Structured
Interviews
J. Kavoussi,
Richard
for Borderline
Personality
M.D., Emil F. Coccaro, M.D., Howard M. Klar, Bernstein, Ph.D., and Larry J. Siever, M.D.
David
denlines
The authors compared three instruments used to diagnose borderline personality disorder-the Diagnostic Interview for Borderline Patients (DIB), the Schedule for Interviewing Borderlines, and the Structured Interview for DSM-III Personality Disorders-in 56 patients with personality disorders. A borderline diagnosis was made according to the DIB in 30%, the Structured Interview for DSM-III Personality Disorders in 48%, and the Schedule for Interviewing Borderlines in 55% of the patients. Diagnostic agreement was only 52%, which has implications for the generalizability of results of validation studies of the borderline diagnosis. Improvement in diagnostic agreement requires modification of current criteria sets and/or the use of dimensional models.
(AmJ
Psychiatry
1990;
147:1522-1525)
B
ondenline personality disorder nosed in clinical settings and most studied personality disorders nomenology, biological markers, family history, this diagnosis
cause
of
and has
inconclusive
across studies. sets are available Knight
outcome. not been
(1),
or
A major problem for the diagnosis,
Kennberg
(2),
is commonly diaghas been one of the in terms of phetreatment response,
Unfortunately, adequately
findings
Gninken
Disorder
however, validated be-
discordant
findings
is that many such as et al.
(3),
criteria those of
(8),
also
based
on
Each of these instruments terrater and test-retest condance
of
making
it
using
the
these
to
different
DIB) na
those
(using
evaluated,
results
of
and
studies
entry
concordance
diagnosed
et al.
good inthe con-
but
been
the
personality
clinicians’
Spitzer
been no studies theme have been
to Gundenson
and
not
to establish
diagnostic
borderline
according
has
measures
the with
of
demonstrated (9-14),
compare
Although there have these structured interviews, tients
criteria
has reliability
interviews
difficult
vestigating
the
M.D.,
criteria.
comparing studies inbetween
disorder
Singer’s
criteria
according
pa-
diagnosed (using
the
to DSM-III
judgments).
The
degree
cniteof
agree-
ment has varied in previous studies (14-20), with kappa values ranging from 0.45 to 0.83. One study (21) compared the borderline diagnoses made by using the DIB with those made by using the Personality Disorder Examination (22), a structured interview based on DSM-III criteria, and found poor concordance between the two instruments in a sample of 22 patients clinically
diagnosed
as
having
borderline
personality
disorder. Given that the DIB, the Structured Interview for DSM-III Personality Disorders, and the Schedule for Interviewing Bonderlines are widely used in studies of borderline personality disorder, the purpose of the present among
study these
was three
of borderline
to measure instruments
personality
the
degree in making
of
agreement a diagnosis
disorder.
Gunderson
and
Singer (4), and Spitzen et al. (5). One solution to this dilemma is to develop operationalized criteria in the form of structured interviews to increase diagnostic reliability from setting to setting. Various semistructuned interviews have been developed to correlate with the most widely used criteria sets. These include the denline Patients (DIB) Singer’s
criteria;
Personality
Spitzer
the
Structured
Disorders
et al.;
and
Diagnostic (6), based
the
Interview on Gunderson Interview
for
for
METHOD The who
Bor-
Presented at the 142nd annual meeting of the American Psychiatnc Association, San Francisco, May 6-1 1, 1989. Received July 6, 1989; revision received April 23, 1990; accepted May 10, 1990. From the Department of Psychiatry, Mount Sinai School of Medicine, New York. Address reprint requests to Dr. Kavoussi, Depression Evaluation Service, New York State Psychiatric Institute, 722 West 168th St., New York, NY 10032. Copyright © 1990 American Psychiatric Association.
1522
of
study
male
they
had
patients
referred
18-65
years
the
psychiatric
from
Administration
center patients
disorders. if
DSM-III mental on drug using
56
a Veterans
or another medical tion program for personality
of
were
consecutively
services
Bonand
DSM-III
(7), based on the criteria Schedule for Interviewing
subjects were
Patients major
medical
for inclusion with clinical were
from
illnesses
or
axis I diagnoses of schizophrenia, disorder, bipolar disorder, or current dependence. Axis I diagnoses were the
phrenia at the
Schedule
for
Affective
Disorders
(SADS) (23). All patients time of the study and gave
center
in an evaluadiagnoses of
excluded
medical
old
and
the major
organic alcohol made by Schizo-
were medication informed consent
free to
participate. Each
fenent
interview
set
of
was
raters
Am
conducted
conduced
J Psychiatry
by
each
147:11,
two
raters.
interview
November
A dif-
for
any
1990
KAVOUSSI,
given patient and were blind to data obtained from the other instruments. The interviews were conducted in random order for any given patient, and 2-7 days separated each interview. Intemraten reliability (kappa) was high for the diagnosis of borderline personality disorder for each of the three instruments (DIB, -0.79;
Structured
Interview
for
DSM-III
Personality
Disorders, -0.78; Schedule for Interviewing Bonderlines, -0.81). Structured Interview for DSM-III Pensonality
Disorders
informant
interviews
were
con-
ducted for 44 (79%) of the patients; these patients granted permission for such interviews. Disagreements on final scores or diagnoses derived from any specific instrument were resolved by consensus at a meeting of all members of the rating team supervised by a senior clinician (H.M.K.) For a categorical diagnosis of borderline personality disorder, patients were required to meet four on more of the eight criteria for borderline personality disorder on the Structured Interview for DSM-lII Personality Disorders, score 7 or more on the DIB, or score 23 or more on the summed severity scales of the schedule for borderline personality portion of the Schedule for Intenviewing
Borderlines.
To
determine
categorical
agree-
COCCARO,
KLAR,
ET AL.
Schizotypal (27 [48.2%] of the patients) and histrionic (27 [48.2%] of the patients) personality disorders were the most common, followed by paranoid (19 [33.9%] of the patients), compulsive (17 [30.4%]), antisocial (nine [16.1%]), and narcissistic (six [10.7%]) personality disorders. Other personality disorders were found in less than 5% of the patients. The three instruments agreed on the presence or absence of borderline personality disorder in 29 (51.8%) of the patients. Thirty-nine (69.6%) of the patients were diagnosed as having borderline personality disorder by at least one instrument; 24 (42.9%) were diagnosed as having borderline personality disorder by at least two instruments, and 12 (21.4%) were so diagnosed by all three. Seventeen (30.4%) of the patients were not diagnosed as having borderline personality disorder by any of the instruments. The DIB diagnosed borderline personality disorder in
17
(30.4%)
of
the
patients,
the
Structured
Inter-
view for DSM-III Personality Disorders in 27 (48.2%), and the Schedule for Interviewing Bordenlines in 31 (55.4%). Fourteen (82.4%) of the 17 patients diagnosed as having borderline personality disorder by the
DIB were
also
diagnosed
as having
the disorder
by the
ment on the overall diagnosis of bordeniine personality disorder and agreement between subscales common to any two instruments, two-by-two kappa values were computed by using the formula K=(pOBS-pCHA)/
Structured Interview dens, and 22 (81.5%) having the disorder
for DSM-III Personality Disorof the 27 patients diagnosed as by the Structured Interview for
DSM-III
Disorders
1-pCHA),
the
where
pCHA=chance
agreement
and
p OBS = observed agreement. Pearson correlations were made by comparing for each subject the number of criteria met on the Structured Interview for DSM-III Personality Disorders, the DIB scaled score, and the total schedule for borderline personality severity scone of the Schedule for Interviewing Bordemlines.
RESULTS The mean age of the patients was 37.7, and all were men. Twenty-seven (48.2%) of the patients were separated or divorced, 18 (32.1%) were currently manned, and I 1 (19.6%) had never been married. The racial composition reflected the general clinic population: 29 (51.8%) were white, 17 (30.4%) were Hispanic, and 10 (17.9%) were black. Affective disorders were common in the history of the 56 men in the study group: 24 (42.9%) had a diagnosis of major depressive disorder in remission. Also common were past substance abuse (17 [30.4%] of the patients) and anxiety disorders (12 [21.4%] of the patients). These diagnoses were not significantly more common in the patients with borderline personality disorder as diagnosed by any one of the instruments under study; however, the small sample sizes limited the power of these analyses. There was significant overlap with other personality disorder diagnoses made by using the Structured Interview for DSM-III Personality Disorders: 46 (82A%) of the patients had more than one personality disorder.
Am
J Psychiatry
147:1
1, November
1990
Personality
were
so diagnosed
by
Schedule for Interviewing Bordenlines. The degree of concordance (kappa) for the diagnosis of borderline personality disorder was 0.50 between the Structured Interview for DSM-III Personality Disorders and the Schedule for Interviewing Bordenlines, 0.42 between the DIB and the Structured Interview for
DSM-III
Personality
Disorders,
and only
0.18
between
the D1B and the Schedule for Interviewing Bordenlines. However, if the cutoff for a diagnosis of borderline personality disorder on the DIB was lowered to a score of 6, concordance was improved (for the Structured Interview for DSM-lII Personality Disorders and the
DIB,
0.47;
for the DIB and
the Schedule
for Interview-
ing Bordenlines, 0.73). Pearson’s correlations were r=0.65, df=54, p