Vous êtes sur la page 1sur 9

Journal of Personality Disorders.

3(1), 10-18, 1989


< 1989 The Guilford Press

THE REVISED DIAGNOSTIC INTERVIEW


FOR BORDERLINES: DISCRIMINATING BPD
FROM OTHER AXIS II DISORDERS

Mary C. Zanarini, EdD, John G. Gunderson, MD,


Frances R. Frankenburg, MD, and Deborah L. Chauncey, AB

The Diagnostic Interview for Borderlines (DIB) was revised to


sharpen its ability to discriminate between clinically diagnosed
borderline patients and patients with other types of Axis II

diagnoses. The discriminant power of both the revised


clinical
DIB (DIB-R) and the DIB itself was then tested in a sample of
237 inpatients and outpatients given an Axis II diagnosis by
their therapists. The DIB-R was administered blind to clinical

diagnosis, while a DIB score was independently derived from


DIB-R and other data using a predetermined scoring
algorithm. At a cutoff of 8, the DIB-R had a sensitivity of .82,
a specificity of .80, a positive predictive power of .74, and a

negative predictive power of .87. Overall, these conditional


probabilities compare favorably to those obtained for the DIB
at its standard cutoff of 7: sensitivity
=
.97,
specificity =
.27,
positive predictive power
=
.47, and negative predictive power
=
.93. They also compare favorably with those obtained in
studies that used semistructured or self-report instruments
based on DSM-III or DSM-III-R criteria for BPD.

There are currently five diagnostic instruments designed solely to identify


patients with borderline personality disorder (BPD): the Diagnostic In
terview for Borderlines (DIB) (Gunderson, Kolb, St Austin, 1981), the Bor
derline Personality (Perry & Cooper, 1985), the Schedule for
Scale (BPS)
Borderline Personalities (SBP) (which is the second half of the Schedule for

Interviewing Borderlines, or SIB) (Baron & Gruen, 1980), the Borderline


Syndrome Index (BSI) (Conte, Plutchik, Karasu, & Jerrett, 1980), and the
Combined Criteria Instrument (CCI) (Nurnberg, Hurt, Feldman, & Suh,
1987). Although each of these instruments attempts to identify clinically
diagnosed borderline patients using criteria sets that are somewhat differ
ent than that of DSM-III, four different formats are used. The DIB is a

semistructured interview, the BPS and the SBP are structured interviews,
the BSI is a self-report questionnaire, and the CCI is a checklist that is rated
after an unstructured clinical interview.

From the Psychosocial Research Program. McLean Hospital, Belmont, Massachusetts: and the
Department of Psychiatry, Harvard Medical School. Boston, Massachusetts. Address reprint
requests to Dr. M. C. Zanarini. McLean Hospital, 115 Mill Street, Belmont, Massachusetts
02178.

10
DISCRIMINATING BPD 1 1

The DIB is the best known and most influential of these instruments, the
DSM-III criteria for BPD having been based, in part, upon those diagnostic
criteria that Gunderson and his associates found to be both highly
characteristic and most discriminating for borderline patients (Gunderson,
1977; Gunderson & Kolb, 1978). It is also the most extensively used of these
instruments: recent MEDLINE and PSYCINFO searches indicate that over
70 studies have used either the DIB or DIB-derived criteria, alone or in

conjunction with DSM-III criteria for BPD, to define their borderline cohort.

psychometric properties, the DIB has repeatedly been shown


In terms of
to discriminate clinically diagnosed borderline patients from
adequately
those with clinical diagnoses of schizophrenia and major depression
(Koenigsberg, Kemberg, & Schomer, 1983; Kolb & Gunderson, 1980: Soloff
& Ulrich, 1981). In addition, both its interrater (k =
.71-.80) (Cornell, Silk,
Ludolph, & Lohr, 1983: Kroll, Pyle, Zander, Martin, Lari, & Sines, 1981)
and test retest reliability (k
=
.71) (Cornell et al., 1983) have been found to
be in the good to excellent range (Fleiss, 1981). However, its ability to
discriminate clinically diagnosed borderlines from those with other types of
Axis II disorders has been seriously questioned (Kolb & Gunderson, 1980;
Soloff 6k Ulrich, 1981); this is a particularly important concern as more
recent studies have begun to explore the relationship between BPD and
other Axis II disorders, particularly antisocial personality disorder (Gunder
son & Zanarini, in press).
Responding to the growing need of researchers to adequately distinguish
clinically diagnosed borderline patients from those judged to meet criteria
for other Axis II disorders, we revised the DIB in November of 1982 and pilot
tested it for the next 10 months, using the information obtained from 15

conjoint interviews to refine its


format, phrasing, and scoring system. We
then tested its discriminant power using a sample of 237 patients (132

inpatients and 105 outpatients) with a full range of Axis II clinical di

agnoses. This instrument, the Revised Diagnostic Interview for Borderlines


(DIB-R), and the results of this study are described in this paper.

DESCRIPTION OF INSTRUMENT

The DIB-R is a semistructured interview that is divided into four sections

thought to be of clinical
importance in diagnosing BPD: affect, cognition,
impulse patterns, and interpersonal relationships. Overall, it has
action
186 questions that inquire into 108 scored areas. The impulse action
patterns section has the fewest questions (19), the interpersonal rela
tionships section has the most questions (80), and the affect (28) and
cognition (59) sections occupy an intermediate position. The information
obtained from these questions is used to rate 22 summary statements that
describe important features of BPD (see Table 1). The summary statements
in each section are added to yield a section score that can range from a high
of 018 in the interpersonal relationships section to a low of 06 in the

cognition section. These scores are then scaled using an


algorithm provided
at the end of each section to yield a scaled section score of 02 in the affect
and cognition sections and 03 in the impulse action patterns and in

terpersonal relationships sections. These scaled section scores are in turn


12 ZANARINI ET AL.

Table 1. DIB-R Summary Statements

Affect section
1. Chronic/major depression
2. Chronic helplessness/hopelessness/worthlessness/guilt
3. Chronic anger/frequent angry acts
4. Chronic anxiety
5. Chronic loneliness/boredom/emptiness

Cognition section
6. Odd thinking/unusual perceptual experiences
7. Nondelusional paranoid experiences
8. Quasipsychotic experiences

Impulse action patterns section

9. Substance abuse/dependence
10. Sexual deviance
11. Self-mutilation
12. Manipulative suicide efforts
13. Other impulsive patterns
Interpersonal relationships section
14. Intolerance of aloneness
15. Abandonment/engulfment/annihilation concerns

16. Counterdependency/serious conflict over help or care

17. Stormy relationships


18. Dependency/masochism
19. DevaluationAnanipulation/sadism
20. Demandingness/entitlement
21. Treatment regressions
22. Countertransference problems/ "special" treatment relationships

added to yield a total DIB-R score of 010. The entire interview, which is
designed for those 18 years of age or older, typically takes from 45 minutes
to an hour to administer. Raters with substantial clinical experience,
whether they have received formal graduate training or not, can usually be
trained to use the DIB-R effectively. Four conjoint interviews are usually
necessary to familiarize such a rater with the content, format, and scoring
system of the DIB-R.
This instrument differs from the DIB in seven main ways. First, the social
adaptation section of the DIB was not included in the DIB-R because studies
have repeatedly shown that it added little to the discriminant power of the
interview (Gunderson et al., 1981; Soloff 6k Ulrich, 1981). Second, the
psychosis section was renamed the
cognition section and expanded to
encompass various forms of disturbed but nonpsychotic thought. Third,
both the impulse action patterns and interpersonal relationships sections
were given more scaled weight than the affect and cognition sections as a

number of studies have shown that the discriminant power of the former
sections is substantially greater than that of the latter sections (Gunderson
et al., 1981; Soloff 6k Ulrich, 1981). Fourth, all questions in the DIB-R refer
to the past 2 years of a patient's life, whereas the five sections of the DIB had
three different time frames, which ranged from 3 months to 3 years. This
lengthy time frame was selected to avoid
mistakenly assessing state phe
nomena as phenomena. Fifth, three summary statements were placed
trait
in more clinically appropriate sections (i.e., demanding, entitled behavior

being transferred from affect to interpersonal relationships, chronic feel-


DISCRIMINATING BPD 13

ings hopelessness/worthlessness now linked with chronic feelings of


of

helplessness/guilt being transferred from cognition to affect, and treat


ment regressions being transferred from cognition to interpersonal rela

tionships). Sixth, four summary statements were added (i.e., chronic


anxiety, odd thinking/unusual perceptual experiences, quasipsychotic ex
periences, and abandonment/engulfment/annihilation concerns). Seventh,
seven summary statements (in addition to the four from the deleted social
adaptation section) were dropped. One deletion had to do with the reputed
hypersensitivity of borderline patients to certain types of drugs (psychotic
experiences secondary to alcohol/marijuana/psychotomimetic abuse). Two
had to do with dissociative experiences (depersonalization, derealization),
which are now rated in conjunction with odd types of thought. The other
four had to do with affective, cognitive, or interpersonal symptoms or with
features prototypic for schizophrenic or bipolar patients: elation/

hypomania, clear-cut hallucinations or nihilistic/grandiose/bizarre de


lusions, mania or widespread/persistent delusions or hallucinations, and
social isolation/loner.

METHOD

All subjects were either inpatients admitted to McLean Hospital between October
1983 and October 1986 outpatients beginning or currently in treatment during
or

this period in one of three outpatient clinics in the metropolitan Boston area (Revere

Community Counseling Center, East Boston/Winthrop Community Counseling


Center, Bunker Hill Health Center in Charlestown, Massachusetts). Patients were

considered eligible for inclusion if they: (1) were between theages of 18 and 40, (2)
had average or better intelligence, (3) had
no history or current symptomatology of a

major psychotic disorder clear-cut


organic condition, (4) were substance-free for
or

at least 10 days, and (5) were given a definite or probable Axis II diagnosis by the

admitting physician or referring clinician. All subjects who agreed to participate in


the study gave written informed consent and were interviewed blind to clinical

diagnosis by one of us (M.C.Z.) within 2 weeks of their admission to the hospital or


referral to the study. Another of us (D.L.C.) then used a predetermined scoring
algorithm to derive a DIB score for each
subject from DIB-R data, the information
from the deleted social adaptationbeing routinely collected as part of a larger
section

study. The primary Axis II diagnosis given a patient by his or her therapist was then
obtained to be used as the standard against which research diagnoses were mea
sured.

RESULTS

Overall, 95 clinically diagnosed borderline patients and 142 patients carry


ing other Axis II clinical diagnoses were interviewed. The demographic
characteristics of these groups can be seen in Table 2. As expected, the
borderline group had a significantly higher percentage of female subjects
than the other personality disorder (OPD) group. However, the groups did
not differ significantly on the variables of race,
age, or socioeconomic sta
tus, revealing an almost all White and, on
average, relatively young, middle-
class sample.
14 ZANARINI ET AL.

Table 2. Demographic Characteristics of Sample (N =


237)

Characteristic Clinical BPD CI nical OPD"

n 95 142
% female 82. lb 41.6
% White 100.0 97.9
Age
Mean 27.7 27.7
SD 6.8 6.1

Hollingshead Redlich socioeconomic status


(1 highest; 5 lowest)
Mean 3.5 3.5
SD 1.4 1.2

aOPD: other personality disorders.


bX2 =
36.4. d( =
1, p < .001.

Figure 1 shows the distribution of DIB-R scores found in the borderline


and nonborderline groups. As can be seen, borderline subjects were more

likely than nonborderline subjects to obtain a score of 8 or more, while


nonborderline subjects were more likely than borderline
subjects to obtain a
score of 7 or less. In addition, there highly significant difference (p <
was a

.0001) in the DIB-R score obtained by these two


mean
groups of subjects;
BPD subjects obtained mean score of 8.52 1.8 and OPD
a
subjects
obtained a mean score of 5.89 2.2.
Table 3 details the conditional probabilities obtained using the DIB-R.
Sensitivity is defined as the percentage of cases correctly identified as cases,
and specificity is defined as the percentage of noncases correctly identified
as noncases. Positive predictive power is defined as the
prevalence-adjusted
ability of a positive test result to predict the presence of a clinical diagnosis,
and negative predictive power is defined as the prevalence-adjusted ability
of a negative test result to predict the absence of a clinical diagnosis. As can
be seen, a cutoff score of 8 had the best overall probabilities. At this score,

40

35 m

30

25
Q BPD
20
? OPD

15

r
10
s
^r~i
5

0 n,^ +sa. M jm
5 6 10

DIB-R Scores

Figure 1. DIB-R score distribution.


DISCRIMINATING BPD 15

Table 3. Conditional Probabilities of DIB-R

Positive Negative
predictive predictive
DIB-R score
Sensitivity Specificity power power

6 .92 .42 .51 .88


7 .88 .55 .57 .88

8 .82 .80 .74 .87


9 .70 .86 .77 .81
10 .31 .96 .83 .67

about 80% of both the borderline and nonborderline patients were correctly
identified. In addition, the negative predictive power found at this cutoff
was very
high (.87), whereas the positive predictive power was somewhat
lower (.74) but still clearly adequate. Above this cutoff, the instrument tends
to sacrifice sensitivity and negative predictive power, and below this cut
off, the instrument tends to sacrifice specificity and positive predictive
power.
The DIB scores obtained through the application of the previously noted

scoring algorithm very different distribution of scores. At this


revealed a

instrument's standard cutoff of 7 or higher, a sensitivity of .97, a specificity


of .27, a negative predictive power of
positive predictive power of .47, and a

.93 were obtained,


revealing a ability than the DIB-R to
somewhat better

correctly identify clinically diagnosed borderlines and a markedly lower

ability than the DIB-R to correctly identify those given other types of Axis II
clinical diagnoses.
The clinicaldiagnoses of the nonborderline subjects are shown in Table 4.
Forty percent of the nonborderlines were "dramatic' cluster subjects, 31%

Table 4. Nonborderline Clinical Diagnoses

Correctly Incorrectly
identified identified
by DIB-R by DIB-R
Clinical Times

diagnosis diagnosed n % n %

Dramatic cluster 53 47 81 10 17
Antisocial PD 40 33 83 7 17
Histrionic PD 6 5 83 1 17
Narcissistic PD 11 9 82 2 18
Anxious cluster 44 32 73 12 27
Avoidant PD 13 10 77 3 23

Compulsive PD 7 5 71 2 29
Dependent PD 22 15 68 7 32
Passive aggressive PD 2 2 100 0 0
Odd cluster 6 5 83 1 17
Paranoid PD 2 2 100 0 0
Schizoid PD 3 3 100 0 0
Schizotypal PD 1 0 0 1 100
Mixed PD 35 30 86 5 14
16 ZANARINI ET AL.

were subjects, 25% had been given a primary Axis II


"anxious" cluster

diagnosis of mixed
personality disorder, and 4% were "odd" cluster subjects.
The DIB-R was able to correctly identify more than 80% of those given a
clinical diagnosis of a "dramatic" cluster personality disorder, an "odd"
cluster personality disorder, and mixed personality disorder. However, it
was somewhat less successful in correctly identifying "anxious" cluster

subjects, being able to correctly identify a little less than 75% of those given
such a diagnosis. In terms of frequently diagnosed disorders, avoidant and
dependent personality disorders were most likely to yield a false-positive
BPD diagnosis on the DIB-R.
The DIB was far moremisidentify these clinically nonborderline
likely to

subjects; only subjects, 30% of the "anxious"


16% of the "dramatic" cluster
cluster subjects, 34% of the mixed personality disorder subjects, and 67% of
the "odd" cluster subjects were correctly identified as having nonborderline
Axis II disorders. Overall, the DIB-R was more accurate in all categories than
the DIB. However, the greatest improvement was found in the "dramatic"
and mixed categories; the DIB-R was almost 5 times more likely than the
DIB to correctly identify a "dramatic" cluster and mixed personality disorder
subject and somewhat more than 2 times as likely as the DIB to correctly
identify an "anxious" or "odd" cluster subject.

DISCUSSION

The Diagnostic Interview for Borderlines


was revised to sharpen its ability

to discriminate clinically diagnosed borderline patients from patients with


other types of Axis II clinical diagnoses. A naturalistic design was employed
that compared the diagnosis given each subject by his or her therapist to
the overall score obtained by each subject on the DIB-R and the DIB. It was
found that the DIB-R was able to make this difficult discrimination about
80% of the time. The conditional probabilities obtained in this study com

pare favorably with those obtained using structured interviews to diagnose


well-established Axis I disorders, such as major depression and schizophre
nia (Endicott St Spitzer, 1978; Robins, Helzer, Ratcliff, St Seyfried, 1982).
They also compare favorably with those obtained using three instruments
that generateDSM-III or DSM-III-R BPD diagnoses: two semistructured
interviews the Personality Disorder Examination (PDE) and the Struc
tured Interview for DSM-III (SCID) (Skodol, Rosnick, Kellman, Oldham, St

Hyler, 1988) and


self-report instrument the Personality Diagnostic
a

Questionnaire (PDQ) (Hurt, Hyler, Frances, Clarkin, 6k Brent, 1984). These


latter results are particularly impressive as the current study used di

agnoses provided by over 50 therapists as its standard of comparison,


whereas these earlier studies used either diagnoses made by one in

terdisciplinary inpatient team or diagnoses generated by a team of research


ers after conducting a somewhat standardized clinical interview that

covered all of the Axis II criteria in DSM-III.


It was also found that the DIB is too overinclusive in diagnosing BPD, a
finding that confirms the reservations expressed by researchers familiar
with the of the DIB (Kolb 6k Gunderson, 1980;
psychometric properties
Soloff 6k Ulrich, 1981). It also confirms the results of an earlier study that
DISCRIMINATING BPD 17

found that a clinically diagnosed nonborderline per


substantial number of
sonality-disordered patients DIB-positive for BPD (Barrash, Kroll,
were

Carey, 6k Sines, 1983). However, this finding contradicts the relatively good
conditional probabilities obtained in several other studies that assessed the
ability of the DIB to discriminate clinically diagnosed borderline patients
from Axis II controls. The explanation for this discrepancy may lie in the fact
that these latter studies either deliberately excluded near-neighbor Axis II
controls (e.g., antisocial patients, narcissistic patients) (Koeningsberg et
al., 1983) or artificially sharpened the distinction between these two groups
of patients by using diagnoses provided by a team of researchers as their
clinical standard (Frances, Clarkin, Gilmore, Hurt, 6k Brown, 1984).

Perhaps the most surprising finding of this study was that "anxious"
cluster patients, who theoretically share few clinical features with border
line patients, were more commonly misdiagnosed as having BPD than
"dramatic" cluster patients, who theoretically share many clinical features
with borderline patients. However, fully two-thirds (8 of 12) of these in

correctly identified patients were inpatients, which suggests that these


patients were considerably more disturbed than most patients who carry
the diagnosis of avoidant, compulsive, or dependent personality disorder. It
may be that these patients were suffering from the sequelae of a state
disorder, such as major depression or cocaine abuse, that the DIB-R mis
takenly identified as BPD. In contrast, the therapists involved may have
been responding to certain particularly pronounced character traits, such
as a strong fear of intimacy or extreme dependency, that led them to
overlook the more obvious borderline diagnosis. It may also be that the
DSM-III guidelines for making these diagnoses need to convey more dis
tinctiveness.
The interrater and testretest reliability of the DIB-R will be described in a
future report. Other future reports will detail the clinical features assessed
by the DIB-R that best discriminate borderline patients from Axis II controls
and the clinical features assessed by the DIB-R, if any, that differentiate
inpatient and outpatient borderlines. Research in the area of borderline
psychopathology is turning more toward the differentiation of BPD from
other Axis II disorders. The DIB-R provides researchers with a reliable
means of identifying a borderline cohort that includes relatively few Axis II
controls, thus facilitating research into the etiology, course, and outcome of
BPD.

REFERENCES

Baron, M., & Gruen, R. (1980). The sched liminary norms. Journal of Nervous and
ule for interviewing borderlines. New Mental Disease. 168. 428-^35.
York: New York State Psychiatric In Cornell, D. G., Silk, K. R., Ludolph, P. S., &
stitute. Lohr, N. E. (1983). Test-retest reliability
Barrash, J., Kroll, J., Carey, K., & Sines, L. of the diagnostic interview for border
(1983). Discriminating borderline dis lines. Archives of General Psychiatry, 40,
order from other personality disorders: 1307-1310.
Cluster analysis of the diagnostic in Endicott, J., & Spitzer, R. L. (1978). A di
terview for borderlines. Archives of agnostic interview: The schedule for
General Psychiatry. 40. 1297-1302. affective disorders and schizophrenia.
Conte, H. R., Plutchik. R., Karasu, T. B., & Archives of General Psychiatry, 35, 837-
Jerrett, I. (1980). A self- report borderline 862.
scale: Discriminative validity and pre Fleiss, J. L. (1981). Statistical methodsfor
18 ZANARINI ET AL.

rates and proportions (2nd ed.). New Archives of General Psychiatry, 40. 49-
York: John Wiley & Sons. 53.
Frances, A., Clarkin, J. F., Gilmore, M., Kolb. J. E.. & Gunderson. J. G. (1980). Di
Hurt, S. W., & Brown, R. (1984). Reliabil agnosing borderline patients with a semi-

ity of criteria for borderline personality structured interview. Archives of General


disorder: A comparison of DSM-III and Psychiatry, 37, 37-41.
the diagnostic interview for borderline Pyle, R.. Zander, J..
Kroll, J.. Martin. K..
patients. American Journal of Psychi Lari. S., & Sines, L. (1981). Borderline
atry. 141, 1080-1083. personality disorder: Interrater reliability
Gunderson, J. G. (1977). Characteristics of of the diagnostic interview for border
borderlines. In P. Hartocollis (Ed.), Bor lines. Schizophrenia Bulletin. 7. 269-
derline personality disorders: The con 272.

cept, the syndrome, the patient (pp. 173 Nurnberg, H. G.. Hurt, S. W., Feldman, A..

192). New York: International Univer & Suh, R. (1987). Efficient diagnosis of
sities Press. borderline personality disorder. Journal
Gunderson, J. G.. & Kolb, J. E. (1978). Dis of Personality Disorders. 1. 307-315.
criminating features of borderline Perry. J. C. & Cooper, S. H. (1985). Psy-
patients. American Journal of Psychi chodynamics, symptoms, and outcome in
atry, 135. 792-796. borderline and antisocial personality dis
Gunderson, J. G., Kolb, J. E., & Austin, V. orders and bipolar type II affective dis
(1981). The diagnostic interview for bor order. InT. H. McGlashan, (Ed.). The bor
derlines. American Journal of Psychi derline: Current empirical research (pp.
atry. 138. 896-903. 19-41). Washington, DC: American Psy
Gunderson, J. G., & Zanarini, M. C. (in chiatric Press.

press). Pathogenesis of borderline per Robins. L. N., Helzer. J. E.. Ratcliff. K. S., &
sonality. In A. Tasman, R. Hales, & A. Seyfried, W. (1982). Validity of the di
Frances (Eds.), American
Psychiatric agnostic interview schedule, version II:
Press review of psychiatry (Vol. 8). DSM-III diagnoses. Psychological Medi

Washington, DC: American Psychiatric cine. 12. 855-870.


Press. Skodol, A. E., Rosnick, L., Kellman, D., Old
Hurt. S. W., Hyler S. E.. Frances, A., Clark ham. J. M.. & Hyler, S. E. (1988). The
in, J. F.. & Brent, R. (1984). Assessing validity of structured assessments of
borderline personality disorder with self- Axis II. Paper presented at the 141st an

report, clinical interview, or semistruc nual meeting of the American Psychiatric


tured interview. American Journal of Association, Montreal, Canada.
Psychiatry, 141, 1228-1231. Soloff, P. H.,&Ulrich, R. F. (1981). Diagnos
Koenigsberg. H. W..Kemberg, O. F.. & tic interview for borderline patients: A
Schomer, J. (1983). Diagnosing border replication study. Archives of General
line conditions in an outpatient setting. Psychiatry. 38, 686-692.

Vous aimerez peut-être aussi