCID-D interview and she did not
report a history of physical or sexual abuse.
The last patient (DES score 37.5; totalSCID-D score 7)
was 37 years old. She was assigned the diagnosis schizophre-
nia and was called "chronic psychotic". She used soft drugs
(marihuana) regularly. She reported recurrent feelings of
depersonalization at the research interview, but these feel-
ings seemed to be closely associated with the use of soft drugs.
DES scores of two patients from the original psychiatric control
condition were identified as having a dissociative disorder
with the SCID-D.
A dissociative disorder was detected in two patients, who
originally participated in the control group. These two patients
had entered the psychiatric control group with a diagnosis
of borderline personality disorder. They earned a high mean
DISSOCIATION.Vol.VI,No.I.Mardi 1993
DRAIJER/BOON
DES score of 42 and 57. If the DES had been used as a screen-
ing instrument these two cases would have been identified
with both instruments.
DISCUSSION
This study validates the Dissociative Experience Scale
(DES) against a structured clinical interview (SCID-D; Steinberg
et al. 1990; 1991) as a standard for systematic comparison.
The results show that the Dutch version of the DES (transla-
tion Boon, Draijer & Van der I-Tart) discriminates at a high
level of significance between patients with and without dis-
sociative disorders. We also found a high overall correlation
between mean DES scores, total SCID-D scores and SCID-D
severity ratings of separate dissociative symptoms.
Our results confirm those found in other studies indi-
cating that the DES is a valid and reliable self-report instru-
ment to measure dissociative pathology (Bernstein & Putnam,
1986; Ross et al. 1988; Ensink & van Otterloo, 1989; Frischholz
et al., 1990; Carlson et al., 1993). Our mean and median
scores for MPD patients were comparable to those of Bernstein
and Putnam's original study and several replication studies
both in North America as well as in the Netherlands
(Bernstein & Putnam, 1986; Ensink & van Otterloo, 1989;
Frischholz et al., 1990).
We investigated the utility of the DES as a screening instru-
ment for the identification of patients at high risk for dis-
sociative disorders and examined several possible cut-off scores
using ROC-analysis. The index of discriminating ability of
the DES, based on the area under the ROC curve, was .96.
This value means that the DES is a test with a very high diag-
nostic utility. The index of diagnostic accuracy is not depen-
dent on prevalence (as is positive predictive value) or on the
cut-off score (as are sensitivity and specificity). Our results
indicate that 25 is the optimal cut-off score, yielding good
to excellent sensitivity (93%) and specificity (86%) in a select-
ed clinical population (N=79). What do these results mean
for the use of the DES as a screener in random clinical sam-
ples? We calculated the estimated positive predictive value
of the DES based on different estimated prevalence rates for
dissociative disorders in a clinical sample. At a base rate of
15% (Saxe et al., 1993), this calculation shows a drop of the
positive predictive value at cut-off score 25 from 89% in our
sample to 54% in a random clinical sample. At a base rate
of 10% it drops even further to 42%, and at 5% it drops to
26%. This means that using the DES as a screener in clini-
cal samples, one certainly needs to use a clinical diagnostic
instrument, such as the SCID-D, to select the `true positives'.
Ransohoff and Feinstein (1978) drew attention to the
fact that many diagnostic tests have proved to be valueless
after optimistic introduction into medical practice, due to
the use of a too narrow spectrum for the `diseased' and 'non dis-
eased' patients in the study population. They state that the
sensitivity of a test should be examined in a broad range of
patients with the disorder and that a test should be chal-
lenged for its specificity in a broad range of patients with-
out the disorder.
We may illustrate this by comparing our results to those
of Steinberg et al. (1991) and Carlson et al. (1993). In the
first study almost identical false negative rates, but much
lower false positive rates were found than we did at a cut-off
score of 25 (7% versus 14%) or 20 (7% versus 22%). One
explanation could be that Steinberg et al. limited their study
to outpatients and excluded patients who were very agitat-
ed, gravely disabled, or at risk of suicide, whereas we inter-
viewed inpatients as well, some of whom were just recover-
ing from a psychotic episode at the time of the interview.
Steinberg saw 21 psychiatric patients which a range of Axis
I diagnoses; we saw 36 control patients among whom sever-
al also had an Axis II diagnosis. On the other hand, Steinberg
studied a range of dissociative disorder patients, whereas in
our sample the dissociative disorders were accidentally lim-
ited to MPD and DDNOS. Evidently more research on the
diagnostic utility of the DES is needed using a wide spectrum
in pathology of dissociative as well as non-dissociative patients.
Carlson et al. (1993) assessed the capacity of the DES to
blindly predict a psychiatric diagnosis of MPD in a large pool
of general psychiatric patients. According to discriminant
analysis on a subgroup of 883 subjects (out of 1051) more
closely representing patients in a typical psychiatric facility
in terms of prevalence rates of MPD, they found a false pos-
itive rate of 15% and false negative rate of 24%. The false
positive rate is almost identical as in our study. The high
false negative rate could be due to the relatively low mean
DES score for MPD-patients (42.8, SD=?19.2), which could
possibly be explained by a high representation of MPD-patients
in early stages of treatment.
The analysis of the false negative cases in our study showed
that some patients with a dissociative disorder are unable to
give an accurate self report, because they are unaware of
their symptoms or deny them. This is, to a certain extent,
also confirmed by the interesting findings (1) that a signif-
icant difference was found in mean DES scores of patients
with MPD and patients with DDNOS, but (2) that these groups
did not differ significantly in severity of dissociative symp-
toms derived with the SCID-D interview. Moreover, at one
year follow up we obtained information on 20 of the 24 patients
with DDNOS: in 19 of those 20 patients the diagnosis MPD
instead of DDNOS was made by the treating clinician and a
description of distinct alterpersonalities could be given! These
findings confirm the following clinical observations: A major-
ity ofMPD patients initially minimizes, denies or is unaware
of their dissociative symptoms (Kluft 1987a, 1987b). A self-
report questionnaire at that stage may be problematic, because
some of these patients deny or may be unaware of their dis-
sociative symptoms and therefore are unable to give an accu-
rate self-report.
When MPD patients have accepted the diagnosis and
are more aware of their dissociative symptoms or do not have
to deny these symptoms so much, this may influence their
scores at a self-report questionnaire. Our hypothesis is that
this phenomenon maybe one of the explanations of the fact
that DES scores of MPD patients have ranged considerably
-from 40.7 to 57-in different studies (Bernstein & Putnam,
1986; Ross, et al., 1988; Ensink & Van Otterloo, 1989; Ross,
Miller, Reagor, et al., 1990; Boon & Draijer, 1993a).
A further analysis of the false positive cases in our study
showed that here was no convergence between the relatively
35
DISS0CL1TI0N. Vol. V1. No. I. March 1993
VALIDATION OF THE DES
high DES score and the scores derived from the SCID-D inter-
view; a dissociative disorder could be easily ruled out with
the SCID-D. Moreover, there were distinct qualitative differ-
ences in the descriptions of the dissociative experiences of
patients with and without a dissociative disorder. Clearly other
mechanisms - for instance suggestibility or a tendency to
aggravate symptoms - may have influenced the relatively
high DES scores of these patients (see also Frankel, 1990).
This asks for some caution in the interpretation of high DES
scores, if no confirmatory diagnostic interviews are done.
CONCLUSION
The Dutch version of the Dissociative Experience Scale
is a reliable and valid instrument to screen for dissociative
pathology. It has a high diagnostic utility according to the
results of ROC-analysis (AUC=.96). A cut-off score of 25 is
optimal, yielding good to excellent sensitivity and specifici-
ty. In spite of those optimistic results, the estimated positive
predictive value of the DES for a random clinical sample is
rather low (26%, 42% or 54%) due to the relatively low esti-
mated base rate of dissociative disorders (5%, 10%, or 15%
respectively). Screening for dissociative disorders in a ran-
dom clinical sample will result in a certain amount of false
positive cases. Clinical assessment or the use of a confirma-
tory interview such as the SCID-D is required in order to diag-
nose the presence or absence of a dissociative disorder. Such
a clinical diagnostic interview is more able to identify cases
of a dissociative disorder with a subtle presentation, who
may go undetected with a self-report questionnaire. These
patients may be resistant to describe their dissociative symp-
toms or may be unaware of such symptoms and therefore
are unable to complete an accurate self-report. 
REFERENCES
Bernstein, E.M. & Putnam, F.W. (1986). Development, reliability,
and validity of a dissociation scale. of Nervous and Mental
Disease, 174,727-735.
Boon, S. & Draijer, N. (1991). Diagnosing dissociative disorders in
the Netherlands: A pilot study with the structured clinical inter-
view for DSM-III-Rdissociative disorders.
148, 4, 458-462.
Boon, S. & Draijer, N. (1992). Reliability and validity
Clinical Interview for Dissociative Disorders: report on a
study in the Netherlands.
on Multiple Personality/Dissociative States.
Boon, S. & Draijer, N. (1993a). Multiple Personality Disorder in
the Netherlands: A clinical investigation of 71 patients.
Journal of Psychiatry. 150, 3,489-494
Boon, S. & Draijer, N. (1993b).Multiple Personality Disorder in the
Netherlands. A study on reliability and validity of the diagnosis.
Zeitlinger B.V., Amsterdam/Lisse
Bridges, KW. & Goldberg, D.P. (1986). The validation of the GHQ-
28 and the use of the MMSE in neurological inpatients. British
Journal of Psychiatry, 148,548.
Carlson, E.B., Putnam, F.W., Ross, C.A., Torem, M., Coons, P? Dill,
D., Loewenstein, R.J., & Braun, B.G. (1993). Validity of the
Dissociative Experiences Scale in screening for MPD: A multicen-
ter study.American Journalof Psychiatry,150,1030-1036
Chu, J.A. & Dill, D.L. (1990). Dissociative symptoms in relation to
childhood physical and sexual abuse.
147, 887-892.
Cronbach, L J. (1971) . Test Validation. In: Thorndyke
Measurement.Second Edition. American Council on Education.Washington DC, 443-507.
Dorfman, D.D. (1982). RSCORE. InJ.E.Swets & R.M.Pickett (Eds.),
Evaluation ofDiagnosticSystems: from SignalDetection Theory.
Academic Press, New York.
Ensink, B J. & Otterloo van, D. (1989). A validation study of the
DES in the Netherlands.DISSOCIATION, I1(4), 221-223.
Frankel, F.H. (1990). Hypnotizability and dissociation.American
Journal of Psychiatry,147, 823-829.
Frischholz, EJ., Braun, B.G., Sachs, G.R., Hopkins, L., Shaeffer,
D.M., Lewis, J., Leavitt, F., Pasquotto,J.N. & Schwartz, D.R. (1990)
The Dissociative Experiences Scale: Further Replication and Vali-
dation.DISSOCIATION, [11(3), 151-153.
Hanley, J.A. & McNeil, B J. (1982). The meaning and use of the
area under a receiver operating characteristics (ROC) Curve.
Radiology,143, 1982, 29-36.
Hsiao, J.K., Bartko, JJ. & Potter, W.Z. (1989). Diagnosing diag-
noses: Receiver Operating Characteristic methods and psychiatry.
Archivesof General Psychiatry, 46,
Kluft, R.P. (1987a). Making the diagnosis of multiple personalitydisorder. In: Flach, F. (Ed.),
Diagnostics and psychopathology,225. NewYork: W.W. Norton.
Kluft, R.P. (1987b). An update on multiple personality disorder.
Hospital and Community Psychiatry,
Mari, J.D J. & Williams, P. (I985) . A Comparison of the Validity of
two Psychiatric Screening Questionnaires (GHQ-12 & SRQ-20) in
Brazil, using Relative Operating Characteristic (ROC) analysis.
Psychological Medicine,
Metz, C.E., Wang, P., ShenJ. & Kronman, H.B. Departmentof Radiology and the Franklin Mclean Memorial Research Institute.
University of Chicago, Chicago Illinois,
Murphy, TM, Berwick, D.M., Weinstein, M.C., Borus,J.F., Budman,
S.H. & Klerman, G.L. Performance of screening and diag-
nostic tests: application of receiver operating characteristic analy-
sis.Arch. Gen. Psychiatry,44, 550.
Pfohl, B., Stangl, D., & Zimmerman, M. (1982) .
for DSM-lllPersonality Disorders (SIDP).and Clinics, Iowa City I.A.
Quimby, L.G. & Putnam, F.W. (1991). Dissociative symptoms and
aggression in a state mental hospital. IV(1), 21-24
36
DISSOCIITION. 1,N1ardi
DRAIJER/BOON
Ransohoff, D.F., Feinstein, A.R. (1978) Problems of Spectrum
and Bias in Evaluating the Efficacy of Diagnostic Tests. The New
England journal of Medicine, 26, 926-930.
Rey, J.M., Morris-Yates, A., Stanislaw, H. (1992). Measuring the
Accuracy of Diagnostic Tests Using Receiver Operating Characteristics
(ROC) Analysis. ofMethods in PsychiatricResearch,
2, 39-50.
Ross, C.A., Anderson, G., Fleisner, W.P., Morton, G.R. (1991).
The frequency of Multiple Personality Disorder among Psychiatric
Inpatients. American Journal of Psychiatry,148, 12, 1717-1720.
Ross, C.A., Heber, S., Norton, R.,  Anderson, G. (1989).
Differences between Multiple Personality Disorder and Other
Diagnostic Groups on Structured Interview.Journal of Nervous and
Mental Disease,177,8, 487-491.
Ross, C.A., Joshi, S., Currie, R. (1991). Dissociative experiences
in the general population: A factor analysis.Hospital and Community
Psychiatry, 42, 3,297-312.
Ross, C.A., Miller, D.S., Reagor, P., Bjornson, L., Fraser, G.A., 
Anderson, G. (1990). Structured interview data on 102 cases of
multiple personality disorder from four centers. American Journal
of Psychiatry,
Ross, C.A., Norton, G.R., Anderson, G. (1988) The dissociative
experiences scale: A replication study.DISSOCIATION, I(3), 21-22.
Saxe, G.N., Kolk van der, B.A., Berkowitz, R., Chinman, G., Hall,
K, Lieberg, G., Schwartz, J. (1993) Dissociative Disorders in
Psychiatric Inpatients, ,150,1037-1042.
Schwartz, D.R., Frischholz, EJ., Braun, B.G., Sachs, R.G. (1991)
A confirmatory factor analytic study of the Dissociative Experience Scale
(DES).
Personality/Dissociative States, Chicago.
Steinberg, M., Rounsaville, B., Cicchetti, D.V. (1990). The
structured clinical interview for DSM-III-Rdissociative disorders:
Preliminary report on a new diagnostic instrument. American
journal of Psychiatry, 147, 76-82.
Steinberg, M., Rounsaville, B., Cicchetti, D. (1991). Detection
of dissociative disorders in psychiatric patients by a screening instru-
ment and a structured diagnostic interview, American Journal of
Psychiatry,148, 8, 1050-1054.
Swets,J.A. (1979). ROC analysis applied to the evaluation of imag-
ing techniques. Investigative Radiology,
Swets, J.A., Pickett, KM., Whitehead, S.F., Getty, D.J. (1979).
Advanced Measurement methods illustrated in a study of computed
tomography of the brain. In Swets J.A. Freeman B.A. (Eds.),
Assessment diagnostic technologies, 205,753-759.
Weinstein, M.G., Berwick, D.M., Goldman, P.A., Murphy, J.M., 
Barsky, A.J. (1989). A Comparison of Three Psychiatric Screening
Tests Using Receiver Operating Characteristic (ROC) Analysis.
Care, 27, 6, 593-607.
Wing, J.K., Cooper, J.E., Sartorius, N. (1974) Measurement and
classification of p.sychiatric symptoms; an instruction manual for the PSE
and Catego program.Cambridge University Press, London.
37
DISSOCIATION, Vol. VI, No. 1. March 1993