Normative data for PID-5 domains, facets, and personality disorder composites from a representative sample and comparison to community and clinical samples

Abstract

The introduction of the Alternative Model of Personality Disorders (AMPD) in the fifth edition of the Diagnostic and Statistical Model of Mental Disorders (DSM-5, APA, 2013) represented a substantive change in how personality disorders (PDs) are diagnosed. One barrier to its adoption (among several) in clinical practice, however, is a lack of information as to what constitutes an elevated score on the 25 domains and facets that comprise Criterion B. Unique sets of facets can be configured to assess any one of six PDs retained in the AMPD; each of these facets can in turn be added to create a PD sum score. In the current study, using the Personality Inventory for DSM-5 (PID-5; Krueger et al., 2012), we report mean scores using this instrument that align with 1.0, 1.5, and 2.0 standard deviation elevations for each of these six PDs on the basis of Krueger and colleagues (2012) representative sample, and compare these to those obtained from a community and a clinical sample. These normative data may be useful to clinicians in determining whether a client has elevated scores on pathological personality domains, facets, or PDs.

Following decades of criticism surrounding the diagnostic system used in the Diagnostic and Statistical Manual of Mental Disorders (DSM; e.g., APA 1980, 1994) for personality disorders (PDs; e.g., Clark, 2007; Trull & Widiger, 2007), the DSM-5 (APA, 2013) included the Alternative Model of Personality Disorders (AMPD). This model was developed by the DSM-5 Personality and Personality Disorder Work Group with the goal of replacing the PD diagnostic system in DSM-IV, which was widely regarded as deeply flawed (Trull & Widiger, 2007). Despite undergoing field trials and existing support for similar models (e.g., Widiger & Costa, 2013), the AMPD was rejected as it was judged to be too new and untested by key oversight and decision-making groups within the American Psychiatric Association. As a result, the AMPD was placed in Section III of the DSM-5 -- “Emerging Measures and Models” with the explicit intent to encourage research to test this model.

The AMPD uses personality dysfunction (Criterion A) in two of four domains; (i.e., Self: Identity; Self-direction; Interpersonal: Empathy; Intimacy) and pathological personality traits (Criterion B; [i.e., elevations on one or more of 25 traits; e.g., Depressivity, Grandiosity, Eccentricity]) as the core defining features of PDs along with inflexibility, pervasiveness and relative stability across time (i.e., Criteria C and D). The pathological trait model largely aligns with other existing five factor models of personality and personality psychopathology including the Five Factor Model of personality (FFM) and the Personality Psychopathology Five (PSY-5) (see e.g., Anderson et al., 2013; Gore & Widiger, 2013; Watters & Bagby, 2018). Different configurations of the 25 pathological traits can be used to diagnose one of six PDs retained in the AMPD (e.g., for Narcissistic PD: summation of attention seeking and grandiosity), which generally correspond with DSM-IV/ DSM-5 Section II PD symptom counts (e.g., Morey et al., 2016; Yam & Simms, 2014) and largely recreate their nomological networks (e.g., Miller et al., 2015). For patient presentations that do not fit one of these six PDs, practitioners can use a PD – Trait Specified diagnosis.

Of the two major AMPD PD criteria, Criterion B (Pathological Personality Traits) has received more attention than Criterion A (Level of Personality Functioning), likely due to the existence of an APA sanctioned instrument developed and widely disseminated even before the release of the DSM-5 -- the Personality Inventory for DSM-5 (PID-5). This self-report instrument was designed and developed by one of the key members of the DSM-5 PD Work Group and this member’s colleagues. Instruments designed to assess Criterion A were not developed until much later and after the publication of DSM-5 (e.g., Morey, 2017). Although Criterion B has been extensively examined and has the potential to advance clinical practice using an empirically informed set of pathological traits, it is infrequently used in clinical practice due to a number of concerns including the lack of validity scales to detect problematic response bias (Hopwood & Sellbom, 2013) and criticism regarding the anchoring used for the Likert scale (Al-Dajani, et al., 2016). Another important barrier to its clinical use is a general lack of clarity as to how one should determine whether a person scores at a pathological level of a given trait, which is key to diagnosing one of the six PDs retained in the AMPD (i.e., Schizotypal, Antisocial, Borderline, Narcissistic, Avoidant, Obsessive-Compulsive) or the use of the PD – Trait Specified diagnosis.

Much like the approach articulated by Widiger and Mullins-Sweatt (2009) for the Five-Factor Model of PDs in which high scores on the NEO PI-R are potentially linked to impairment, we suggest that for the AMPD one could similarly assess each trait, identify the traits for which there is some meaningful degree of elevations, and then determine whether these elevations are associated with clinically significant impairment (e.g., in relational, occupational, or school functioning). Such an approach, however, requires information as to what constitutes an elevation on a given trait. Using the 220-item PID-5 or more abbreviated versions (e.g., 100-items: Maples et al., 2015), one could choose a point on the 0 to 3 Likert scale on this instrument as a threshold. For instance, a score of “2” (Sometimes or Somewhat True) or “3” (Very True or Often True) could be taken as evidence of clinically meaningful elevation (e.g., Samuel et al., 2013). One difficulty with such an approach, however, is the lack of information on endorsement rates – that is, how often do individuals actually have a score that elevated? For that determination, one needs information on the distribution of responses in the population. For instance, using PID-5 data from Krueger et al. (2012; “round 2”), one would find that a score of 2 on the facet Depressivity corresponds to the 98th percentile among a sample of individuals who had received psychiatric/psychological care, which suggests that a 2 on this trait is quite extreme. Even for the facet of Submissiveness, which had the highest mean score in the Krueger and colleagues’ representative sample, a score of 2 represents a small proportion of individuals (i.e., 88^th percentile).¹

Given the quantitative strategies used in the vast majority of personality research, a lack of normative data for a given assessment has no real consequences. Most research questions use continuous data to ask questions about the association of variables or relative elevations; in such instances there is no need for normative data. However, such data are important in clinical and forensic settings in which a particular individual is being assessed and diagnosed. Clinicians require normative data to contextualize an individual’s score and determine diagnoses.

Personality scores will hold little relevance to clinicians and clients without the context provided by normative comparisons. Normative data exist for many personality-related measures including the various iterations of the Minnesota Multiphasic Personality Inventory (e.g., MMPIRF; MMPI-3), Personality Assessment Inventory (PAI, Morey, 1991) and NEO Personality Inventory – 3 (McCrae, Costa, & Martin, 2005). However, these are proprietary and pay to use instruments, which can have an inhibitory effect on their use in both research and clinical settings and none were developed with the goal of assessing the DSM-5 AMPD.

Current Study

In the current study, we use information drawn from Krueger and colleagues (2012) report on the representative sample (N = 264; “round 3”) to present means and standard deviations for the 25 AMPD facet traits, five domain traits, and six PD trait composites included in the AMPD, as well as the scores that would fall at 1.0, 1.5, and 2.0 SDs above the mean. We provide this same information from two moderately-sized samples – one a community sample with a history of mental health treatment and the other an outpatient psychiatric sample. This study was not pre-registered.

Participants

Community Sample.

Participants were recruited by distributing flyers at mental health clinics across Western New York–were eligible to participate if they reported psychiatric treatment within the past two years (Yam & Simms, 2014). The final sample included 628 participants, 454 of whom completed the measures used here. The average age of the sample was 42.0 years (SD=12.6, range = 18 to 77), 65% female, 68% Caucasian. This subsample differed from the full sample in terms of race, χ 2 (4, N=624) = 28.75, p < .01, and age, t(626) = 3.68, p < .01, but not sex, χ 2 (1, N=627) = 1.56, p = .21. Those excluded were more likely to be African American and older. Participants attended a four-hour session and completed self-report measures using computers in privacy carrels. Procedures were approved by the Social and Behavioral Sciences IRB at the University at Buffalo.

Psychiatric Sample.

Participants were recruited from a patient research registry maintained at a university-affiliated addictions and mental health hospital in Toronto, Canada. This sample included 435 participants (who provided informed consent and completed the PID-5). The use of this patient data for research purposes was approved by the Research Ethics Board (REB) at the Centre for Addiction and Mental Health. The average age of the sample was 43.21 years (SD = 13.78, range = 19 – 81), 50.3% (219) were female, and 68.3% (297) self-identified as White or Caucasian. The sample was diagnostically heterogeneous; the most common primary diagnoses in descending order were depressive disorders (32.0%), bipolar disorders (13.7%), anxiety disorders (13.3), schizophrenia (7.3%), and borderline personality disorder (7.1%). The PID-5 data used for this study have been a part of several other publications (e.g., Sellbom, Solomon-Krakus, Bach, & Bagby, 2020; Watters, Sellbom, Bagby, 2019); however, the analyses and research questions used here are novel for these data.

Measures

Personality Inventory for DSM-5 (PID-5; Krueger et al., 2012), a 220-item questionnaire with a four-point response scale, was used to measure the DSM-5 AMPD personality facet traits. In Sample 1, omegas ranged from .75 to .96 (Mdn = .88) across the facets. In Sample 2, omegas ranged from .58 to .87 (Mdn = .77).

Results

PID-5 items are scored on a four-point scale ranging from 0 to 3 with Likert anchors of Very False or Often False (0), Sometimes or Somewhat False (1), Sometimes True or Somewhat True (2), and Very True or Often True (3). Using the normative data presented in Krueger et al. (2012), we present the values for the PID-5 facets and domains that represent values 1.0, 1.5, and 2.0 standard deviations above the mean (see left hand portion of Table 1). Within these data, scores above 1.0 are rare; only 5 facets have average scores above 1. In fact, at 1.0 SD above the mean, no facets have a mean of 2.0. Using a cut-off of 1.5 SDs, which is equal to a T-score of 65 and is considered clinically significant on the MMPI-2-RF, only 5 of 25 facets have a mean of 2.0 or higher (mean 1.74). One does not find scores approximately 2.0 or higher until working with scores that are 2 SDs above Krueger et al.’s means. Even at 2.0 SDs above the mean, several facets have scores that are substantially lower than this value including Irresponsibility (M = 1.37), Callousness (M = 1.40), Perceptual Dysregulation (M = 1.40), Deceitfulness (M = 1.60), and Depressivity (M = 1.77). In terms of domains, at 2.0 SDs, only one PID-5 domain has a mean > 2.0 (i.e., Negative Affectivity).

Table 1:

PID-5 descriptive data across three samples including cut-offs associated with 1, 1.5 and 2 standard deviation elevations

	Normative Sample					Community Sample		Clinical Sample		Effect size (d)b
	Mean	SD	1 SD	1.5 SD	2 SD	Mean	SD	Mean	SD	Norm v Comm	Norm v Clin
Domains
Negative Affectivity	0.92	0.62	1.54	1.85	2.16	1.32	0.69	1.41	0.51	−0.60**	−0.86**
Detachment	0.84	0.56	1.40	1.68	1.96	1.05	0.59	1.26	0.63	−0.36**	−0.71**
Antagonism	0.71	0.52	1.23	1.49	1.75	0.72	0.54	0.71	0.47	−0.02	0.00
Disinhibition	0.66	0.50	1.16	1.41	1.66	0.97	0.60	1.31	0.39	−0.55**	−1.44**
Psychoticism	0.63	0.56	1.19	1.47	1.75	0.77	0.60	0.94	0.62	−0.24*	−0.53**
Facets
Anhedonia	0.89	0.64	1.53	1.85	2.17	1.22	0.77	1.46	0.80	−0.46**	−0.79**
Anxiousness	1.02	0.73	1.75	2.12	2.48	1.52	0.80	1.69	0.84	−0.65**	−0.86**
Attention Seeking	0.81	0.65	1.46	1.79	2.11	0.94	0.75	0.95	0.75	−0.18	−0.20
Callousness	0.40	0.50	0.90	1.15	1.40	0.45	0.49	0.47	0.47	−0.10	−0.14**
Deceitfulness	0.52	0.54	1.06	1.33	1.60	0.62	0.56	0.77	0.68	−0.18	−0.41**
Depressivity	0.53	0.62	1.15	1.46	1.77	0.95	0.72	1.23	0.83	−0.61**	−0.96**
Distractibility	0.86	0.69	1.55	1.90	2.24	1.22	0.77	1.43	0.80	−0.49**	−0.77**
Eccentricity	0.82	0.76	1.58	1.96	2.34	1.00	0.82	1.17	0.81	−0.23*	−0.45**
Emotional Lability	0.94	0.74	1.68	2.05	2.42	1.41	0.82	1.46	0.86	−0.59**	−0.65**
Grandiosity	0.82	0.58	1.40	1.69	1.98	0.71	0.60	0.70	0.61	0.19	0.20
Hostility	0.91	0.67	1.58	1.92	2.25	1.09	0.69	1.17	0.70	−0.26*	−0.38**
Impulsivity	0.77	0.57	1.34	1.63	1.91	1.01	0.73	1.12	0.78	−0.36**	−0.52**
Intimacy Avoidance	0.61	0.65	1.26	1.59	1.91	0.72	0.73	0.97	0.82	−0.16	−0.49**
Irresponsibility	0.39	0.49	0.88	1.13	1.37	0.66	0.58	0.77	0.64	−0.49**	−0.67**
Manipulativeness	0.80	0.67	1.47	1.81	2.14	0.84	0.72	0.85	0.76	−0.06	−0.07
Perceptual Dysregulation	0.44	0.48	0.92	1.16	1.40	0.66	0.57	0.79	0.58	−0.41**	−0.66**
Perseveration	0.82	0.62	1.44	1.75	2.06	1.11	0.67	1.22	0.68	−0.44**	−0.62**
Restricted Affectivity	0.97	0.56	1.53	1.81	2.09	0.90	0.59	0.97	0.69	0.12	0.00
Rigid Perfectionism	1.05	0.68	1.73	2.07	2.41	1.28	0.73	1.12	0.72	−0.32**	−0.10
Risk Taking	1.05	0.51	1.56	1.82	2.07	1.22	0.59	1.19	0.59	−0.30**	−0.25
Separation Insecurity	0.80	0.68	1.48	1.82	2.16	1.04	0.78	1.00	0.81	−0.32**	−0.27*
Submissiveness	1.17	0.66	1.83	2.16	2.49	1.26	0.73	1.33	0.76	−0.13	−0.23
Suspiciousness	0.95	0.58	1.53	1.82	2.11	1.18	0.68	1.23	0.64	−0.36**	−0.46**
Unusual Beliefs and Experiences	0.64	0.63	1.27	1.59	1.90	0.63	0.62	0.77	0.67	0.02	−0.20
Withdrawal	1.01	0.72	1.73	2.09	2.45	1.22	0.73	1.33	0.79	−0.29**	−0.42**
AMPD Composites
Schizotypal (6)a	4.83	2.87	7.70	9.13	10.57	5.61	3.01	6.26	3.06	−0.26*	−0.48**
Antisocial (7)	4.84	3.01	7.85	9.35	10.85	5.89	3.26	6.34	3.42	−0.33**	−0.47**
w/ PS (10)	9.62	2.90	12.52	13.97	15.42	10.08	3.38	10.26	3.55	−0.14	−0.20
Borderline (7)	6.02	3.25	9.27	10.90	12.52	8.24	3.78	8.85	3.95	−0.62**	−0.79**
Narcissistic (2)	1.63	1.09	2.72	3.26	3.80	1.65	1.17	1.64	1.17	−0.02	−0.01
Avoidant (4)	3.53	2.24	5.77	6.89	8.00	4.68	2.32	5.44	2.51	−0.50**	−0.81**
OCPD (4)	3.45	1.82	5.27	6.18	7.08	4.02	1.92	4.27	1.99	−0.26**	−0.48**

Note. Normative data drawn from Krueger et al., 2012. Means and SDs for the domains differ from those reported by Krueger et al. as the current values reflect the updated official scoring protocol (i.e., domains scored as the means of only 3 facets per domain). Ns = 264 ,454, and 234 for the Normative, Community, and Clinical samples respectively.

^a.Number in parentheses indicates the number of facets included in the PD composite.

^b.Asterisks indicate significance level for two-tailed independent samples t-tests.

^*indicates p < .005,

^**indicates p < .0005.

Table 1 also presents means and standard deviations for the community and clinical samples, as well as effect sizes (d) and results from independent samples t-tests for the differences between these two samples and the normative sample. In general, and as expected, mean scores for domains and facets were lower in the normative sample than in the community sample with Cohen’s ds ranging from −.60 to −.02 with a mean of −.35 for the domains and from −.65 to .19 with a mean of −.28 for the facets. Four of the 5 domains and 16 of the 25 facets differed statistically significantly at p < .01, two-tailed. Although generally higher than the mean scores for the normative sample, the mean scores for the community sample are still not that high relative to the PID-5 anchors. Only 2 of the 5 domains and 14 of the 20 have average scores above 1.0; none have a mean of 2.0 on the 0 to 3 scale. In fact, scores of 2 represent, on average, 1.75 and 1.53 SDs above the community sample mean.

The case is much the same for the clinical sample. Means for the domains are, on average, .71 d units lower in the normative sample (range from −1.44 to 0); 4 of the 5 domains and 16 of the 25 facets differed statistically significantly at p < .01, two-tailed. Even in this sample, only 3 of the 5 domains and 14 of the 25 facets have average scores above 1.0; none have average scores above 1.5 on the 0 to 3 scale. In this sample, scores of 2 (i.e., “Sometimes True or Somewhat True”) are reached, on average, at 1.71 and 1.33 SDs above the clinical sample mean for domains and facets respectively.

The bottom of Table 1 presents the means and SDs for the six PDs scored in the DSM-5 AMPD, as well as the scores at 1.0, 1.5, and 2.0 SDs for the normative sample.² As in the top portion of the table, for purposes of comparison, we also present the means and SDs for the community and clinical samples, as well as effect sizes (d) and results from independent samples t-tests for the differences between these two samples and the normative sample. Given the number of facets included – from a low of 2 (Narcissistic PD) to a high of 10 (Antisocial PD with the Psychopathy specifier), the means are not comparable across PD trait composites in the same way as the basic facets.³ The data in Table 1 can be used, however, to examine the degree of statistical deviance of an individual PD score, in the same way one could use the information in the Table to understand the extremity of the 25 basic traits and the 5 overarching domains. In general, the PD composites were generally higher in the community (d’s ranged from −.02 to −.62) and clinical samples (d’s ranged from −.01 to −.81), although there were no substantive differences for Narcissistic PD and Antisocial PD with the Psychopathy specifier.

Discussion

The inclusion of a substantively transformed diagnostic system for PDs in DSM-5 was long overdue and for many there was disappointment that the proposed model (the AMPD) was relegated to Section III (“Emerging Models and Measures”) rather than being placed in the main text of the DSM-5 (i.e., “Section II”). Yet, as noted by Widiger (2013) it should be viewed as an important incremental step toward improving the diagnostic approach to PDs. Heeding the APA call for further study of the AMPD, this model has been highly generative with the literally hundreds of empirical research papers testing various aspects of this new model (Zimmerman et al., 2019). Despite this effort it is not clear that it has had much effect on it’s applied assessment and diagnosis. In our opinion, the lack of normative data to guide interpretation of measures of Criteria A and B has been a substantial obstacle to the instruments use in applied settings given that clinicians have no way of contextualizing a given individual’s scores. As a result, clinicians may be more likely to use instruments like the MMPI-2-RF, PAI, or NEO PI-R that provide information as to the degree of non-normativeness of a client’s scores on constructs of interest.

Our primary aim of this study was to transform the data from Krueger and colleagues’ (2012) representative sample so as to present values at 1.0, 1.5 and 2.0 SDs above the mean, as well as sum those facets into their PD assignments articulated in the DSM-5 AMPD (e.g., Narcissistic PD = Attention Seeking + Grandiosity) and present those values at 1.0, 1.5, and 2.0 SDs. We similarly present the same data from a community sample with a history of mental health treatment and a clinical sample to facilitate comparison to the normative sample. As expected, scores in the community and clinical samples were higher than scores observed in the normative sample. These values from the normative sample can be used by clinicians to quickly and easily determine if a client’s scores on Criterion B – both individual traits (e.g., emotional lability) or as PD constructs (e.g., borderline PD) represent meaningful, at least in the statistical sense, deviations towards the pathological end. Armed with this information, interested mental health practitioners can determine if further assessment is needed and incorporate this information into their case conceptualization, diagnosis, and treatment planning.

In general, for many clinicians, we believe working at the individual facet or domain level may be more helpful than using the PD trait composites. For instance, scores of 1.0 or higher are relatively infrequent for the Antagonism domain and facets (e.g., Callousness; Deceitfulness) and thus scores of 1.5 (close to 2.0 SDs above the norm) or higher would tell the clinician that this domain may well be related to substantial impairment in the client’s interpersonal functioning (e.g., Lynam & Miller, 2019) and may even play out in the treatment setting. Conversely, Negative Affectivity and its facets (e.g., Anxiousness; Emotional Lability) tend to be much higher on average across the current sampling approaches. In general, higher scores on PID-5 domains are associated with poorer psychological functioning (e.g., Anderson et al., 2018; Wright et al., 2012), as one would expect, although the specific impairments associated with these traits and/or PD composites would have to be assessed specifically for each client if this approach was used in an applied setting. One could use the PID-5 and these norms in a multi-step approach like that outlined by Widiger, Costa, and McCrae (2002). Step 1 requires the assessment of a client’s personality; here, Widiger and colleagues were describing the more generally-oriented FFM domains and facets, but the PID-5 could be used for this purpose. Using the current data, practitioners could determine which traits are statistically elevated enough to proceed to Step 2, in which the practitioner works with the client to identify “problems, difficulties, and impairments that are second to each trait” (p. 435). That is, they could decide that a score of 1.80 on Grandiosity represents a potentially meaningful elevation as it falls above the 1.5 SD norm from the normative sample and thus decide to assess if problems are associated with this elevation. At Step 3, the practitioner would determine if any impairments identified are clinically significant in relation to distress caused or functional difficulties. Finally, at Step 4, one can decide if the traits match one of the six AMPD PD configurations.

Overall, we believe the current data will prove useful to clinicians who are interested in adopting the DSM-5 AMPD approach and/or want to use the PID-5 measure to inform their clinical work by providing explicit contextual information surrounding the degree of elevation of a client’s scores. Clinicians are much more likely to use such a tool in clinical work when there are normative data like those presented here that can be used to contextualize a client’s scores. If this process proves helpful to those working in applied settings it could be applied to alternative measures in PID-5 family including the 100-item version (Maples et al., 2014), the PID-5 Brief Format (25-items; Krueger et al., 2013) and the PID-5 Informant Report (Markon et al., 2013). We note that the current values from the representative sample are from a relatively small sample (N = 264) and thus should be viewed with some caution. The means across facets from this sample are relatively closely aligned with the means from the other samples in terms of relative elevations (profile rs = .83 and .69, respectively) suggesting some general consistency in which traits and domains tend to be higher or lower across sampling approaches (i.e., higher Negative Affectivity means, lower Antagonism means). That being said, it would be ideal if data were collected from a much larger representative sample at some point, which could be used to provide revised estimates. Additionally, it would be helpful if representative samples from other regions were collected so that they could be used in clinical settings outside North America. Nonetheless, we hope the current data help transform the PID-5 from an entirely research based tool to one that has some clinical utility in applied settings.

Public Significance Statement:

In this study we present different cut scores for a commonly used measure of personality pathology, the PID-5, based on the representative sample presented with the initial study (Krueger et al., 2012). These scores can be used to assist clinicians in determining whether a client has an elevated score on a pathological trait domain, facet, or personality disorder construct, relative to a non-patient normative sample.