Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Journal of Personality Assessment

ISSN: 0022-3891 (Print) 1532-7752 (Online) Journal homepage: http://www.tandfonline.com/loi/hjpa20

Exploring the Assessment of the DSM–5


Alternative Model for Personality Disorders With
the Personality Assessment Inventory

Alexander J. Busch, Leslie C. Morey & Christopher J. Hopwood

To cite this article: Alexander J. Busch, Leslie C. Morey & Christopher J. Hopwood (2016):
Exploring the Assessment of the DSM–5 Alternative Model for Personality Disorders With the
Personality Assessment Inventory, Journal of Personality Assessment

To link to this article: http://dx.doi.org/10.1080/00223891.2016.1217872

Published online: 06 Sep 2016.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=hjpa20

Download by: [Cornell University Library] Date: 07 September 2016, At: 08:14
JOURNAL OF PERSONALITY ASSESSMENT
http://dx.doi.org/10.1080/00223891.2016.1217872

Exploring the Assessment of the DSM–5 Alternative Model for Personality Disorders
With the Personality Assessment Inventory
Alexander J. Busch,1 Leslie C. Morey,1 and Christopher J. Hopwood2
1
Department of Psychology, Texas A&M University; 2Department of Psychology, Michigan State University

ABSTRACT ARTICLE HISTORY


Section III of the Diagnostic and Statistical Manual of Mental Disorders (5th ed. [DSM–5]; American Received 29 July 2015
Psychiatric Association, 2013) contains an alternative model for the diagnosis of personality disorder Revised 24 June 2016
involving the assessment of 25 traits and a global level of overall personality functioning. There is hope
that this model will be increasingly used in clinical and research settings, and the ability to apply
established instruments to assess these concepts could facilitate this process. This study sought to
develop scoring algorithms for these alternative model concepts using scales from the Personality
Assessment Inventory (PAI). A multiple regression strategy used to predict scores in 2 undergraduate
samples on DSM–5 alternative model instruments: the Personality Inventory for the DSM–5 (PID–5) and the
General Personality Pathology scale (GPP; Morey et al., 2011). These regression functions resulted in scores
that demonstrated promising convergent and discriminant validity across the alternative model concepts,
as well as a factor structure in a cross-validation sample that was congruent with the putative structure of
the alternative model traits. Results were linked to the PAI community normative data to provide
normative information regarding these alternative model concepts that can be used to identify elevated
traits and personality functioning level scores.

Section II of the fifth edition of the Diagnostic and Statistical established validity in basic personality science (Markon,
Manual of Mental Disorders (DSM–5; American Psychiatric Krueger, & Watson, 2005; Samuel & Widiger, 2008; Widiger &
Association, 2013) retains a categorical model for the diagnosis Costa, 1994; Widiger & Simonsen, 2005; Widiger & Trull, 2007;
of personality disorders (PDs). However, the DSM–5 also Wiggins & Pincus, 1989). Therefore the inclusion of each of these
includes an alternative model for PD diagnosis within Section elements was part of deliberations on the DSM–5 process from
III, Emerging Measures and Models. This new model uses very early on (Skodol, 2011). The Personality and Personality
dimensional measures of self and interpersonal dysfunction Disorder Work Group attempted to develop initial descriptors of
(Criterion A) as well as hierarchically organized traits (Criterion these characteristics that were operationalized, evaluated, and
B) as a means of conceptualizing personality pathology and spe- ultimately refined using self-report measures. The Personality
cific forms of PD. The aim of the model was to provide a broad Inventory for the DSM–5 (PID–5; Krueger, Derringer, Markon,
but coherent system for identifying personality psychopathol- Watson, & Skodol, 2012) was used to assess and evaluate the trait
ogy, quantifying its severity, and characterizing its myriad clini- dimensions. Similarly, a composite General Personality Pathol-
cal manifestations in terms of impairments in personality ogy (GPP) scale (Morey et al., 2011) derived from two instru-
functioning and pathological personality traits (Skodol, Morey, ments, the Structured Inventory for Personality Pathology
Bender, & Oldham, 2013). Early research results have been (SIPP–118; Verhuel et al., 2008) and the General Assessment of
promising (Krueger & Markon, 2014; Morey, Benson, Busch, & Personality Disorder (GAPD; Livesley, 2006), was used to reflect
Skodol, 2015). However, one practical obstacle to its adoption severity. The PID–5 is made up of 25 relatively homogeneous
involves the availability and novelty of validated assessment traits that can be organized hierarchically under five broad fac-
approaches for assessing Section III features. The purpose of tors: negative affectivity, detachment, antagonism, disinhibition,
this study was to develop and evaluate methods to assess Crite- and psychoticism (Krueger et al., 2012; Wright et al., 2012). Ini-
rion B traits and the Criterion A level of disordered personality tial research with the PID–5 (see review by Krueger & Markon,
severity with a widely used clinical assessment tool, the Person- 2014) suggests that this structure replicates fairly well across
ality Assessment Inventory (PAI; Morey, 1991). samples (e.g., Wright et al., 2012), resembles that found in clini-
A large body of research demonstrates the benefits of consid- cal judgments of these 25 traits (Morey, Krueger, & Skodol,
ering a general severity dimension of personality pathology (e.g., 2013), and converges reasonably well with the Five-Factor Model
Bender, Morey, & Skodol, 2011; Hopwood et al., 2011) as well as of normative personality (Gore & Widiger, 2013; Thomas et al.,
reorganizing variants of PDs using personality traits with 2013). Furthermore, a number of studies indicate that PID–5

CONTACT Leslie C. Morey morey@tamu.edu Department of Psychology, Texas A&M University, 282 Psychology Building, College Station, TX 77843-4235.

© 2016 Taylor & Francis


2 BUSCH, MOREY, HOPWOOD

scores correlate as expected with measures of PDs and other strongly related to DSM–5 withdrawal. This might suggest that
forms of psychopathology and dysfunction (Hopwood et al., DSM–5 withdrawal involves elements of aloof detachment tinged
2013; Krueger & Markon, 2014). Similarly, the GPP scale has with concerns about closeness and experiences of negative affect.
been shown to be a sensitive and specific indicator of whether Thus, the degree to which single PAI scales or collections of PAI
patients received a diagnosis of DSM–IV PD, and higher scores scales would optimally indicate DSM–5 traits and personality func-
on this scale were found to significantly predict PD comorbidity tioning is an empirical question. To address this question, we sought
(Morey et al., 2011). to develop predictive functions to optimize measurement of particu-
Given the relatively recent introduction of this alternative model lar constructs (in this case, DSM–5 traits and levels of personality
for formulating PDs and the importance of supporting additional functioning) using multiple regression. It was anticipated that this
research using this model, there is a need for additional strategies empirically driven approach (in contrast to an attempt to rationally
for assessing these constructs. At present, the PID–5 and the GPP select scale combinations) would provide predictive functions with
scale represent initial forays into this area, but there are other instru- reasonable convergent and divergent validity as calibrated against
ments that could conceivably be adapted for this purpose. This the PID–5 and GPP scale scores. PAI regression functions could be
would be expected given the presumption that these traits and this then subsequently translated into interpretable scores using norms
global severity dimension permeate most measures of personality against the census-matched community sample as well as the clini-
problems, and there are a number of measures that are given rou- cal normative of the PAI.
tinely in clinical practice to assess such problems, such as the
MMPI–2/MMPI–2–RF (Sellbom, Anderson, & Bagby, 2013). This
article seeks to provide such a strategy using the scales of the PAI Methods
(Morey, 1991), an instrument widely used in many clinical settings
Participants
(e.g., Archer, Buffington-Vollum, Stredny, & Handel 2006) and one
that, unlike the PID–5 or GPP, include validity scales that can iden- Two samples of participants were examined in this study. The
tify problematic distortions in responding as well as providing PID–5 sample included 1,187 college students recruited to par-
assessments of several other important diagnostic and clinical con- ticipate in exchange for course credit. These participants were
structs. Initial research suggests that the PAI scales have relatively administered the PID–5 and the PAI. Participant scores were
strong and systematic correlations with the PID–5 (Hopwood et al., excluded from the study if missing more than 10% of the items,
2013). In that study, conjoint exploratory factor analysis (EFA) of a for scoring above 67T on the PAI Positive Impression Manage-
large sample of undergraduate students followed by exploratory ment (PIM) scale, or for scoring above 81T (a cut score identi-
structural equation modeling (ESEM) with target rotation replicated fied as optimal in a meta-analysis by Hawes & Boccaccini,
convergence at the broad level around five factors also seen in other 2009) on the PAI Negative Impression Management (NIM)
studies involving the PID–5 (Krueger et al., 2012; Wright et al., scale,1 leaving 922 valid cases. A total of 72% of the included
2012). Furthermore, the bivariate correlations between PID–5 and participants were women, and the mean age was 19.63 years
PAI scales suggested a level of specificity that speaks to the potential (SD D 2.31). Eight-hundred forty-one (84%) of the participants
utility of the PAI for DSM–5 trait assessment. Specifically, appreci- were White, 38 (4%) were African American, 32 (3%) were
able correlations between PID–5 traits and PAI scales suggests that Asian American, 23 (2%) were multiracial, 21 (2%) were His-
particular collections of PAI scales could be used to develop sensitive panic, 32 (3%) were “other,” and 14 participants did not report
and discriminating indicators of DSM–5 traits and trait domains as their ethnicity.
measured by the PID–5. Along similar lines, preliminary research To assess the level of personality functioning, a different
has indicated that the GPP scale demonstrated strong associations sample was used that included 104 college students enrolled in
with the PAI, particularly the Borderline Features (PAI-BOR) scale an upper level course on personality psychology. Participation
(Lowmaster & Morey, 2011), and such borderline characteristics represented an optional educational research project as an
have been identified in other studies as indicators of core personality opportunity to learn more about personality traits and disor-
impairments (e.g., Sharp et al., 2015; Turkheimer, Ford, & ders; 104 of the 115 students in the course (90.4%) elected to
Oltmanns, 2008). However, these initial studies did not describe participate in the project. These participants were administered
specific scoring algorithms that could allow a direct translation of the Borderline Features (BOR) scale of the PAI and the GPP.
PAI scale scores into the constructs of the DSM–5 alternative model. The average age of the sample was 21.4 years; 64.4% were
Accordingly, a purpose of this study was to build on this initial women and 35.6% were men.
research to develop and test a specific method for assessing such
constructs using the PAI.
A number of approaches have been suggested from which one Measures
can make inferences about personality and psychopathology con- Personality Inventory for DSM–5
structs using the PAI (e.g., Morey, 1996), with the most direct The PID–5 is comprised of 220 items and contains 25 nonover-
approach involving the use of individual PAI scales for interpreta- lapping scales that can be combined into five higher order
tion. However, many of the DSM–5 alternative model constructs dimensions of negative affect, detachment, antagonism,
seem to reflect a higher order combination of features measured by
the PAI. For instance, Table 1 in Hopwood et al. (2013) suggests
1
that although the DSM–5 trait facet of withdrawal is strongly corre- Scores on the PAI Inconsistency and Infrequency scales were not available in this
sample; as such, a more conservative score on NIM was selected because PAI
lated with PAI social detachment (SCZ-S), as would be expected, protocols with considerable random responding will typically obtain NIM scores
other PAI scales related to depression and low warmth are also at or above such scores (Morey, 2007).
ASSESSING DSM–5 PERSONALITY PATHOLOGY WITH THE PAI 3

disinhibition, and psychoticism (Krueger et al., 2012; Wright To assess discriminant validity, the predicted scores from each
et al., 2012). Internal consistencies across all trait scales as mea- construct in the DSM–5 alternative model were saved for each
sured in this sample ranged from .40 to .84. Each PID–5 item is participant, and then correlated with all other constructs (both
endorsed on a 4-point scale and receives an item score ranging individual facet traits and the general personality pathology indi-
from 0 (very false) to 3 (very true); scale scores are computed to cator). The mean of the absolute values of these discriminant
reflect the average item score on that scale, with higher values validity correlations (using the r to z transformation in calculat-
indicating greater degrees of the indicated trait. ing these means) provided estimates of discriminant validity at a
global level, whereas any particularly problematic discriminant
Personality Assessment Inventory validity correlations (instances where PAI-predicted scores for a
The PAI includes 344 items that are combined into 22 full certain DSM–5 construct actually displayed a larger correlation
scales and 31 subscales to provide broad coverage of psychopa- with a different DSM–5 construct) were noted.
thology and clinical constructs. In this sample, the internal con- To provide a reference point for determinations regarding “ele-
sistencies for all PAI full scales ranged from .69 to .91 and from vation” of a particular DSM–5 construct as predicted by PAI
.56 to .86 for all PAI subscales. scales, the predictive functions for the DSM–5 constructs were
applied to the census-matched community and the clinical nor-
General Personality Pathology scale mative samples of the PAI (Morey, 1991) and means and standard
The GPP consists of 65 items that were selected from the SIPP–118 deviations were calculated. The community sample estimates are
(Verheul et al., 2008) and the GAPD (Livesley, 2006). These two of particular interest given the stipulation in the alternative model
measures are self-report questionnaires designed to assess the sever- that to be of clinical significance, traits should be determined to be
ity and core components of personality pathology. Items for the “elevated” in comparison to population norms. At this point, there
GPP were initially selected from these instruments on the basis of are few empirical methods for making such a determination, and
their conceptual relationship with the core impairments in person- thus the norms from the PAI community sample provide an
ality pathology as described in the DSM–5 alternative model important opportunity for such comparisons.
(Bender et al., 2011), including impairments in identity, self-direc- Finally, the factor structure of the PAI-predicted DSM–5
tion, empathy, and intimacy. This initial scale was subsequently pathological trait constructs was also examined to determine
refined using item response theory to provide a unidimensional whether it corresponded to the anticipated five-factor structure
marker of impairments in personality functioning at various levels. described by that model. This factor analysis involved an extrac-
Items on the GPP are endorsed on a 4-point scale ranging from 1 tion of five factors with a principal axis factoring followed by an
(strongly disagree) to 4 (strongly agree), with the total score reflecting oblique Equamax rotation, the approach used by Krueger et al.
the sum of items scores such that higher scores indicate greater per- (2012) in the initial derivation of the PID–5. The resulting factor
sonality pathology. The GPP demonstrated an internal consistency structure loadings were then compared to that described for the
of .97 in this sample. PID–5 by using coefficients of congruence, examining congru-
ence both for the Equamax rotation as well as a Procrustes rota-
tion to Krueger et al.’s solution to evaluate structure replication.
Analysis
Data analyses involved stepwise entry multiple regression (p < .05
to enter, p > .10 to remove) using the PAI scales or subscales Results
(excluding the validity scales ICN and INF) to predict the individual
scores resulting from the PID–5 and the GPP. To predict PID–5 The Appendix provides the derived regression functions for pre-
scores, all PAI scales other than the ICN and INF validity scales dicting the 25 DSM–5 pathological traits and the global level of
(which are not measures of substantive constructs) were included in personality pathology using PAI scales or subscales. These func-
these analyses, as there were no explicit hypotheses about relation- tions provide empirically selected PAI indicators that can be
ships between PAI and PID–5 constructs. The analyses predicting applied to obtained PAI scores as a means of assessing the
GPP scores focused exclusively on the four PAI-BOR subscales, as respondent’s DSM–5 traits and level of personality functioning.2
these subscales have previously been hypothesized to be related to To determine the effectiveness of these selected PAI scales in
general personality pathology (e.g., Morey, 1991). To estimate the predicting DSM–5 traits and personality pathology severity,
generalizability of these predictive functions to a cross-validation Table 1 provides estimates of convergent and discriminant
sample, we employed the predicted residual sums of squares
(PRESS) cross-validation method to correct for potential model 2
Given the complexity of the estimation formulas provided in the Appendix, we
overfitting (Stevens, 2002). A PRESS analysis builds a model with also tested unit-weighted functions using significant PAI scales as predictors, as
data from every participant except the person whose score is being suggested by Cohen (1990), who observed that weighted and (more parsimoni-
ous) unweighted functions tend to yield results that are very highly correlated.
predicted, and this recurs for every participant. Model effect size is Two unit-weighting approaches were examined, one using all significant predic-
estimated based on the observed residuals in the entire sample, pro- tors from the regression models, and the second using only unit-weighted pre-
viding a cross-validation uninfluenced by overfitting resulting from dictors demonstrating a zero-order correlation of at least .40 with the criterion
PID scale. Both of these approaches yielded significant reductions in the differen-
larger numbers of predictors in the regression model. Such an tiation of convergent and discriminant validity correlations in the resulting multi-
approach is particularly useful to control for overfitting artifacts trait–multimethod matrix, relative to that provided by the empirical regression
arising from the use of correlated predictor variables (Weisberg, function weights, after jack-knifed cross-validation within this sample (complete
details are available on request). It is an open question whether unit weights
1985), which is typically the case when studying mental health vari- would demonstrate comparable generalization in a new sample with different
ables as predictors. characteristics.
4 BUSCH, MOREY, HOPWOOD

Table 1. Regression functions and descriptive statistics predicting DSM–5 traits from Personality Assessment Inventory scales.

Communitya Clinicalb Mean discriminant validity


Convergent PRESS
Alternative model construct M SD M SD multiple R correlation R Worst

Submissiveness 1.31 0.32 1.44 0.41 .53 .51 .08 .35


Depressivity 0.25 0.39 0.79 0.65 .86 .85 .49 .81
Separation insecurity 0.88 0.41 1.33 0.53 .61 .60 .28 .51
Anxiousness 1.18 0.47 1.70 0.66 .82 .82 .29 .69
Emotional lability 0.87 0.51 1.44 0.73 .74 .73 .35 .62
Suspiciousness 0.84 0.40 1.19 0.53 .77 .76 .46 .62
Restricted affectivity 0.86 0.35 0.98 0.42 .60 .59 .26 .47
Withdrawal 0.80 0.42 1.13 0.57 .79 .79 .40 .67
Intimacy avoidance 0.41 0.32 0.64 0.38 .52 .50 .29 .49
Anhedonia 0.92 0.41 1.45 0.64 .84 .84 .44 .70
Manipulativeness 0.98 0.41 1.15 0.48 .61 .60 .30 .53
Deceitfulness 0.69 0.37 0.92 0.43 .66 .65 .43 .61
Hostility 0.99 0.43 1.26 0.55 .79 .78 .44 .60
Callousness 0.41 0.31 0.57 0.39 .73 .72 .45 .69
Attention seeking 1.03 0.40 1.16 0.46 .59 .57 .20 .43
Grandiosity 0.78 0.35 0.80 0.39 .60 .59 .25 .51
Irresponsibility 0.41 0.33 0.69 0.43 .70 .69 .47 .68
Impulsivity 0.73 0.41 1.03 0.55 .70 .69 .38 .59
Distractibility 0.95 0.41 1.34 0.60 .73 .72 .38 .56
Perseveration 0.91 0.37 1.26 0.53 .68 .67 .43 .60
Rigid perfectionism 1.21 0.46 1.32 0.56 .77 .77 .18 .32
Risk taking 1.21 0.37 1.26 0.48 .81 .80 .11 .56
Eccentricity 1.02 0.44 1.29 0.55 .66 .64 .42 .60
Perceptual dysregulation 0.61 0.34 0.93 0.50 .75 .74 .51 .75
Unusual beliefs 0.71 0.41 0.92 0.58 .73 .72 .44 .72
Personality disorder severity 111.13 22.34 140.98 30.48 .86 .84 .38 .80

n D 1,000. bn D 1,246.
a

validity. The convergent multiple correlation provides the cor- Section III model. The means and standard deviations of these
relation between the PID–5 and GPP scales and the PAI com- composites as applied to the PAI normative data are also pre-
posite constructed to predict that scale score. The PRESS sented in Table 1. It is worth noting that estimated PID–5
correlations provide the cross-validated multiple correlation scores of two or greater, which indicate an average response of
convergent values; as expected, there is some shrinkage in the “present” across the items on a scale, represent a score at least 1
convergent validity values after cross-validation. For discrimi- SD above the community mean for all traits. Furthermore, as
nant validity, the average (following an r to z transformation) expected, the mean scores for the PAI clinical normative sam-
correlation between a given PAI composite and all other PID–5 ple were higher than that for the community sample on every
scales is presented. The median convergent validity correlation DSM–5 alternative model construct.
for these constructs was .71 (PRESS r average D .70), demon- As a final gauge of the verisimilitude of the PAI-estimated
strating a reasonably strong relationship between PAI compo- DSM–5 alternative model scores to the target constructs, an
sites and the target traits that displayed only modest shrinkage EFA of the estimated scores for the pathological trait constructs
in PRESS cross-validation. The PAI scale combination predict- was conducted, as applied to the PAI community normative
ing level of personality functioning had the highest convergent sample (Morey, 2007). This factor analysis involved a principal
multiple r value, at .86. In contrast, the mean discriminant axis factoring of the 25 PAI-estimated PID–5 scores, followed
validity values had an average of .35 (Table 1). Table 1 also by an oblique Equamax rotation of five factors following the
presents the most problematic discriminant validity correlation approach of Krueger et al. (2012) in the derivation of the PID–
for each scale; only one instance (of 325 total discriminant 5. The factor loadings from this model are presented in Table 2.
validity correlations) was noted where a discriminant validity In Table 2, trait facets as assigned to the relevant domains in
correlation exceeded the cross-validated PRESS multiple corre- DSM–5 Section II are shown in bold. The pattern of factor
lation, this being the PAI perceptual dysregulation composite, loadings is consistent with the structural organization of the
which demonstrated a slightly larger correlation with the PID– pathological traits presented in DSM–5. As a summary of this
5 eccentricity trait than with the target perceptual dysregulation convergence, Tucker coefficients of congruence between these
scale. However, in general, the PAI composites demonstrated a PAI-estimated scale factor loadings and those presented by
solid combination of convergent and discriminant validity, sug- Krueger et al. (2012) in the derivation of the PID–5 are pre-
gesting that these selected PAI scale combinations are poten- sented in the final row of Table 2, with four of the five congru-
tially useful predictors of DSM–5 traits and level of personality ence coefficients exceeding .85. Following the method of
functioning. Paunonen (1997), a Procrustes rotation to the Krueger et al.
Applying these PAI scale composites to the PAI community (2012) solution yielded five factors with Tucker coefficients
and clinical normative data provides an opportunity to estab- ranging from .91 to .96, well above the 99% confidence interval
lish norms for the trait and functional constructs of the DSM–5 for replication provided by Paunonen. These results signify
ASSESSING DSM–5 PERSONALITY PATHOLOGY WITH THE PAI 5

Table 2. Five-factor exploratory factor analysis solution from Personality Assessment Inventory (PAI)-estimated DSM–5 trait scores in the PAI community normative
sample.
Facet trait Negative affectivity Detachment Antagonism Disinhibition Psychoticism

Emotional lability .78 .06 .24 .22 .30


Anxiousness .82 .19 .09 .02 .33
Separation insecurity .79 .06 .17 .14 .38
Perseveration .68 .24 .26 .09 .59
Submissiveness .53 ¡.07 ¡.23 ¡.18 .37
Hostility .53 .23 .47 .36 .14
Restricted affectivity ¡.20a .82 .17 .22 .11
Depressivity .50 .57 ¡.01 .25 .42
Suspiciousness .50 .43 .42 .15 .32
Withdrawal .22 .86 .05 .01 .18
Anhedonia .47 .71 ¡.01 .21 .26
Intimacy avoidance .01 .74 .10 .12 .32
Manipulativeness .10 .20 .76 .50 .14
Deceitfulness .28 .48 .48 .54 .22
Grandiosity ¡.05 .12 .85 .24 .20
Attention seeking .14 ¡.34 .76 .38 .18
Callousness .13 .56 .52 .49 .18
Irresponsibility .26 .51 .23 .54 .44
Impulsivity .16 .23 .28 .78 .36
Rigid perfectionism .46 .10 .35 ¡.31y .10
Distractibility .47 .25 ¡.10 .37 .65
Risk taking ¡.25 ¡.03 .30 .80 .03
Unusual beliefs and experiences .13 .23 .50 .13 .74
Eccentricity .28 .20 .44 .39 .59
Cognitive and perceptual dysregulation .27 .37 .32 .26 .78
Congruence with Krueger et al. (2012) .88 .85 .87 .85 .91

Note. N D 1,000. DSM–5-based hypothesized factor specifications are shown in bold.


a
Trait relationship hypothesized to be negative.

reasonable cross-instrument and cross-sample congruence in Morey 1997). It should be noted that there are a number of
the factor structure of the PID–5 and that of the PAI-estimated potential benefits to providing such indicators in the context of
DSM–5 trait scores. a broadband clinical measure such as the PAI. For example, the
PAI provides measures of a host of other diagnostic constructs
related, for example, to substance misuse, psychotic disorders,
Discussion
mood disorders, and somatic disorders, that can provide
The DSM–5 Section III model attempts to link empirical broader diagnostic perspective than PD-specific measures such
research in personality with the widely used DSM system of as the PID–5 or the GPP. Another advantage in the use of the
psychiatric diagnosis (Krueger & Markon, 2014). The model PAI is the provision of normative scores for these composites,
has the potential to improve the validity and clinical utility of as calculated in a large, national, census-matched community
personality diagnosis and more effectively connect clinical sample that serves as the normative basis for the PAI itself.
practice with basic research on normal personality and psycho- Such norms are essential for an understanding of these con-
pathology. Furthermore, this model provides a more explicit cepts, given that the Section III model notes that for these
structural framework that can connect dimensional psychologi- dimensional concepts, “the judgment that a specific trait is ele-
cal assessment data with personality-related diagnostic con- vated (and therefore is present for diagnostic purposes) could
cepts. Research that connects this new model to existing involve comparison of individuals’ personality trait levels with
personality measures that are well validated and widely used in population norms” (American Psychiatric Association, 2013,
psychopathology will facilitate research and application using p. 774). The provision of descriptive statistics for these DSM–5
this model. alternative model score composites provides a much needed
The goal of this study was to develop a strategy for using the foundation for making such determinations of elevation, a step
PAI to assess DSM–5 traits and personality functioning, pro- critical in diagnostic determinations. Finally, the availability of
viding a currently unique opportunity to assess both compo- a host of empirically supported indicators of PAI profile valid-
nents of the DSM–5 alternative model for PD. A series of ity augments the clinical utility of the information derived from
regression analyses were used to develop composites of PAI the DSM–5 alternative model.
scales that could be used as indicators of DSM–5 Section III The convergence of these PAI-estimated DSM–5 model
concepts. The evidence of convergent and discriminant validity scores with the model itself was also examined through the use
of these composites with respect to the PID–5 and GPP meas- of EFA to compare the obtained structure with that hypothe-
ures provides an important first step in attempting to establish sized in DSM–5 Section III. In most respects, there was appreci-
their construct validity for the DSM–5 Section III model, and able convergence between these PAI-estimated scores and the
the obtained convergent and discriminant validity estimates hypothesized five-factor structure of the DSM–5 pathological
tend to be equal to or better than similar values found when trait model. Certain traits demonstrated some unanticipated
measuring traditional PD concepts (e.g., Clark, Livesley, & cross-factor loadings—most noticeably distractibility, a putative
6 BUSCH, MOREY, HOPWOOD

disinhibition trait that displayed appreciable cross-loadings on phenomena that they were validated against. It was the goal of
negative affectivity and psychoticism. However, at the global the DSM–5 Personality and Personality Disorders Work Group
level there was clear similarity between the structure of these to provide a comprehensive model of personality structure and
PAI-estimated scores and those generated by the PID–5. For dysfunction that would tie to important concepts in contempo-
example, the median coefficient of congruence between the rary personality science. To continue to evaluate this model,
PAI-estimated traits score factors, as calculated in the PAI nor- objective and replicable methods to measure the concepts of
mative sample, and the factor loadings presented in the original this model are needed. The results from this study provide sup-
PID–5 article (Krueger et al., 2012) was .87. By comparison, port for the notion that the concepts from the alternative model
Wright et al. (2012) attempted to replicate the Krueger et al. could be estimated using combinations of scores from the PAI,
(2012) five-factor EFA solution in a sample of college under- and thus application of this strategy might assist the field in
graduates, and obtained a median coefficient of congruence of determining whether the DSM–5 alternative model is indeed a
.82 across the five factors. It is noteworthy that the structural viable and useful model for conceptualizing and assessing per-
replication for these PAI-estimated scores, which were derived sonality psychopathology.
from a different sample and a different instrument from
Krueger et al. (2012), provided stronger replication of this
structure using PAI composites than did Wright et al. (2012)’s Disclosure
cross-sample EFA replication of the PID–5 itself. This confir- Leslie C. Morey is the author of the Personality Assessment Inventory and
mation of a hypothesized pattern of anticipated interrelation- derives royalties from its sale.
ships is an important step in the process described by Westen
and Rosenthal (2003), who suggested a means by which con-
struct validity could be quantified in a way that allows for com- References
parisons across studies and measures. American Psychiatric Association. (2013). Diagnostic and statistical man-
It is important to underscore that this study is an initial step ual of mental disorders (5th ed.). Arlington, VA: Author.
toward a strategy for specifying DSM–5 alternative model con- Archer, R. P., Buffington-Vollum, J. K., Stredny, R. V., & Handel, R. W.
cepts using a widely used clinical tool—the PAI. There is clearly (2006). A survey of psychological test use patterns among forensic psy-
a need for additional research, both on these PAI scores as well chologists. Journal of Personality Assessment, 87, 84–94.
Bender, D. S., Morey, L. C., & Skodol, A. E. (2011). Toward a model for
as on the alternative model, and it is hoped that the former assessing level of personality functioning in DSM–5, part I: A review of
might help serve as a catalyst for the latter. It will be particu- theory and methods. Journal of Personality Assessment, 93, 332–346.
larly important to evaluate these indicators against clinical Clark, L. A., Livesley, W. J., & Morey, L. C. (1997). Personality disorder
judgments of DSM–5 alternative model concepts, such as the assessment: The challenge of construct validity. Journal of Personality
Level of Personality Functioning scale provided in the DSM. It Disorders, 11, 205–231.
Cohen, J. (1990). Things I have learned (so far). American Psychologist, 45,
should also be acknowledged that the reference point for the 1304–1312.
development of these PAI indicators, the PID–5 and the GPP, Crego, C., Gore, W. L., Rojas, S. L., & Widiger, T. A. (2015). The discrimi-
are both relatively recently introduced measures with a limited nant (and convergent) validity of the Personality Inventory for DSM–
body of validity evidence behind them. In particular, the dis- 5. Personality Disorders: Theory, Research, and Treatment, 6, 321–335.
criminant validity of these reference measures has not been Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological
tests. Psychological Bulletin, 52, 281–302.
well-established; for example, Markon, Quilty, Bagby, and Gore, W. L., & Widiger, T. A. (2013). The DSM–5 dimensional trait model
Krueger (2013) and Crego, Gore, Rojas, and Widiger (2015) and five factor models of general personality. Journal of Abnormal Psy-
both noted some discriminant validity issues with the PID–5 chology, 122, 816–821.
scales that present an important consideration in the further Hawes, S. W., & Boccaccini, M. T. (2009). Detection of overreporting of
study and refinement of these PAI composites. In terms of this psychopathology on the Personality Assessment Inventory: A meta-
analytic review. Psychological Assessment, 21(1), 112–124.
study, there was a tendency for PAI scales to be included in Hopwood, C. J., Malone, J. C., Ansell, E. B., Sanislow, C. A., Grilo, C. M.,
multiple composites across 26 personality pathology constructs, McGlashan, T. H., … Morey, L. C. (2011). Personality assessment in
with the median PAI scale involved in 7 of the 26 predictive DSM–V: Empirical support for rating severity, style, and traits. Journal
functions. However, a diverse set of PAI indicators were of Personality Disorder, 25, 305–320.
involved, with 38 of 42 PAI scales or subscales being repre- Hopwood, C. J., Wright, A. G., Krueger, R. F., Schade, N., Markon, K. E., &
Morey, L. C. (2013). DSM–5 pathological personality traits and the
sented on at least one composite. It might be this diversity of Personality Assessment Inventory. Assessment, 20, 269–285.
predictors that allowed the generally good discriminant validity Krueger, R. F., Derringer, J., Markon, K. E., Watson, D., & Skodol, A. E.
results across scales that were observed in Table 1. (2012). Initial construction of a maladaptive personality trait model
Furthermore, a major limitation of the study that should and inventory for DSM–5. Psychological Medicine, 42, 1879–1890.
also be noted is that the functions in this study were derived Krueger, R. F., & Markon, K. E. (2014). The role of the DSM–5 personality
trait model in moving toward a quantitative and empirically-based
using university student samples, which might limit the imme- approach to classifying personality and psychopathology. Annual
diate direct clinical utility of the results. Cross-validation in a Review of Clinical Psychology, 10, 477–501.
clinical sample with significant personality pathology will be an Livesley, W. J. (2006). General assessment of personality disorder (GAPD).
important next step for research. Despite such limitations, as Vancouver, Canada: Department of Psychiatry, University of British
noted by Cronbach and Meehl (1955), the concept of boot- Columbia.
Lowmaster, S. E., & Morey, L. C. (2011, March). Support for a core person-
strapping is often necessary in the initial development and sub- ality pathology impairment dimension in an undergraduate sample.
sequent refinement of measures, whereby various indicators Paper presented at the annual meeting of the Society for Personality
might ultimately prove to be more valid than the initial Assessment, Boston, MA.
ASSESSING DSM–5 PERSONALITY PATHOLOGY WITH THE PAI 7

Markon, K. E., Krueger, R. F., & Watson, D. (2005). Delineating the Wright, A. G. C., Thomas, K. M., Hopwood, C. J., Markon, K. E., Pincus,
structure of normal and abnormal personality: An integrative hier- A. L., & Krueger, R. F. (2012). The hierarchical structure of DSM–5
archical approach. Journal of Personality and Social Psychology, 88, pathological personality traits. Journal of Abnormal Psychology, 121,
139–157. 951–957.
Markon, K. E., Quilty, L. C., Bagby, R. M., & Krueger, R. F. (2013). The
development and psychometric properties of an informant-report form
of the Personality Inventory for DSM–5 (PID–5). Assessment, 20, Appendix: PAI regression equations for DSM–5 alter-
370–383. native PD model constructs
Morey, L. C. (1991). Professional manual for the Personality Assessment
Inventory. Odessa, FL: Psychological Assessment Resources. To obtain PAI indicator scores on the DSM–5 alternative
Morey, L. C. (1996). An interpretive guide to the Personality Assessment model, the T scores (using standard PAI community norms)
Inventory. Odessa, FL: Psychological Assessment Resources.
Morey, L. C. (2007). Personality Assessment Inventory professional manual for all indicated PAI scales and subscales should be multiplied
(2nd ed.). Odessa, FL: Psychological Assessment Resources. by the corresponding regression weight and then summed, as
Morey, L. C., Benson, K. T., Busch, A. J., & Skodol, A. E. (2015). Personal- presented in the formulas listed here. All PAI scale and subscale
ity disorders in DSM–5: Emerging research on the alternative model. acronyms are as delineated in the PAI test manual (Morey,
Current Psychiatry Reports, 17(4), 1–9. doi:10.1007/s11920-015-0558-0 2007).
Morey, L. C., Berghuis, H., Bender, D. S., Verheul, R., Krueger, R. F., &
Skodol, A. E. (2011). Toward a model for assessing level of personality
 Submissiveness D (TARDO  .005520) C (TMANI 
functioning in DSM–5, Part II: Empirical articulation of a core dimen- .013345) C (TPARH  –.005810) C (TANTA 
sion of personality pathology. Journal of Personality Assessment, 93, –.007323) C (TBORI  .006113) C (TALC  .004922) C
347–353. (TSTR  –.006354) C (TRXR  –.008159) C (TDOM 
Morey, L. C., Krueger, R. F., & Skodol, A. E. (2013). The hierarchical struc- –.029352) C (TWRM  .009416) C (2.194113).
ture of clinician ratings of proposed DSM–5 pathological personality
traits. Journal of Abnormal Psychology, 122, 836–841.
 Depressivity D (TNIM  .004155) C (TPIM  ¡.003700)
Paunonen, S. V. (1997). On chance and factor congruence following C (TDEPA  .011554) C (TDEPC  .017550) C
orthogonal Procrustes rotation. Educational and Psychological Mea- (TMANG  .003136) C (TMANI  –.005167) C (TDRG
surement, 57, 33–59.  –.002799) C (TSUI  .008055) C (TNON  .003176) C
Samuel, D. B., & Widiger, T. A. (2008). A meta-analytic review of the relation- (TWRM  –.003899) C (–1.345765).
ships between the Five-Factor Model and DSM–IV–TR personality disor-
ders: A facet level analysis. Clinical Psychology Review, 28, 1326–1342.
 Separation insecurity D (TANXC  .006378) C (TMANI
Sellbom, M., Anderson, J. L., & Bagby, R. M. (2013). Assessing DSM–5  .007556) C (TSCZS  –.012544) C (TBORI  .028201)
Section III personality traits and disorders with the MMPI–2–RF. C (TANTS  –.005552) C (TAGGA  .006029) C
Assessment, 20, 709–722. (TNON  .007672) C (TDOM  –.009658) C (–.519245).
Sharp, C., Wright, A. C., Fowler, J. C., Frueh, B. C., Allen, J. G., Oldham, J.,  Anxiousness D (TNIM  –.005511) C (TPIM  –.006017)
& Clark, L. A. (2015). The structure of personality pathology: Both gen-
eral (‘g’) and specific (‘s’) factors? Journal of Abnormal Psychology, 124, C (TANXA  .007727) C (TANXC  .033481) C (TPARH
387–398.  .005490) C (TSCZP  –.004210) C (TBORI  .007347) C
Skodol, A. E. (2011). Scientific issues in the revision of personality disor- (TANTA  –.004620) C (TDOM  –.003988) C
ders for DSM–5. Personality and Mental Health, 5, 97–111. (–.304736).
Skodol, A. E., Morey, L. C., Bender, D. S., & Oldham, J. E. (2013). The  Emotional lability D (TNIM  .008713) C (TSOMC 
ironic fate of personality disorders in DSM–5. Personality Disorders:
Theory, Research and Treatment, 4, 342–349. –.004824) C (TANXA  .009669) C (TANXC  .005332)
Stevens, J. (2002). Applied multivariate statistics for the social sciences (4th C (TPARH  –.006651) C (TPARR  .007335) C
ed.). Mahwah, NJ: Erlbaum. (TBORA  .037376) C (TBORI  .010142) C (TANTA 
Thomas, K. M., Yalch, M. M., Krueger, R. F., Wright, A., Markon, K. E., & –.008144) C (TSUI  –.005632) C (TWRM  .009092) C
Hopwood, C. J. (2013). The convergent structure of DSM–5 personality (–2.250505).
trait facets and Five-Factor Model trait domains. Assessment, 20,
308–311.
 Suspiciousness D (TANXC  .004636) C (TARDO 
Turkheimer, E., Ford, D. C., & Oltmanns, T. F. (2008). Regional analysis of .002381) C (TMANG  .002693) C (TPARH  .012267)
self-reported personality disorder criteria. Journal of Personality, 76, C (TPARP  .015801) C (TPARR  .008535) C (TBORN
1587–1622.  .005899) C (TANTA  –.004247) C (TANTE 
Verheul, R., Andrea, H., Berghout, C., Dolan, C. C., van Busschbach, J. J., .003045) C (TDOM  –.005427) C (–1.438204).
Van Der Kroft, P. J. A., … Fonagy, P. (2008). Severity Indices of Per-
sonality Problems (SIPP–118): Development, factor structure, reliabil-
 Restricted affectivity D (TANXC  –.008010) C (TARDP
ity, and validity. Psychological Assessment, 20, 23–34.  –.006660) C (TDEPA  .009528) C (TMANA 
Weisberg, S. (1985). Applied linear regression (2nd ed.). New York, NY: .006978) C (TPARH  .007878) C (TBORA  –.012317)
Wiley. C (TANTE  .008854) C (TANTS  .009177) C (TAGGV
Westen, D., & Rosenthal, R. (2003). Quantifying construct validity: Two  –.008450) C (TWRM  –.026460) C (1.835072).
simple measures. Journal of Personality and Social Psychology, 84,
608–618.
 Withdrawal D (TPIM  –.003320) C (TARDT  .003969)
Widiger, T. A., & Costa, P. T. (1994). Personality and personality disor- C (TPARH  .003317) C (TSCZS  .026469) C (TANTA
ders. Journal of Abnormal Psychology, 103, 78–91.  –.004228) C (TANTE  .005883) C (TAGGV 
Widiger, T. A., & Simonsen, E. (2005). Alternative dimensional models of –.005817) C (TSUI  –.003035) C (TWRM  –.015729)
personality disorder: Finding a common ground. Journal of Personality C (.423567).
Disorders, 19, 110–130.
Widiger, T. A., & Trull, T. J. (2007). Plate tectonics in the classification of
 Intimacy avoidance D (TNIM  .006642) C (TPIM 
personality disorder. American Psychologist, 62, 71–83. .007632) C (TDEPP  .005648) C (TMANI  –.008867) C
Wiggins, J. S., & Pincus, A. L. (1989). Conceptions of personality disorders (TPARH  .005113) C (TSCZT  .006911) C (TBORS 
and dimensions of personality. Psychological Assessment, 1, 305–316. –.005706) C (TANTE  .018430) C (TALC  .003732) C
8 BUSCH, MOREY, HOPWOOD

(TAGGV  –.005913) C (TWRM  –.012467) C  Impulsivity D (TSOMC  –.005311) C (TARDO 


(–.644794). –.003881) C (TDEPC  .005768) C (TSCZT  .005651)
 Anhedonia D (TDEPA  .026289) C (TDEPC  .005579) C (TPARP  .004304) C (TBORS  .018269) C (TANTA
C (TPARP  –.004497) C (TANTS  .002200) C (TNON  .006842) C (TANTS  .018401) C (TSUI  –.003445) C
 .005400) C (TWRM  –.014167) C (–.122591). (–1.607169).
 Manipulativeness D (TANXP  –.004399) C (TMANG   Distractibility D (TPIM  –.011481) C (TARDO 
.006847) C (TMANI  .007985) C (TPARR  .005714) C –.006347) C (TARDT  –.004194) C (TDEPC  .018527)
(TBORS  .006571) C (TANTA  .008268) C (TANTE  C (TMANG  .004864) C (TPARP  –.006178) C
.021988) C (TAGGV  .005879) C (TDRG  –.004059) (TSCZP  –.005413) C (TSCZT  .026363) C (TBORS 
C (TDOM  .005185) C (–2.020781). .004762) C (TANTS  .004191) C (TSUI  –.007085) C
 Deceitfulness D (TPIM  –.006776) C (TARDT  –.006000) (TDOM  –.008551) C (.477284).
C (TPARR  .007805) C (TBORA  .004758) C (TBORS   Perseveration D (TPIM  –.007551) C (TARDO 
.004780) C (TANTA  .014409) C (TANTE  .017846) C .006440) C (TARDP  .004509) C (TDEPC  .008228) C
(TDRG  –.005428) C (TWRM  –.005229) C (–.625298). (TMANG  .005658) C (TMANI  .005453) C (TPARR
 Hostility D (TPIM  –.011202) C (TANXC  –.003552)  .003130) C (TSCZT  .013790) C (TBORA  .005268)
C (TMANI  .012324) C (TPARR  .003906) C (TBORA C (TSUI  –.006588) C (TDOM  –.009031) C
 .012637) C (TDRG  –.004163) C (TAGGA  .010586) (–.560214).
C (TAGGV  .010102) C (TRXR  .004766) C (TWRM  Rigid perfectionism D (TANXC  .004767) C (TARDO 
 –.007767) C (–.390267). .043882) C (TARDP  .004911) C (TARDT  –.006748)
 Callousness D (TARDT  –.002781) C (TDEPA  C (TPARR  .006521) C (–1.520032).
.004529) C (TDEPC  .003582) C (TDEPP  –.002813)  Risk taking D (TSOMC  –.004372) C (TARDO 
C (TMANG  .004133) C (TPARP  .004414) C (TSCZP –.005489) C (TARDP  –.008047) C (TMANI 
 .003645) C (TBORI  –.003510) C (TANTA  .006537) –.002802) C (TSCZS  –.003053) C (TBORS 
C (TANTE  .007457) C (TAGGP  .009431) C .003992) C (TANTS  .030160) C (TDRG  .003772)
(TAGGV  .005411) C (TSUI  –.003047) C (TRXR  C (TAGGV  .004242) C (TDOM  .003304) C
.004306) C (TWRM  –.011116) C (–1.100108). (.119584).
 Attention seeking D (TDEPP  –.006087) C (TMANA   Eccentricity D (TNIM  –.008873) C (TPIM  –.006347)
.006489) C (TMANG  .014967) C (TMANI  .005221) C (TARDO  –.005108) C (TDEPC  .007788) C
C (TPARR  .005334) C (TSCZS  –.007167) C (TBORA (TMANA  .008291) C (TMANG  .009087) C (TSCZP
 .007999) C (TANTE  .008072) C (TALC  .005530) C  .017789) C (TSCZT  .007732) C (TBORA  .008147)
(TAGGV  .005569) C (TNON  .007573) C (TWRM  C (TBORS  –.006573) C (TANTS  .012707) C (TALC
.016511) C (–2.473992).  –.006430) C (TRXR  –.006484) C (TWRM 
 Grandiosity D (TARDT  –.005302) C (TMANG  –.009265) C (–.102238).
.023614) C (TMANI  .005727) C (TPARP  .004050) C  Perceptual dysregulation D (TSOMC  .005476) C
(TPARR  .004483) C (TSCZP  .004800) C (TSCZS  (TDEPC  .005083) C (TMANG  .003798) C (TSCZP 
.007054) C (TANTE  .005448) C (TAGGV  .004619) .010402) C (TSCZT  .014835) C (TANTA  .002809) C
C (TRXR  .006211) C (–2.260406). (TANTS  .002752) C (TWRM  –.006233) C
 Irresponsibility D (TARDO  –.005286) C (TARDP  (–1.333644).
.003907) C (TARDT  –.003692) C (TDEPC  .005077)  Unusual beliefs D (TSOMC  .009420) C (TMANG 
C (TPARP  .007542) C (TPARR  .003215) C (TSCZT .009505) C (TSCZP  .025136) C (TSCZT  .009724) C
 .006791) C (TBORN  –.005237) C (TBORS  (TANTA  .004603) C (TALC  –.006416) C (TWRM 
.006072) C (TANTA  .003726) C (TANTE  .006465) –.007738) C (–1.508141).
C (TALC  .003672) C (TDRG  .003150) C (TAGGP   PD severity D (TBORA  .962733) C (TBORI 
.003562) C (TWRM  –.004271) C (–1.328749). 1.244508) C (TBORN  .358478) C (–16.937825).

You might also like