PsychologicalReports, 1991, 69, 75-81
O Psychological Reports 1991
MMPI CHARACTERISTICS O F T H E DSM-111-R AVOIDANT PERSONALITY DISORDER ' CHRIS SCHOTTE, MICHAEL MAES, RAYMOND CLWDTS, DIRK DE DONCKER, MACHTELD CLAES AND PAUL COSYNS Antwerp University Hospihl (UZA), Belgium Summary.-The clinical description provided by the MMPI Scale 0 (Social Introversion, Si) reflects an apparent and striking affinity with the DSM-111-R concept of the avoidant personality disorder which was here investlgaccd by comparing the MMPI variables of 22 avoidant with 60 other, nonavoidanc personditv disordered inpatients. Both of the groups were diagnosed by use of a semistructured interview (SCID-11). The avoidant group obtained for mean MMPI profile a 2-6-7 (D-Pa-Pt) code. Significant differences were found between groups on MMPI Scales 2 (D),7 (Pt), 9 (Ma), and 0 (Si). ROC analysis of the diagnostic performance showed the importance and utility of the evaluation of Scale 0 (St) in the psychodiagnostic assessment of the avoidant personality disorder.
The Minnesota Multiphasic Personality Inventory (MMPI) is considered to be the most popular self-report personality inventory among clinicians and researchers. The clinical scales, profiles, and codetypes provide, based on their empirical correlates, descriptions of and predictions for patients. Widiger and Frances (1988) state that the MMPI validity and clinical scales were developed on the basis of their abllity to diagnose clinical syndromes recognized in the 1730s and 1940s. These MMPI scales may therefore not provide the optimal set of dimensions or items with which to identify and differentiate among the DSM-111-R personality disorders. Morey, Waugh, and Blashfield (1785) developed a set of eleven new MMPI scales, derived from rational and internal consistency analyses, which yield dimensional scores corresponding to DSM-I11 personality definitions. However, the examination of the clinical descriptions, provided by the v&dity and clinical MMPI scales, and of the DSM-111-R personality disorder definitions shows an apparent and striking affinity between the MMPI Scale 0 (Social Introversion, Si) and the DSM-111-R concept of avoidant personality disorder. The DSM-111-R (American Psychiatric Association, 1787) definition of the avoidant personality disorder reflects a pattern of inhibited, introverted, and anxious behaviour, with a characteristic low self-esteem, hypersensitivity to rejection, apprehension and mistrust, social awkwardness, timidity, social discomfort, self-conscious fears of being embarrassed or acting foolish, and exaggeration of risks and dangers in everyday life (Widiger & Frances,
'Address correspondence to C. Schotte, Department of Psychiatry, Universitair Ziekenhuis Antwerpen, Wilrijkstraat 10, 2650 Edegem, Belgium.
76
C. SCHOTTE, ETAL.
1988). High scorers on the MMPI Scale 0 (Si) are described (Greene, 1980) as socially introverted, self-deprecating, shy, and socially inept individuals. They tend to withdraw from and to avoid social activity and are aloof and anxious in their social interactions. Scale 0 (Si) consists of 70 items, selected to assess a social introversion-extroversion dimension, with the high scores reflecting social introversion. The scale was constructed by a contrast of college students, who scored at the extremes of the social introversion-extroversion scale of the Minnesota TS-S Inventory (Drake, 1946). Many (26 out of 70) of the rather face-valid items are unique to that scale. Scale 0 (Si) is one of the few MMPI scales in which heritability studies with twins have supported a genetic basis for a substantial portion of its variance. Gottesman (1962) reports in a study of 68 pairs of normal, adolescent twins an heritabhty index-defined as the proportion of within-family variance associated with genotype-of .69 for Scale 0 (Si). This result suggests a substantial genetic component for the social introversion-extroversion dimension, as measured by the MMPI Scale O (Si). The scale is considered to be a fairly stable index of comfort in interpersonal relations: test-retest reliability coefficients for intervals up to one year ranged from .54 to .76 (Dahlstrom, Welsh, & Dahlstrom, 1975). This relative stability could indicate that the scale assesses a trait characteristic, corresponding to the DSM-111 and DSM-111-R concept of a personality disorder as inflexible, deeply ingrained patterns of behaviour. The aim of the present exploratory study was to investigate the MMPI characteristics of-and especially the diagnostic performance of the MMPI Si Scale for-the DSM-111-R avoidant personahty disorder among a population of personality-disordered inpatients.
METHOD
AU
subjects were inpatients admitted to the psychiatric ward. The MMPI (566-item Dutch form) was individually administered to each patient in the second week of admission to the unit. The raw scores on the validity and clinical scales were computed, K-corrected, and converted into T scores. Eighty-two inpatients were assessed with the Structured Clinical Interview for DSM-111-R (Spitzer & Williams, 1987), using the questionnaire and the semistructured clinical interview, constructed for the diagnosis of the DSM111-R personality disorders. Based on the SCID-I1 interview results, two samples were derived from this inpatient group: (a) 22 patients fulfilling the DSM-111-R criteria for avoidant personality disorder and (b) 60 patients obtaining a nonavoidant (fewer than 4 DSM-111-R avoidant criteria fulfilled) personality disorder diagnosis. Differences between the group means on the MMPI scales were tested with a one-way analysis of variance. The diagnostic performance of the individual MMPI scales for the avoidant diagnosis, i.e., the sensitivity, speci-
MMPI: AVOIDANT PERSONALITY DISORDER
77
ficity, and predictive value for a positive and a negative test result, were calculated for the combined sample of 82 personality disordered inpatients in this study. The statistical significance of the diagnostic performance was studied by means of the kappa-statistic for nominal scales (Cohen, 1960) and by means of the Y-statistic (Spitznagel & Helzer, 1985). A receiver operating characteristics (ROC) analysis (Swets, 1979; Hsiao, Bartko, & Potter, 1989) was performed to estimate the discriminative power of the MMPI scales. The ROC analysis generates indices that describe all of a test's possible sensitivities and specificities. The ROC curve is obtained by plotting the true-positive rate versus the false-positive rate for several cut-off values of a diagnostic test. The most widely used ROC index of accuracy which summarises the discriminative power of a diagnostic system is the AUC index (Mossman & Somoza, 1989). T h s index (area under the ROC curve, AUC) corresponds to the probability that a given MMPI scale will correctly sort a pair of subjects, one chosen randomly from the avoidant and the other from the nonavoidant sample. An AUC index of .5 indicates a performance no better than chance; an AUC index of 1.0 reflects a perfect diagnostic performance.
RESULTS No significant differences were obtained between the avoidant and nonavoidant groups on age (t,, = 1.91; mean age of one total sample = 34.5 f 11.8 yr.), or sex ratio [x12(N = 82) = .001, p = ,971. In both groups a higher prevalence of female patients with a malelfemale ratio of .42 in the total sample was observed. The mean numbers of SCID-I1 avoidant criteria were significantly different (F,,8, = 186.69, p < suggesting that both diagnostic personality hsorder groups are well differentiated with respect to the DSM-111-R avoidant features. Analysis of the prevalence of the SCID-I1 nonavoidant personality disorder diagnoses in both samples indicated no significance differences in the prevalence rates for the two groups [maximal (N= 82) = 2.211. The distribution rates of the Axis I1 diagnoses in the total sample were: Cluster A diagnoses 9.40%, Cluster B diagnoses 37.60%, Cluster C diagnoses 45.60%, and NOS Axis I1 diagnoses 7.40%. Concomitant clinical DSM-111-R Axis I diagnoses were obtained for each patient; no significant differences in the prevalence rates were observed between both groups [maximal x,Z (N = 82) = 1.381. The following diagnostic Axis I prevalence rates were observed in the total sample: for no AXISI diagnosis 17Y0, Adjustment Disorder 21.18%, Major Depression 21.18%, Substance Abuse Disorders 15.29%, Dysthymic Disorder 12.94%, and Somatoform Disorders 10.60%. Table 1 displays the mean T scores of the MMPI scales and the summaries of the analyses of variance. These analyses were performed to examine
78
C. SCHOTTE, ETAL. TABLE 1 MEANT SCORESOF THE MMPI SCALESAND ANALYSES OF VARIANCE
MMPI Scales
Personality Disorder Avoidant Nonavoidant
M
SD
M
F
P
SD
the effects of a diagnostic category on the T scores of the 13 MMPI scales. As a correction for experiment-wise error, the acceptable level of alpha was set at p69T for Scale 0 (Si), a sensitivity of 59%, w ~ t ha specificity of 87To and a positive predic-
79
MMPI: AVOIDANT PERSONALITY DISORDER TABLE 2 SUMMARY OF THE RECENEROPERATMG CHARACTERISTICS (ROC) ANALYSES AM
MMPI Scales
AUC Index
S
Note.-M: difference in population means ex ressed in units of standard deviation of the nonavoidant population; S: ratio of the standard Seviations of the two populations, S,,../S., .
tive value of 82%, together with a negative predictive value of 68%, were obtained. Scale O T scores were highly correlated for the total sample with the number of SCID-I1 avoidant criteria (Spearman r = ,588; t = 6.50, p < On the other hand, the correlation of the Scale O scores with the number of the SCID-I1 criteria for the other personality disorders was rather low, with a maximal correlation obtained for the schizotypal personahty disorder (Spearman r = .247; t = 2.28, p = .02). TABLE 3 DIAGNOSTIC PERFORMANCE OF THE SELECTED MMPI SCALES MMPI Scale
Cut-off Score
%
Sens.
Spec.
PV +
K
To
t
Y
t
PV-
0 (5)
>69 T 59 87 82 68 .46 2.57* .51 2.81t 73 3.03t .55 3.28t 82 80 75 >67 T .51 82 72 74 2.877 .54 3.46t >64 T 80 .45 93 .42 3.50t .70 4.86t 95 60 70 >59 T 7.42t 100 48 66 100 2.48* 1.00 254 T 33 82 80 80 81 3.46t .62 3.86t ,557 2(D) >84T .35 64 ?5 2.05* .39 72 67 2.30* 7 (Pt) >73 T Note.-Sens. and Spec.: sensitivity and specificity in %; PV + and PV-: predictive value for a positive (PV + ) or a negative (PV-) test result in %; adjusted for 50% prevalence rate of avoidant personalior disorder, K: Kappa character- characteristic and t ratio; Y Yule-Soitznaeel . istic and i ratio. . *p
O Psychological Reports 1991
MMPI CHARACTERISTICS O F T H E DSM-111-R AVOIDANT PERSONALITY DISORDER ' CHRIS SCHOTTE, MICHAEL MAES, RAYMOND CLWDTS, DIRK DE DONCKER, MACHTELD CLAES AND PAUL COSYNS Antwerp University Hospihl (UZA), Belgium Summary.-The clinical description provided by the MMPI Scale 0 (Social Introversion, Si) reflects an apparent and striking affinity with the DSM-111-R concept of the avoidant personality disorder which was here investlgaccd by comparing the MMPI variables of 22 avoidant with 60 other, nonavoidanc personditv disordered inpatients. Both of the groups were diagnosed by use of a semistructured interview (SCID-11). The avoidant group obtained for mean MMPI profile a 2-6-7 (D-Pa-Pt) code. Significant differences were found between groups on MMPI Scales 2 (D),7 (Pt), 9 (Ma), and 0 (Si). ROC analysis of the diagnostic performance showed the importance and utility of the evaluation of Scale 0 (St) in the psychodiagnostic assessment of the avoidant personality disorder.
The Minnesota Multiphasic Personality Inventory (MMPI) is considered to be the most popular self-report personality inventory among clinicians and researchers. The clinical scales, profiles, and codetypes provide, based on their empirical correlates, descriptions of and predictions for patients. Widiger and Frances (1988) state that the MMPI validity and clinical scales were developed on the basis of their abllity to diagnose clinical syndromes recognized in the 1730s and 1940s. These MMPI scales may therefore not provide the optimal set of dimensions or items with which to identify and differentiate among the DSM-111-R personality disorders. Morey, Waugh, and Blashfield (1785) developed a set of eleven new MMPI scales, derived from rational and internal consistency analyses, which yield dimensional scores corresponding to DSM-I11 personality definitions. However, the examination of the clinical descriptions, provided by the v&dity and clinical MMPI scales, and of the DSM-111-R personality disorder definitions shows an apparent and striking affinity between the MMPI Scale 0 (Social Introversion, Si) and the DSM-111-R concept of avoidant personality disorder. The DSM-111-R (American Psychiatric Association, 1787) definition of the avoidant personality disorder reflects a pattern of inhibited, introverted, and anxious behaviour, with a characteristic low self-esteem, hypersensitivity to rejection, apprehension and mistrust, social awkwardness, timidity, social discomfort, self-conscious fears of being embarrassed or acting foolish, and exaggeration of risks and dangers in everyday life (Widiger & Frances,
'Address correspondence to C. Schotte, Department of Psychiatry, Universitair Ziekenhuis Antwerpen, Wilrijkstraat 10, 2650 Edegem, Belgium.
76
C. SCHOTTE, ETAL.
1988). High scorers on the MMPI Scale 0 (Si) are described (Greene, 1980) as socially introverted, self-deprecating, shy, and socially inept individuals. They tend to withdraw from and to avoid social activity and are aloof and anxious in their social interactions. Scale 0 (Si) consists of 70 items, selected to assess a social introversion-extroversion dimension, with the high scores reflecting social introversion. The scale was constructed by a contrast of college students, who scored at the extremes of the social introversion-extroversion scale of the Minnesota TS-S Inventory (Drake, 1946). Many (26 out of 70) of the rather face-valid items are unique to that scale. Scale 0 (Si) is one of the few MMPI scales in which heritability studies with twins have supported a genetic basis for a substantial portion of its variance. Gottesman (1962) reports in a study of 68 pairs of normal, adolescent twins an heritabhty index-defined as the proportion of within-family variance associated with genotype-of .69 for Scale 0 (Si). This result suggests a substantial genetic component for the social introversion-extroversion dimension, as measured by the MMPI Scale O (Si). The scale is considered to be a fairly stable index of comfort in interpersonal relations: test-retest reliability coefficients for intervals up to one year ranged from .54 to .76 (Dahlstrom, Welsh, & Dahlstrom, 1975). This relative stability could indicate that the scale assesses a trait characteristic, corresponding to the DSM-111 and DSM-111-R concept of a personality disorder as inflexible, deeply ingrained patterns of behaviour. The aim of the present exploratory study was to investigate the MMPI characteristics of-and especially the diagnostic performance of the MMPI Si Scale for-the DSM-111-R avoidant personahty disorder among a population of personality-disordered inpatients.
METHOD
AU
subjects were inpatients admitted to the psychiatric ward. The MMPI (566-item Dutch form) was individually administered to each patient in the second week of admission to the unit. The raw scores on the validity and clinical scales were computed, K-corrected, and converted into T scores. Eighty-two inpatients were assessed with the Structured Clinical Interview for DSM-111-R (Spitzer & Williams, 1987), using the questionnaire and the semistructured clinical interview, constructed for the diagnosis of the DSM111-R personality disorders. Based on the SCID-I1 interview results, two samples were derived from this inpatient group: (a) 22 patients fulfilling the DSM-111-R criteria for avoidant personality disorder and (b) 60 patients obtaining a nonavoidant (fewer than 4 DSM-111-R avoidant criteria fulfilled) personality disorder diagnosis. Differences between the group means on the MMPI scales were tested with a one-way analysis of variance. The diagnostic performance of the individual MMPI scales for the avoidant diagnosis, i.e., the sensitivity, speci-
MMPI: AVOIDANT PERSONALITY DISORDER
77
ficity, and predictive value for a positive and a negative test result, were calculated for the combined sample of 82 personality disordered inpatients in this study. The statistical significance of the diagnostic performance was studied by means of the kappa-statistic for nominal scales (Cohen, 1960) and by means of the Y-statistic (Spitznagel & Helzer, 1985). A receiver operating characteristics (ROC) analysis (Swets, 1979; Hsiao, Bartko, & Potter, 1989) was performed to estimate the discriminative power of the MMPI scales. The ROC analysis generates indices that describe all of a test's possible sensitivities and specificities. The ROC curve is obtained by plotting the true-positive rate versus the false-positive rate for several cut-off values of a diagnostic test. The most widely used ROC index of accuracy which summarises the discriminative power of a diagnostic system is the AUC index (Mossman & Somoza, 1989). T h s index (area under the ROC curve, AUC) corresponds to the probability that a given MMPI scale will correctly sort a pair of subjects, one chosen randomly from the avoidant and the other from the nonavoidant sample. An AUC index of .5 indicates a performance no better than chance; an AUC index of 1.0 reflects a perfect diagnostic performance.
RESULTS No significant differences were obtained between the avoidant and nonavoidant groups on age (t,, = 1.91; mean age of one total sample = 34.5 f 11.8 yr.), or sex ratio [x12(N = 82) = .001, p = ,971. In both groups a higher prevalence of female patients with a malelfemale ratio of .42 in the total sample was observed. The mean numbers of SCID-I1 avoidant criteria were significantly different (F,,8, = 186.69, p < suggesting that both diagnostic personality hsorder groups are well differentiated with respect to the DSM-111-R avoidant features. Analysis of the prevalence of the SCID-I1 nonavoidant personality disorder diagnoses in both samples indicated no significance differences in the prevalence rates for the two groups [maximal (N= 82) = 2.211. The distribution rates of the Axis I1 diagnoses in the total sample were: Cluster A diagnoses 9.40%, Cluster B diagnoses 37.60%, Cluster C diagnoses 45.60%, and NOS Axis I1 diagnoses 7.40%. Concomitant clinical DSM-111-R Axis I diagnoses were obtained for each patient; no significant differences in the prevalence rates were observed between both groups [maximal x,Z (N = 82) = 1.381. The following diagnostic Axis I prevalence rates were observed in the total sample: for no AXISI diagnosis 17Y0, Adjustment Disorder 21.18%, Major Depression 21.18%, Substance Abuse Disorders 15.29%, Dysthymic Disorder 12.94%, and Somatoform Disorders 10.60%. Table 1 displays the mean T scores of the MMPI scales and the summaries of the analyses of variance. These analyses were performed to examine
78
C. SCHOTTE, ETAL. TABLE 1 MEANT SCORESOF THE MMPI SCALESAND ANALYSES OF VARIANCE
MMPI Scales
Personality Disorder Avoidant Nonavoidant
M
SD
M
F
P
SD
the effects of a diagnostic category on the T scores of the 13 MMPI scales. As a correction for experiment-wise error, the acceptable level of alpha was set at p69T for Scale 0 (Si), a sensitivity of 59%, w ~ t ha specificity of 87To and a positive predic-
79
MMPI: AVOIDANT PERSONALITY DISORDER TABLE 2 SUMMARY OF THE RECENEROPERATMG CHARACTERISTICS (ROC) ANALYSES AM
MMPI Scales
AUC Index
S
Note.-M: difference in population means ex ressed in units of standard deviation of the nonavoidant population; S: ratio of the standard Seviations of the two populations, S,,../S., .
tive value of 82%, together with a negative predictive value of 68%, were obtained. Scale O T scores were highly correlated for the total sample with the number of SCID-I1 avoidant criteria (Spearman r = ,588; t = 6.50, p < On the other hand, the correlation of the Scale O scores with the number of the SCID-I1 criteria for the other personality disorders was rather low, with a maximal correlation obtained for the schizotypal personahty disorder (Spearman r = .247; t = 2.28, p = .02). TABLE 3 DIAGNOSTIC PERFORMANCE OF THE SELECTED MMPI SCALES MMPI Scale
Cut-off Score
%
Sens.
Spec.
PV +
K
To
t
Y
t
PV-
0 (5)
>69 T 59 87 82 68 .46 2.57* .51 2.81t 73 3.03t .55 3.28t 82 80 75 >67 T .51 82 72 74 2.877 .54 3.46t >64 T 80 .45 93 .42 3.50t .70 4.86t 95 60 70 >59 T 7.42t 100 48 66 100 2.48* 1.00 254 T 33 82 80 80 81 3.46t .62 3.86t ,557 2(D) >84T .35 64 ?5 2.05* .39 72 67 2.30* 7 (Pt) >73 T Note.-Sens. and Spec.: sensitivity and specificity in %; PV + and PV-: predictive value for a positive (PV + ) or a negative (PV-) test result in %; adjusted for 50% prevalence rate of avoidant personalior disorder, K: Kappa character- characteristic and t ratio; Y Yule-Soitznaeel . istic and i ratio. . *p
