The Effects of Changing Roger Blashfield,
Axis II Diagnostic
Criteria
Nancee Blum, and Bruce F’fohl
This study examined whether changes in diagnostic criteria from the DSM-III to the DSM-III-R for personality disorders (PDs) had the intended effects. Seventytwo subjects at the University of Iowa from three research studies and one clinical sample were administered two structured interviews (the Structured Interview for DSM-III Personality [SIDP] and the revised SIDP [SIDP-R]) to assess DSM-III and DSM-III-R criteria. Major changes in rates of diagnoses were observed between the DSM-III and DSM-III-R criteria with kappasfor agreement ranging between -.025 and .571. As expected, the switch from monothetic to polythetic
definitions had an effect on which patients were assigned a given diagnosis. However, not all of the other revisions associated with the DSM-III-R had the intended effects. For instance, the frequency of the diagnosis of schizoid PD did not increase, nor did the overlap between borderline and histrionic PDs decrease. In addition, there was an unintended increase in the rate of paranoid PD. An analysis of individual criteria showed how small, apparently minor changes in the wording of criteria can sometimes have major effects on which patients received a diagnosis of PD. Copyright 0 1992 by W. B. Saunders Company
EN THE DSM-III was revised to form the DSM-III-R, most of the changes concerned the diagnostic criteria used to define the various mental disorders. Widiger et al.] discussed the changes in the DSM-III-R definitions of the personality disorders (PDs). For instance, the definition of schizoid PD was changed from three criteria with a monothetic form (i.e., all three were required) in the DSMIII to seven criteria with a polythetic form (four of seven required) in the DSM-III-R. In contrast, schizotypal PD had a polythetic definition in both classifications, and about half of the criteria were the same across the two systems. In the early 1980s Pfohl and colleagues? developed the Structured Interview for DSMIII Personality (SIDP), a semistructured interview to assess the personality disorders. The SIDP contained 18.5 probe questions to elicit information for scoring the 96 DSM-III diagnostic criteria for the 11 PDs. After the publication of the DSM-III-R, the SIDP was revised to the SIDP-R.3 The SIDP-R maintained a design and structure similar to the SIDP. Many of the probe questions were identical across the two interviews. The SIDP-R contains 189 probe questions to elicit information about the 120 DSM-III-R PD criteria. The current study compares the SIDP versus SIDP-R results on 72 individuals who were administered both semistructured interviews. Since the results of structured interviews to assess the DSM-III and the DSM-III-R criteria are available on all 72 subjects, it is possible to analyze the intended effect of changes made in diagnostic criteria by the DSM-III-R committee. For instance, Widiger et al.’ argued that the
change in the definition of schizoid was intended to increase the frequency with which this diagnosis is made, since patients rarely met the DSM-III definition of schizoid. The data from this study can indicate whether the revised definition of schizoid increased the frequency of this diagnosis.
W
Comprehensive
METHOD
Subjects Seventy-two subjects over 18 years of age were selected from three research studies at the University of Iowa and from inpatients in the University of Iowa Psychiatric Hospital. The patients, who were chosen from previous research studies, were selected primarily because of proximity to the University of Iowa so that performing a second interview would be feasible. Fifty-two subjects previously participated in earlier research studies at the University of lowa.4-” For these 52 individuals, the SIDP interviews were administered first. with a median time interval of 22 weeks (interquartile range, 60 weeks). Of these 52 patients, 22 had an axis I diagnosis of obsessive-compulsive disorder, another 20 patients had an axis I diagnosis of generalized anxiety disorder, and IO individuals had served as normal control subjects in a study of obsessive-compulsive disorder. The final 20 subjects were nonpsychotic, psychiatric inpatients on whom a consultation about a PD diagnosis was requested. Sixteen of the 20 inpatients had an axis I diagnosis of major depressive disorder. The reason for selecting individuals with different axis I disorders was to insure that the results would not be limited
From the Department of Psychiatry, University of Florida, Gainesville. FL: and the Department of Pvchiatty, Unil*ersir?, of Iowa, Iowa City, IA Address reprint requests to Roger Blashfield, Ph.D., Department of Psychiatry. University of Iowa, 500 Newton Rd, Iowa City, IA 52242. Copyright 0 1992 by W.B. Saunders Cornpar!,: OOIO-440X19213304-OOlI$O3,OOlO
Psychiatry, Vol. 33, No. 4 (July/August), 1992: pp 245-252
245
246
to patients with one type of axis I psychopathology. The use of a more heterogeneous sample should provide more generalizable results than would be possible if the subjects had the same axis I diagnosis.
BLASHFIELD, BLUM. AND PFOHL
diagnostic criteria: the subject’s responses to the questions, the informant’s responses to a subset probe questions, and any information available patient’s chart. All diagnostic criteria were scored sent), 1 (present), or 2 (severely present).
probe of the in the 0 (ab-
Procedure The 52 subjects, who had been previously administered a SIDP, were called by one of the authors (N.B.) and asked to participate in the study. If the subject resided within 25 miles of the Universi~ of Iowa, the SIDP-R was administered to the patient in a face-to-face interview by N.B. For 20 of the research subjects who lived outside that radius, the SIDP-R was administered by phone. All subjects were asked to suggest a knowledgeable informant who was familiar with them and who would be willing to answer questions about them. All informants were interviewed by phone. For the 20 inpatients, the SIDP-R was usually administered first (n = 15). because the consultation was being requested in order to assign a DSM-III-R diagnosis to the patients. Seventeen of the 20 repeated interviews were performed while the patient was still hospitalized. The remaining three were performed over the phone. The time between the two interviews was usually 1 to 3 weeks. All informants of the inpatients were interviewed by phone. One author (N.B.) administered both interviews. In an attempt to reduce any diagnostic bias associated with the memory of the interviewer, neither the SIDP-R nor the SIDP were scored until both interviews were completed.
scoring The SIDP and SIDP-R are semistructured. Both use probe questions to elicit information. However, clinicians must consider all available information, as well as their clinical judgment, when determiningwhetheror not diagnostic criteria are present. In parti~uIar, the subject’s selfreport is not the only basis for making a decision about a diagnostic criterion. For instance, one subject responded “yes” when asked the probe question: “Do you keep changing your mind about who you want for friends or what career you want?” The subject was then asked for an example, to which she said, that after 22 years of being an elementary school teacher, she was thinking of managing a restaurant instead. When the subject’s husband was asked the same question, i.e., “Does your wife keep changing her mind.. . ?,” he responded with “no,” but said that recently his wife had been bored with teaching school. The DSM-III diagnostic criterion associated with this probe question is the borderline criterion “identity disturbance manifested by uncertainty about several issues relating to identity, such as self-image, gender identity, long-term goals or loyalties, e.g., ‘Who am I?,’ ‘I feel like I am my sister when I am good.’ ” Even though the subject answered “yes” to one of the probe questions associated with this criterion, the criterion would not be scored as present because the subject’s recent consideration of a career change does not appear to reflect a chronic and troubling concern about self-identity. For both the SIDP and SIDP-R, three sources of information were combined by the interviewer when scoring the
RESULTS Subjects The subjects ranged in age from 21 to 65 years. The mean age was 41.8 years (SD 12.4). There were 20 men and 52 women in the study. Most of the subjects were married (n = 46; 64%); 11 were single (15%); 14 were divorced or separated (20%); and one was widowed. The subjects constituted a relatively educated sample, with a mean of 14.1 years of formal schooling (SD 2.9). All subjects were white.
Table 1 displays the concordant between the DSM-III PD diagnoses using the SIDP with the DSM-III-R diagnoses using the SIDP-R. The DSM-III diagnoses are shown as columns in this table. The first number in each column refers to the total number of subjects who met the DSM-III criteria for a particular personality disorder. The DSM-III-R diagnoses are shown as rows in Table 1. The first number on the left side of each row refers to the total number of subjects who met a particular DSM-III-R diagnosis. The numbers in parentheses along the main diagonal are the number of subjects diagnosed with the same disorder on both classifications. As might be expected, the disagreements between the DSM-III and DSM-III-R diagnoses were not haphazard. First, somewhat more PD diagnoses were assigned using the DSM-III-R, rather than the DSM-III, even though the SIDP was administered first when the patient was more likely to manifest symptoms of an axis I disorder. In addition, more of the SIDP diagnoses were confirmed by the SIDP-R than the reverse. That is, 60% of the SIDP/DSM-III diagnoses were confirmed by the DSM-III-R. In contrast, only 48% of the SIDP-R/DSM-III-R diagnoses had been made earlier using the SIDP. Table 2 displays the estimates of agreement between the DSM-III and DSM-III-R diagnoses. The first column shows the number of
247
CHANGING AXIS II
Table 1. Diagnostic Concordance Between SIDP/DSM-III
and SIDP-WDSM-III-R
SIDP Par
Szd
B0r
His
N.V
AW
1
stv 8
Ant
2
2
16
15
4
9
DIP 16
Corn
P.3S
Mix
11
19
7
SIDP-R Par 16
(2)
1
5
1
9
a
3
3
4
6
7
0
Szd 2
1
(1)
1
1
1
1
0
0
1
1
1
0
sty 1
0
0
(0)
1
1
1
0
0
1
1
1
0
Ant 2
0
0
0
(2)
2
1
0
0
1
1
1
0
Bor 7
0
0
1
2
(7)
3
1
1
1
2
3
0
His 16
1
0
1
2
11
(101
4
2
5
5
7
1
Nar 6
1
0
1
1
5
4
(31
2
3
2
5
0
Avo 24
0
0
5
1
10
7
2
(8)
10
a
12
1
Dep 12
0
0
0
1
5
5
3
3
(8)
6
0
Obc 20
2
1
5
1
7
5
1
5
7
(9)
a a
Pas 17
1
0
1
1
a
7
3
4
6
7
(10)
1
Sad 1
1
0
0
0
1
0
1
0
0
1
1
0
a
0
0
1
1
6
3
2
0
2
2
3
0
Mix 4
0
0
0
0
2
2
1
1
1
0
2
(0)
.sd
diagnostic criteria that were virtually the same in the two classifications divided by the total number of DSM-III-R criteria. Column 2 lists those disorders that were given a monothetic definition in the DSM-III and polythetic definition in the DSM-III-R. The third column lists the percentage of patients assigned to each DSM-III diagnosis. The percentages for the DSM-III-R are shown in column 4. The fifth column contains the kappa estimates of agreement between the SIDP/DSM-III and the SIDP-
1
R/DSM-III-R diagnoses. The sixth column contains the Pearson product moment correlations between the number of positive criteria per diagnosis for the two classifications. The seventh and eighth columns report the kappa and correlation values for the n = 28 patients who were administered the SIDP and SIDP-R within 1 month of each other. For this reduced sample, changes in clinical status are less likely to affect the estimates of diagnostic concordance. Concerning diagnostic agreement between
Table 2. Estimates of Agreement Between DSM-III and DSM-III-R Diagnoses No.
Entire Sample
of
Criteria in Common*
Monothetic Polythetict
to
% DSM-III$
l-Month
(n = 72)I
% DSM-III-RI
Sample (n = 28)//
K
r
K
,783
,169 (NS)
r
Paranoid
317
No
3
22
Schizoid
217
Yes
1
3
Schizotypal
419
No
11
1
Antisocial
14122
No
3
3
Borderline
518
No
22
10
Histrionic
118
No
21
22
Narcissistic
519
No
6
a
Avoidant
317
Yes
13
33
,370
,743
,410
,853
Dependent
119
Yes
22
17
,471
,622
,340
,577
Ob-Comp
419
No
15
28
,478
,696
,462
,808
Pass-Aggr
3/9
No
26
24
,408
,550
,650
,770
,182 (NS) n
Axis II Diagnostic
Criteria
Nancee Blum, and Bruce F’fohl
This study examined whether changes in diagnostic criteria from the DSM-III to the DSM-III-R for personality disorders (PDs) had the intended effects. Seventytwo subjects at the University of Iowa from three research studies and one clinical sample were administered two structured interviews (the Structured Interview for DSM-III Personality [SIDP] and the revised SIDP [SIDP-R]) to assess DSM-III and DSM-III-R criteria. Major changes in rates of diagnoses were observed between the DSM-III and DSM-III-R criteria with kappasfor agreement ranging between -.025 and .571. As expected, the switch from monothetic to polythetic
definitions had an effect on which patients were assigned a given diagnosis. However, not all of the other revisions associated with the DSM-III-R had the intended effects. For instance, the frequency of the diagnosis of schizoid PD did not increase, nor did the overlap between borderline and histrionic PDs decrease. In addition, there was an unintended increase in the rate of paranoid PD. An analysis of individual criteria showed how small, apparently minor changes in the wording of criteria can sometimes have major effects on which patients received a diagnosis of PD. Copyright 0 1992 by W. B. Saunders Company
EN THE DSM-III was revised to form the DSM-III-R, most of the changes concerned the diagnostic criteria used to define the various mental disorders. Widiger et al.] discussed the changes in the DSM-III-R definitions of the personality disorders (PDs). For instance, the definition of schizoid PD was changed from three criteria with a monothetic form (i.e., all three were required) in the DSMIII to seven criteria with a polythetic form (four of seven required) in the DSM-III-R. In contrast, schizotypal PD had a polythetic definition in both classifications, and about half of the criteria were the same across the two systems. In the early 1980s Pfohl and colleagues? developed the Structured Interview for DSMIII Personality (SIDP), a semistructured interview to assess the personality disorders. The SIDP contained 18.5 probe questions to elicit information for scoring the 96 DSM-III diagnostic criteria for the 11 PDs. After the publication of the DSM-III-R, the SIDP was revised to the SIDP-R.3 The SIDP-R maintained a design and structure similar to the SIDP. Many of the probe questions were identical across the two interviews. The SIDP-R contains 189 probe questions to elicit information about the 120 DSM-III-R PD criteria. The current study compares the SIDP versus SIDP-R results on 72 individuals who were administered both semistructured interviews. Since the results of structured interviews to assess the DSM-III and the DSM-III-R criteria are available on all 72 subjects, it is possible to analyze the intended effect of changes made in diagnostic criteria by the DSM-III-R committee. For instance, Widiger et al.’ argued that the
change in the definition of schizoid was intended to increase the frequency with which this diagnosis is made, since patients rarely met the DSM-III definition of schizoid. The data from this study can indicate whether the revised definition of schizoid increased the frequency of this diagnosis.
W
Comprehensive
METHOD
Subjects Seventy-two subjects over 18 years of age were selected from three research studies at the University of Iowa and from inpatients in the University of Iowa Psychiatric Hospital. The patients, who were chosen from previous research studies, were selected primarily because of proximity to the University of Iowa so that performing a second interview would be feasible. Fifty-two subjects previously participated in earlier research studies at the University of lowa.4-” For these 52 individuals, the SIDP interviews were administered first. with a median time interval of 22 weeks (interquartile range, 60 weeks). Of these 52 patients, 22 had an axis I diagnosis of obsessive-compulsive disorder, another 20 patients had an axis I diagnosis of generalized anxiety disorder, and IO individuals had served as normal control subjects in a study of obsessive-compulsive disorder. The final 20 subjects were nonpsychotic, psychiatric inpatients on whom a consultation about a PD diagnosis was requested. Sixteen of the 20 inpatients had an axis I diagnosis of major depressive disorder. The reason for selecting individuals with different axis I disorders was to insure that the results would not be limited
From the Department of Psychiatry, University of Florida, Gainesville. FL: and the Department of Pvchiatty, Unil*ersir?, of Iowa, Iowa City, IA Address reprint requests to Roger Blashfield, Ph.D., Department of Psychiatry. University of Iowa, 500 Newton Rd, Iowa City, IA 52242. Copyright 0 1992 by W.B. Saunders Cornpar!,: OOIO-440X19213304-OOlI$O3,OOlO
Psychiatry, Vol. 33, No. 4 (July/August), 1992: pp 245-252
245
246
to patients with one type of axis I psychopathology. The use of a more heterogeneous sample should provide more generalizable results than would be possible if the subjects had the same axis I diagnosis.
BLASHFIELD, BLUM. AND PFOHL
diagnostic criteria: the subject’s responses to the questions, the informant’s responses to a subset probe questions, and any information available patient’s chart. All diagnostic criteria were scored sent), 1 (present), or 2 (severely present).
probe of the in the 0 (ab-
Procedure The 52 subjects, who had been previously administered a SIDP, were called by one of the authors (N.B.) and asked to participate in the study. If the subject resided within 25 miles of the Universi~ of Iowa, the SIDP-R was administered to the patient in a face-to-face interview by N.B. For 20 of the research subjects who lived outside that radius, the SIDP-R was administered by phone. All subjects were asked to suggest a knowledgeable informant who was familiar with them and who would be willing to answer questions about them. All informants were interviewed by phone. For the 20 inpatients, the SIDP-R was usually administered first (n = 15). because the consultation was being requested in order to assign a DSM-III-R diagnosis to the patients. Seventeen of the 20 repeated interviews were performed while the patient was still hospitalized. The remaining three were performed over the phone. The time between the two interviews was usually 1 to 3 weeks. All informants of the inpatients were interviewed by phone. One author (N.B.) administered both interviews. In an attempt to reduce any diagnostic bias associated with the memory of the interviewer, neither the SIDP-R nor the SIDP were scored until both interviews were completed.
scoring The SIDP and SIDP-R are semistructured. Both use probe questions to elicit information. However, clinicians must consider all available information, as well as their clinical judgment, when determiningwhetheror not diagnostic criteria are present. In parti~uIar, the subject’s selfreport is not the only basis for making a decision about a diagnostic criterion. For instance, one subject responded “yes” when asked the probe question: “Do you keep changing your mind about who you want for friends or what career you want?” The subject was then asked for an example, to which she said, that after 22 years of being an elementary school teacher, she was thinking of managing a restaurant instead. When the subject’s husband was asked the same question, i.e., “Does your wife keep changing her mind.. . ?,” he responded with “no,” but said that recently his wife had been bored with teaching school. The DSM-III diagnostic criterion associated with this probe question is the borderline criterion “identity disturbance manifested by uncertainty about several issues relating to identity, such as self-image, gender identity, long-term goals or loyalties, e.g., ‘Who am I?,’ ‘I feel like I am my sister when I am good.’ ” Even though the subject answered “yes” to one of the probe questions associated with this criterion, the criterion would not be scored as present because the subject’s recent consideration of a career change does not appear to reflect a chronic and troubling concern about self-identity. For both the SIDP and SIDP-R, three sources of information were combined by the interviewer when scoring the
RESULTS Subjects The subjects ranged in age from 21 to 65 years. The mean age was 41.8 years (SD 12.4). There were 20 men and 52 women in the study. Most of the subjects were married (n = 46; 64%); 11 were single (15%); 14 were divorced or separated (20%); and one was widowed. The subjects constituted a relatively educated sample, with a mean of 14.1 years of formal schooling (SD 2.9). All subjects were white.
Table 1 displays the concordant between the DSM-III PD diagnoses using the SIDP with the DSM-III-R diagnoses using the SIDP-R. The DSM-III diagnoses are shown as columns in this table. The first number in each column refers to the total number of subjects who met the DSM-III criteria for a particular personality disorder. The DSM-III-R diagnoses are shown as rows in Table 1. The first number on the left side of each row refers to the total number of subjects who met a particular DSM-III-R diagnosis. The numbers in parentheses along the main diagonal are the number of subjects diagnosed with the same disorder on both classifications. As might be expected, the disagreements between the DSM-III and DSM-III-R diagnoses were not haphazard. First, somewhat more PD diagnoses were assigned using the DSM-III-R, rather than the DSM-III, even though the SIDP was administered first when the patient was more likely to manifest symptoms of an axis I disorder. In addition, more of the SIDP diagnoses were confirmed by the SIDP-R than the reverse. That is, 60% of the SIDP/DSM-III diagnoses were confirmed by the DSM-III-R. In contrast, only 48% of the SIDP-R/DSM-III-R diagnoses had been made earlier using the SIDP. Table 2 displays the estimates of agreement between the DSM-III and DSM-III-R diagnoses. The first column shows the number of
247
CHANGING AXIS II
Table 1. Diagnostic Concordance Between SIDP/DSM-III
and SIDP-WDSM-III-R
SIDP Par
Szd
B0r
His
N.V
AW
1
stv 8
Ant
2
2
16
15
4
9
DIP 16
Corn
P.3S
Mix
11
19
7
SIDP-R Par 16
(2)
1
5
1
9
a
3
3
4
6
7
0
Szd 2
1
(1)
1
1
1
1
0
0
1
1
1
0
sty 1
0
0
(0)
1
1
1
0
0
1
1
1
0
Ant 2
0
0
0
(2)
2
1
0
0
1
1
1
0
Bor 7
0
0
1
2
(7)
3
1
1
1
2
3
0
His 16
1
0
1
2
11
(101
4
2
5
5
7
1
Nar 6
1
0
1
1
5
4
(31
2
3
2
5
0
Avo 24
0
0
5
1
10
7
2
(8)
10
a
12
1
Dep 12
0
0
0
1
5
5
3
3
(8)
6
0
Obc 20
2
1
5
1
7
5
1
5
7
(9)
a a
Pas 17
1
0
1
1
a
7
3
4
6
7
(10)
1
Sad 1
1
0
0
0
1
0
1
0
0
1
1
0
a
0
0
1
1
6
3
2
0
2
2
3
0
Mix 4
0
0
0
0
2
2
1
1
1
0
2
(0)
.sd
diagnostic criteria that were virtually the same in the two classifications divided by the total number of DSM-III-R criteria. Column 2 lists those disorders that were given a monothetic definition in the DSM-III and polythetic definition in the DSM-III-R. The third column lists the percentage of patients assigned to each DSM-III diagnosis. The percentages for the DSM-III-R are shown in column 4. The fifth column contains the kappa estimates of agreement between the SIDP/DSM-III and the SIDP-
1
R/DSM-III-R diagnoses. The sixth column contains the Pearson product moment correlations between the number of positive criteria per diagnosis for the two classifications. The seventh and eighth columns report the kappa and correlation values for the n = 28 patients who were administered the SIDP and SIDP-R within 1 month of each other. For this reduced sample, changes in clinical status are less likely to affect the estimates of diagnostic concordance. Concerning diagnostic agreement between
Table 2. Estimates of Agreement Between DSM-III and DSM-III-R Diagnoses No.
Entire Sample
of
Criteria in Common*
Monothetic Polythetict
to
% DSM-III$
l-Month
(n = 72)I
% DSM-III-RI
Sample (n = 28)//
K
r
K
,783
,169 (NS)
r
Paranoid
317
No
3
22
Schizoid
217
Yes
1
3
Schizotypal
419
No
11
1
Antisocial
14122
No
3
3
Borderline
518
No
22
10
Histrionic
118
No
21
22
Narcissistic
519
No
6
a
Avoidant
317
Yes
13
33
,370
,743
,410
,853
Dependent
119
Yes
22
17
,471
,622
,340
,577
Ob-Comp
419
No
15
28
,478
,696
,462
,808
Pass-Aggr
3/9
No
26
24
,408
,550
,650
,770
,182 (NS) n
