The Brave New World of Personality Disorder-Trait Specified: Effects of Additional Definitions on Coverage, Prevalence, and Comorbidity

Abstract

The alternative dimensional model for personality disorder (PD) in DSM-5, Section III (DSM-5-III) includes two main criteria: (A) personality-functioning impairment, and (B) personality-trait pathology; provides specific functioning-and-trait criteria for six PD-type diagnoses; and introduces PD-trait specified (PD-TS), which requires meeting the general PD criteria and not meeting criteria for any specific PD type. We termed this Simple PD-TS and developed two additional definitions: Mixed PD-TS, meeting criteria for one or two PD types and having five or more additional pathological traits; and Complex PD-TS, meeting criteria for three or more PD types. In a mixed sample of 165 outpatients and 215 community adults screened to be at high-risk for PD, we investigated the effect of these additional definitions on prevalence, coverage, comorbidity, and within-diagnosis heterogeneity, and conclude that eliminating the PD-type diagnoses and thus having PD-TS as the only PD diagnosis would be both more parsimonious and more useful clinically.

An alternative model for personality disorder (PD) appears in Section III, Emerging Measures and Models, of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5-III; American Psychiatric Association [APA], 2013), whereas the main text, Section II (DSM-5-II), has retained the DSM-IV-TR (4th ed., text rev.; APA, 2000) categorical diagnostic system for PD¹. The alternative model was created to address the well-documented major shortcomings of the DSM PD categorical system (e.g., excessive comorbidity, within-PD heterogeneity; Clark, 2007; Skodol et al., 2011).

The DSM-5-III model adopts a dimensional approach to conceptualizing PD and includes two main criteria: (A) personality-functioning impairment, and (B) personality-trait pathology. Criterion A, personality dysfunction, is conceptualized as a hierarchical construct with two higher order domains: self, including the two subdomains of identity and self-direction, and interpersonal, including the two subdomains of empathy and intimacy. Criterion B, personality-trait pathology, is based on a hierarchical trait model with five domain-level traits and 25 facet-level traits.

The model could have been conceptualized as purely dimensional, with a PD diagnosis specified by individuals’ domains and severity of both their personality-functioning impairment and their pathological trait(s). In such an idealized model, a single PD diagnosis, PD-trait specified (PD-TS), would fully capture individuals’ unique characteristics. However, to ease the transition to the new conceptualization, the DSM-5 Personality and PD Work Group derived a hybrid model by mapping six of the DSM-5-II categorical diagnoses (antisocial, avoidant, borderline, narcissistic, obsessive-compulsive, and schizotypal) using the DSM-5-III model of personality dysfunction and pathological traits. Four disorders were excluded from this hybrid model on the basis of the simplicity of their trait component (e.g., histrionic PD is essentially captured by emotional lability and attention seeking) and limited empirical validity (Clark, 2013; Skodol et al., 2011). Finally, a diagnosis of PD-TS was provided to denote personality trait-and-functioning pathology in the absence of one of the six specified PDs.

Problems with the Current Categorical PD Diagnostic System

Comorbidity

As mentioned above, current categorical approaches to PD diagnosis have many drawbacks. First, comorbidity is widely known as a pervasive issue in the classification of all mental disorders, but it particularly causes difficulties with both PD conceptualization (i.e., individuals have one personality², so what does it mean to have two or more PDs?) and interpreting PD research results as discussed further below. Problems associated with comorbidity exist not only among PDs, but also between PD and clinical syndromes, with typical rates in clinical samples of 50% or more in both cases (see Clark, 2005, 2007; Widiger & Shea, 1991, for discussions of these issues).

Within-PD comorbidity rates are high (1) in non-patient as well as patient samples (Zimmerman & Coryell, 1989), (2) using different assessment instruments (Oldham et al., 1992), and (3) across various Western cultures (Marinangeli et al., 2000). This excessive comorbidity undermines the validity and utility of the DSM-5-II PD diagnostic system, in which the 10 PD criterion sets are intended to reflect “qualitatively distinct clinical syndromes” (APA, 2013, p. 646). However, if the DSM-5-II PDs are distinct disorders with specific etiologies, then the frequency of comorbidity would be simply the probability of the disorders co-occurring by chance (i.e., the product of their baserates). That PD comorbidity far exceeds this rate indicates that these putatively discrete disorders actually share substantial common variance. Considering overlapping disorders as separate entities leads to misinterpreting findings as limited to specific PDs when, in fact, the findings are more broadly generalizable. Thus, failure to recognize that the DSM-5-II PDs share considerable variance hinders our further understanding of the PD domain.

For example, Bornovalova, Hick, Iacono, and McGue (2013) found that common risk factors explained substance abuse–borderline PD (BPD) comorbidity, whereas Jahng and colleagues (2011) demonstrated that there is considerable shared pathology in substance dependence-PD comorbidity. Given the known high within-PD comorbidity, the latter findings indicate that the former findings should not be interpreted solely in terms of BPD, but likely should be attributed to personality pathology more generally. This is merely one example of literally thousands of studies that purport to concern BPD (the most studied PD; Blashfield & Intoccia, 2000), but whose findings likely are not limited to BPD, but apply more broadly to personality pathology in general.

Similar conclusions can be drawn about a wide range of personality and clinical pathology. For example, the personality trait of neuroticism has been shown not only to account for comorbidity between anxiety and depressive disorders (Middeldorp, Cath, Van Dyck, & Boomsma, 2005) but, more generally, to underlie comorbidity of these disorders with PD (e.g., Hink et al., 2013; Jylhä, Melartin, & Isometsä, 2009). In contrast to approaches that rely on categorical diagnoses, a dimensional classification system would allow the complexity of personality pathology that is evidenced by high comorbidity rates to be captured fully and represented with greater clarity and specificity, thus furthering our understanding of a wide range of psychopathology.

Within-PD heterogeneity

Another shortcoming of the current categorical approach to PD diagnosis is that it leads to within-diagnosis heterogeneity; individuals with the same PD diagnosis may have significantly different sets of traits. For example, there are 126 different ways to meet the required five of nine criteria for DSM-5-II BPD, and two individuals both diagnosed with BPD could share only a single criterion, so the diagnosis is not necessarily informative. Lenzenweger, Clarkin, & Yeomans (2008) reported, based on finite mixture modeling, that there may even be three distinct BPD subgroups. Within-diagnosis heterogeneity has been documented specifically for narcissistic and antisocial PD as well (Cunliffe & Gacono, 2005; Curtis & Susman, 1994), but beginning with DSM-III-R, all PD diagnoses have used a polythetic system, in which a diagnosis requires x of y criteria (e.g., five of nine, as in the example above), so heterogeneity assuredly occurs in all PD categories, which limits the clinical utility of PD diagnoses, especially when confounded by comorbidity.

It must be noted that the use of a polythetic system is not unique to PD, but rather is used throughout the DSM. However, with such disorders as major depression, panic disorder, and so on, all the criteria are intended to reflect the same psychopathology, whereas the criterion sets of the specific PDs in DSM-5-II do not necessarily all reflect the same trait, and some criteria reflect acute dysfunctional behaviors rather than traits (McGlashan et al, 2005). For example, schizotypal PD is said to reflect “social and interpersonal deficits marked by acute discomfort with, and reduced capacity for, close relationships” and “cognitive or perceptual distortions and eccentricities of behavior” (APA, p. 2013, p. 655); five of nine manifestations are required for diagnosis. However, because five of the nine criteria reflect cognitive or perceptual distortions and eccentricities of behavior, some individuals may receive this diagnosis without meeting any of the criteria related to social and interpersonal deficits, whereas others may have marked social and interpersonal deficits and minimal cognitive/ perceptual distortions or eccentricity.

Poor coverage and PD-not otherwise specified

One of the most significant problems with the DSM-5-II PD diagnostic system is the poor coverage that the system provides, necessitating considerable use of PD-not otherwise specified (PD-NOS)³. The guidelines for diagnosing PD-NOS (i.e., meeting the general diagnostic criteria for PD but not for any specific PD) are such that the category is necessarily heterogeneous. For example, PD-NOS may be used to denote clinically significant features of (1) multiple PD categories (i.e., mixed features in the absence of meeting full criteria for any specific PD), (2) a specific PD category not included in the manual (e.g., depressive PD), or (3) a single PD category but not fully meeting the criteria (i.e., atypical PD) (Verheul & Widiger, 2004; Wilberg, Hummelen, Pedersen, & Karterud, 2008). Consequently, there is a wide range of reported prevalence rates for PD-NOS, depending on how it is defined and whether co-occurrence with a specific PD is allowed. A meta-analysis by Verheul and Widiger (2004) found that PD-NOS was the most frequently used category when assessed with non-structured interviews, accounting for 22%–30% of all PD diagnoses in non-patient and 20%–50% in patient samples, with overall prevalences of 4%–6% in non-patient, and 8%–32% in patient samples.

Since that meta-analysis was published, Wilberg et al. (2008) investigated a sample of 1516 inpatients, 80% of whom had one or more PD diagnoses. Using the DSM-IV-TR definition of PD-NOS, they reported a prevalence of 17%, the third most frequent PD diagnosis (after avoidant and borderline PD), comprising nearly a quarter (22%) of all PD diagnoses. Severity of psychosocial impairment in this group was intermediate between no PD and one or more specific PDs. However, severity increased and prevalence decreased if PD-NOS was defined more stringently as either one criterion below threshold on two or more specific PDs, or as meeting 10 or more PD criteria. Thus, how PD-NOS is defined affects the import of the diagnosis.

Similarly, Verheul, Bartak, and Widiger (2007) investigated 1760 patients referred for psychotherapy and found that 21.6% met criteria for PD-NOS, when defined as (1) meeting criteria for either 10 or more PD criteria (mixed PD) or a specific “other” PD (depressive, passive-aggressive, or self-defeating PD), and (2) not meeting criteria for any of the 10 “official” PDs. Notably, the prevalence rate rose to 70.8% if the latter restriction was removed. Mixed PD was the most frequent subcategory, with a prevalence of 18.4% when it was the only diagnosis, rising to 57.8% when comorbidity with a specific PD was allowed. Importantly, though not surprisingly, when patients with only specific PD diagnoses were compared with those who also met criteria for PD-NOS, the latter showed significantly greater problem severity (e.g., using the Outcome Questionnaire-45, Lambert, Lunnen, Umphress, Hansen, & Burlingame, 1996), which again indicates that how PD-NOS is defined affects its import. The authors suggested that many studies underestimate PD-NOS prevalence, and that it would be much higher than is typically reported if all PD-NOS subtypes (i.e., mixed, other, and atypical) were included, and especially if co-occurrence with official PDs were allowed, which would have the added effect of identifying more severe PD. In sum, the high prevalence of PD-NOS—and that it actually may be even more prevalent and severe than as currently diagnosed—clearly indicates the present system’s inadequate coverage of PD.

Personality Disorder-Trait Specified

Taken together, the problems with the current system are strong evidence of the need for a better way to diagnose personality pathology. The trait model in DSM-5-III in general—and PD-TS specifically—have the potential to solve these problems. For example, a notable strength of PD-TS is that it makes full use of the trait-dimensional model, transforming the largely uninformative PD-NOS diagnosis into meaningful descriptions of individuals’ pathological traits. At present, the DSM-5-III definition of PD-TS is similar to that of PD-NOS in previous editions (and, as noted earlier, it also is similar to that of DSM-5-II other specified PD and unspecified PD). That is, PD-TS currently is intended to be used only when the criteria for a specific PD are not met. However, the work of Verheul et al. (2007) and Wilberg et al. (2008) suggests that this may be too limited a definition. Verheul et al. (2007) demonstrated that it was meaningful to diagnose PD-NOS in addition to specific PDs, and both groups of researchers showed that using different PD-NOS definitions affected both the severity and prevalence of the diagnosis. Similarly, in this article, we report on the effects of using various PD-TS definitions.

As far as we can discern, there is as of yet virtually no empirical research on PD-TS. Most of the literature on the DSM-5-III PD model has investigated either the six defined PDs (hereafter, “PD type[s]”) or simply the traits within Criterion B. Mention of PD-TS has been limited largely to definitional descriptions (e.g., Skodol et al., 2011) or opinions and commentary about the diagnosis. For example, Livesley (2012) thought that PD-TS would be the most prevalent diagnosis and wondered whether the type diagnoses would add any information beyond that provided by the trait dimensions. Schmeck, Schlüter-Müller, Foelsch, and Doering (2013) provided two case-study examples of individuals who met criteria for DSM-5-II dependent and narcissistic PD (NPD), respectively. They diagnosed both individuals using the DSM-5-III model, and stated that the new system provided more information and was more helpful for treatment planning than the single diagnoses. In the first case, they diagnosed PD-TS, and in the second case, they listed 14 pathological traits in addition to the two traits comprising the diagnosis and stated that NPD “alone would not really characterize his broad personality pathology” (p. 9).

In contrast, Wakefield (2013) expressed concern that “the PD-trait-specified category opens the door to massive false positives” (p. 182) because, among other reasons, meeting Criterion B requires only one elevated trait. However, we note that a PD diagnosis requires meeting not just Criterion B, but also the general PD criteria, most particularly the personality-functioning impairment required by Criterion A. It may be that many, perhaps most, individuals with a single pathological trait would not also show the required impairment in personality functioning, which would help to limit false positives. In sum, to date, PD-TS has received little more than passing commentary, and one of our goals with this article is to rectify this situation.

Challenges in Defining PD-TS

The DSM-5-III text regarding diagnosing PD states,

“Individuals who have a pattern of impairment in personality functioning and maladaptive traits that matches one of the six defined personality disorders should be diagnosed with that personality disorder. If an individual also has one or even several prominent traits that may have clinical relevance in addition to those required for the diagnosis…, the option exists for these to be noted as specifiers. Individuals whose personality functioning or trait pattern is substantially different from that of any of the six specific personality disorders should be diagnosed with PD-TS. The individual may not meet the required number of A or B criteria and, thus, have a subthreshold presentation of a personality disorder. The individual may have a mix of features of personality disorder types or some features that are less characteristic of a type and more accurately considered a mixed or atypical presentation.” (p. 771)

On first read, it appears that PD-TS is not meant to be assigned in addition to the six types, regardless of the number of elevated “extra-diagnostic” traits (i.e., pathological traits that are in addition to those contributing to a PD-type diagnosis), because such traits can be specified under the types. That is, in the “Specifiers” text following each type’s criteria, the DSM-5-III text states, “Trait and personality functioning specifiers may be used to record additional personality features that may be present in X personality disorder but are not required for the diagnosis. For example, traits of Y are not diagnostic criteria for X personality disorder (see Criterion B) but can be specified when appropriate” (e.g., p. 765, emphasis added). Elsewhere, as cited in the block text above, the text implies that one should specify extra traits that “may have clinical relevance” (p. 771). Whether or not extra traits have “clinical relevance” or whether it is “appropriate” to specify extra traits is left to clinical judgment.

On the other hand, one might argue that, for example, (a) meeting criteria for four types or (b) meeting criteria for one or two types and having a large number (e.g., eight) of additional pathological trait features, constitutes “a mix of features of personality disorder types” that “is substantially different from that of any of the six specific personality disorders” and would be “more accurately considered a mixed or atypical presentation.” Thus, we maintain that whether complex PD cases are better characterized by multiple type diagnoses with or without additional trait specifiers, or simply by PD-TS, remains an empirical research question, which we address in this study.

Different definitions of PD-TS would affect not only the prevalence and comorbidity of other PD diagnoses, but also how one might conceptualize clinical cases. For example, suppose that a patient meets criteria for NPD, for which Criterion B requires two traits, grandiosity and attention seeking, and also has a number of other pathological traits that do not combine to form a PD type, for example manipulativeness, deceitfulness, suspiciousness, rigid perfectionism, and perseveration⁴. Should one (a) diagnose NPD and specify all five extra traits under this diagnosis; (b) diagnose NPD and specify a subset of the extra traits deemed appropriate or clinically relevant (e.g., manipulativeness, deceitfulness, and rigid perfectionism); (c) diagnose NPD and also PD-TS, specifying the five extra traits, to highlight the considerable personality pathology outside NPD; or simply (d) diagnose PD-TS, listing all seven pathological traits and making a case formulation based on the entire set of pathological traits? A narrow reading of the DSM-5-III text provides only for (a) or (b), with the clinical conceptualization based primarily on NPD, but a broader reading could include (c) or (d), as the clinical picture is substantially different from prototypic NPD and is best conceptualized as “a mixed or atypical presentation.” Should this simply be left to clinical judgment (e.g., dependent on whether the clinician deems the narcissistic or extra traits as more central to the patient’s pathology) or can empirical data be brought to bear on the issue, for example, the severity of the narcissistic versus extra traits or, broadening beyond this particular example, the sheer number of extra traits? To address these issues, we examine three potential definitions of PD-TS, described subsequently, with regard to their prevalence, coverage, and severity (e.g., levels of personality functioning and traits).

Method

Participants

Participants were 214 “high-risk” community adults and 165 psychiatric outpatients. Eligibility requirements included age 18 and older, no diagnosis of dementia or intellectual disability, no active psychosis, and able to respond to interviews and questionnaires in English.

Recruitment

The “high-risk” community adults were contacted initially via random dialing of telephone numbers, both landlines and cell phones, in the greater South Bend, Indiana area. The study was described briefly and those contacted were asked if they were potentially interested in participating in the study. If so, they were then screened to ensure that they met the additional eligibility requirements of (a) not currently in mental-health treatment (to distinguish them from the patient subsample), and (b) at or above the recommended cut-point (2 or more positive responses) on the Iowa Personality Disorder Screen (IPDS), which was shown to have good sensitivity and specificity in identifying PD (Langbehn et al., 1999).

Patients were referred primarily from a community mental health center and local practitioners; a small minority were recruited using listservs, newsletters, and mass emails sent to University of Notre Dame staff, faculty, and graduate students, as well as by word of mouth, as long as we were able to verify current mental-health patient status. Hereafter, we refer to the subsamples as the “high-risk” and “patient” subsamples.

Demographics

Mean age was 48.0 years (SD = 12.5; range = 18–84), with the high-risk subsample being slightly older than the patient subsample (Ms = 49.3 ± 13.5 vs. 46.3 ± 10.8, p < .02). The percentage of women (59.1%) did not differ across the subsamples. The patient sample had significantly more racial/ ethnic minorities (37.0%) than the high-risk sample (21.6%; p < .001; 28.3% overall), with Black/ African-American the largest minority group (15.5% of the high-risk, 26.1% of the patient subsamples; 20.1% overall, p < .01). The modal participant had some post-high-school education (47.0%), with no subsample difference in educational level. The modal high-risk participant was employed (43.0% vs. 21.8% of the patients; 33.8% overall), whereas the modal patient was on disability (44.2% vs. 15.0% of the high-risk sample; 27.7% overall). Well over half (76.8%) of the patient, but less than a third (29.1%) of the high-risk subsample, had a family income less than $20,000. Conversely, a third (32.9%) of the high-risk, but only 7.9% of the patient subsample, had a family income of $50,000 or more. The modal high-risk participant was married or living with a partner (50.9% vs. 24.9% of patients; 39.6% overall); the modal patient was single-never married (38.8% vs. 19.2% of high-risk participants; 27.7% overall). Similar percentages were divorced, separated, or widowed (30.0% and 36.4% of the high-risk and patient subsamples, respectively; 32.7% overall). Subsamples differences in employment status, income level, and relationship status were all significant (ps < .0001).

Procedure

All participants came to the research facility of the Center for Advanced Measurement of Personality and Psychopathology (CAMPP) and gave written informed consent before beginning the study. The vast majority completed the questionnaires alone on a computer; the remaining few were assisted by a research assistant because of participant unfamiliarity with computers, poor eyesight, or other reasons. Interview-based measures were also part of the larger project from which the data reported in this article derive, but are not reported on here. Participation required approximately 6 hours and was completed in one or two sessions, either in 1 day or over 2 days, with ample breaks and food/ beverages to reduce fatigue. All procedures were approved by the Institutional Review Board of the University of Notre Dame.

Measures

The self-report battery used in the larger project comprised 14 instruments, five of which we report on in this article: two personality trait measures and three personality functioning measures. Descriptive statistics (means, standard deviations, coefficient alphas) by subsample are provided in Tables 1 through 3, along with the results of t-tests (p < .01, given the large number of scales) and Cohen’s d, to evaluate mean-level differences between subsamples.

Personality traits

Schedule for Nonadaptive and Adaptive Personality, 2^nd Edition

(SNAP-2; Clark, Simms, Wu, & Casillas, in press). The SNAP-2 is a factor analytically derived T-F format measure that assesses 15 personality traits across the normal–abnormal range. Its scales form three higher order dimensions: Negative Affectivity, Positive Affectivity (vs. Detachment), and Disinhibition (vs. Constraint). The SNAP-2 also yields dimensional and categorical scores for the 12 DSM-IV-TR (and thus, the 10 DSM-5-II) PDs, but these are not reported on in this article. The trait scales have shown excellent cross-sectional and longitudinal predictive power for personality pathology and impaired functioning (e.g., Morey et al., 2012). Supplemental Table 1 provides brief descriptions of high and low scorers on the 15 SNAP-2 personality trait scales.

Personality Inventory for DSM-5 (PID-5; Krueger, Derringer, Markon, Watson, & Skodol, 2012) consists of 220 items rated using a 4-point scale ranging from 0 (very false or often false) to 3 (very true or often true). It assesses 25 trait facets organized in five domains, and was developed in conjunction with the DSM-5-III alternative dimensional model for PD. Initial data support the facet and domain scales’ internal consistency; the measure’s intended five-factor structure, for the most part (e.g., Defruyt et al., 2013); and reasonably good correspondence between the PID-5 domains and the Psychopathology-5 domains (PSY-5; Anderson et al., 2013), as well as with at least four of the five-factor model domains, with mixed results for Psychoticism–Openness correspondence (e.g., De Fruyt et al., 2013; Quilty, Ayearst, Chmielewski, Pollock, & Bagby, 2013; Watson, Stasik, Ro, & Clark, 2013). There also is reasonably good correspondence between respective PD types in DSM-5 Sections II and III (e.g., Morey & Skodol, 2013; Samuel, Hopwood, Krueger, Thomas, & Ruggero, 2013).

Personality functioning

We used three measures that were developed relatively recently to assess personality functioning. We used two scales from each measure, one each to assess self and interpersonal pathology, respectively.

General Assessment of Personality Disorder (GAPD; Livesley, 2010)

The 83-item GAPD contains 15 subscales to assess Self Pathology and four to assess Interpersonal Pathology. Respondents rate items based on how they “usually are, think, feel, believe, or act” using a 5-point Likert-type scale (very unlike me—very like me). However, scale-level factor analysis indicated a maximum of two factors, and subsequent analyses indicated that almost all the variance of these factors was captured by a 15-item Self-pathology (e.g., wonder who real me is, powerless to influence what happens to me) scale and an 11-item Interpersonal-pathology (e.g., no close relationships, don’t work to cooperate with people) scale, which we used in all analyses.

Measure of Disordered Personality Functioning (MDPF; Parker et al., 2004)

We administered the 30-item version of this measure described in Ro and Clark (2013), and used 22 items comprising two scales that represent the core of Parker’s Non-coping (e.g., fail more often than succeed, cope poorly; i.e., self pathology) and Non-cooperativeness (e.g., can be difficult in dealing with others vs. nice, good-hearted, caring; i.e., interpersonal pathology) dimensions. These scales correlate with each other similarly to Parker’s original scales (in the .36-.40 range), but are less broad and thus have greater internal consistency. The MDPF uses a 4-point Likert-type format (definitely false—definitely true) and a general time frame.

Severity Indices of Personality Problems-Short Form (SIPP; Verheul et al., 2008)

Ro and Clark (2013) reduced the 60-item SIPP-Short Form to 44 items comprising four scales via replicated factor analysis. We used the two scales—(Self-) Identity (e.g., confused about kind of person I am) and (Interpersonal) Relationships (e.g., hard to show affection)—that loaded most strongly on their Well-Being versus Self Pathology and their Poor Social/ Interpersonal Functioning factors, respectively. The measure uses a 4-point Likert scale (fully agree—fully disagree) with a past-3-months time frame.

Derivation of DSM-5-III Personality Disorder Diagnoses

Impairment in personality functioning (Criterion A)

The first general PD criterion of the DSM-5-III model PD is “Moderate or greater impairment in personality (self/ interpersonal) functioning” (p. 761), defined as a rating of 2 or 3 on a scale ranging from 0 to 3. We operationalized this using our three personality-functioning measures, two of which had been used by Morey et al. (2011) in developing the Level of Personality Functioning Scale (LPFS) that guides clinical ratings of DSM-5-III Criterion A. We first standardized and averaged the self-pathology and the interpersonal-pathology scales, respectively, of the measures. Scale scores were created by averaging across multiple items, so their values were not simply the integers 0 to 3 as in the LPFS). Therefore, we determined the Z score (~0.75) that corresponded to a cut point of 1.5 (i.e., values that rounded to 2 or more) for each composite. The DSM-5-III model permits Criterion A to be met by manifestation of either self or interpersonal pathology, so we considered Criterion A to be met if either composite exceeded a 0.75 Z-score.

Although a Z-score of .75 would be low were it based on a normative sample, this Z-score is based on our combined patient and high-risk sample, which would be expected to have a significantly higher mean score than a normal-range sample. Nonetheless, to ensure that this cut-point would not produce excessive false positives, we compared it to data we had previously collected on these same scales in a volunteer community sample (see Clark & Ro, 2014, for details). When norms based on our previously collected data were applied to the current sample, the equivalent of a .75 Z-score was a Z-score of 3.11. Our previous community-adult sample, being volunteers for psychological research, may have been better adjusted than a random community sample, but even correcting for this possibility, it seems reasonable to assume that using the selected cut-point in these data did not yield a large percentage of false positives, especially given that it corresponded to the cut-point recommended in DSM-5-III.

Pathological traits (Criterion B)

The second general PD criterion in DSM-5-III is exhibiting “one or more pathological personality traits” (p. 761) for which we used the PID-5 (Krueger et al., 2012), the Criterion B measure provided in DSM-5. However, the DSM-5-III does not provide explicit guidelines for the level at which a trait should be considered pathological. Therefore, we used the same cut point to define pathological personality traits as we did to define personality-functioning impairment: Greater than 1.5 on any of the 25 PD facet traits, assessed with the PID-5, which also uses a 0-3 scale (very false or often false—very true or often true). For the PID-5 trait scales, because they were not used to form composites as the personality functioning scales had been, we did not need to rely on Z-scores. Rather, we simply used the scales’ raw scores.

Operational Definitions of PD-TS

We initially formulated four operational definitions of PD-TS, which we later collapsed into three due to very high overlap between two of them. Below we provide each PD-TS definition/ operationalization.

Simple PD-TS

Our first definition followed DSM-5-III: Meets the general PD criteria but does not meet the criteria for any PD type. Operationally this meant being at or above threshold on both our measure of personality-functioning impairment (Criterion A) and at least one facet of the PID-5 (Krueger et al., 2012); empirically, no one in our sample met Criterion A and had only one pathological trait.

Mixed PD-TS

According to the DSM-5-III PD model, when a person has pathological traits that are not part of a PD-type diagnosis, these should be listed as additional specifiers. However, we reasoned that if there were a large number of such “extra” pathological traits, it might be both more parsimonious and more useful clinically simply to use the PD-TS designation, and describe the person’s trait profile, which would represent a mix of both specific diagnostic and additional pathological traits. Therefore, our second definition, which we termed Mixed PD-TS, was that a person met criteria for one or two PD types and had at least five additional pathological traits not accounted for by their PD-type diagnosis/ diagnoses. We chose that number because the PD types require, on average, 4 (3.7) traits to meet criteria, so we reasoned that if the number of additional pathological traits exceeded that average, the clinical picture was sufficiently complex to warrant an “additional diagnosis” which, in this case, meant substituting Mixed PD-TS for the PD-type diagnosis.

Complex PD-TS

Similarly, in DSM-5-III, regardless of the number of PD types whose criteria are met, a person should be assigned all such PD types. However, we reasoned that if a person met criteria for three or more PD types, they would have at least 8 pathological traits, and likely additional traits, beyond those constituting their PD-type diagnoses, so it would be both more parsimonious and more useful clinically simply to use the PD-TS designation, and describe the person’s trait profile, which would represent their full, complex set of traits. Therefore, our third definition, which we termed Complex PD-TS, was that a person met criteria for three or more PD types, regardless of the number of additional pathological traits.

Our fourth definition was based on the idea that if a person’s “extra” pathological traits (i.e., those not part of a PD-type diagnosis) outnumbered their pathological traits that were part of their PD-type diagnosis/ diagnoses then, again, it might be both more parsimonious and more useful clinically simply to use the PD-TS designation. However, when we operationalized this definition, all but 5 individuals who met it also met our definition of Mixed PD-TS. Therefore, we eliminated this definition from further consideration.

Results

Preliminary Analyses: Descriptive Statistics

Personality pathology scales

Means, standard deviations, and coefficient alphas separately by sample are provided in Table 1 for the PID-5, Table 2 for the SNAP-2, and Table 3 for the personality functioning scales. The results of t-tests and effect sizes for differences between subsamples also are shown. We used a .01 alpha-level cut point because of the large number of comparisons. Patients scored significantly higher on most (19/25) the PID-5 scales, whereas of the SNAP-2 trait scales, they differed significantly only on Self-harm (and its subscales) and Disinhibition (both the overlapping and non-overlapping versions). Three (of six) SNAP-2 validity scale scores, plus the overall Index of Invalidity, were significantly higher in the patient subsample. Patients also scored significantly higher on five of the six scales assessing personality functioning.

Table 1.

Descriptive Statistics for the PID-5 Facet Scales in the Patient and Community Samples

PID-5		Patient			Community
Domain	Facet	M	SD	α	M	SD	α	da
NA	Emotional lability	1.46	0.76	.87	1.22	0.72	.87	.33
NA	Anxiousness	1.62	0.77	.90	1.39	0.71	.88	.31
NA	Separation insecurity	1.11	0.76	.86	0.91	0.67	.82	.30
NA	Submissiveness	1.16	0.71	.76	1.20	0.70	.79	.06
NA/ANT	Hostility	1.09	0.64	.86	1.01	0.58	.85	.14
NA	Perseveration	1.21	0.59	.80	1.06	0.59	.84	.25
NA/DET	Depressivity	1.09	0.75	.94	0.80	0.60	.92	.48
NA/DET	Suspiciousness	1.26	0.68	.81	1.08	0.57	.77	.32
−NA/DET	Restricted affectivity	1.00	0.58	.70	0.96	0.56	.75	.07
DET	Withdrawal	1.30	0.78	.93	1.16	0.69	.92	.20
DET	Intimacy avoidance	0.91	0.82	.87	0.71	0.69	.82	.29
DET	Anhedonia	1.26	0.72	.88	1.10	0.65	.87	.25
ANT	Manipulativeness	0.79	0.68	.81	0.71	0.60	.79	.13
ANT	Deceitfulness	0.61	0.60	.89	0.50	0.50	.88	.22
ANT	Grandiosity	0.77	0.64	.80	0.62	0.53	.76	.28
ANT	Attention seeking	0.96	0.76	.90	0.77	0.70	.91	.27
ANT	Callousness	0.46	0.51	.89	0.33	0.36	.84	.36
DIS	Irresponsibility	0.70	0.53	.73	0.51	0.44	.68	.43
DIS	Impulsivity	1.15	0.68	.83	0.91	0.66	.84	.36
DIS	Distractibility	1.38	0.74	.90	1.21	0.69	.89	.25
DIS	Risk taking	1.13	0.61	.90	1.07	0.53	.87	.11
−DIS	Rigid perfectionism	1.32	0.71	.90	1.23	0.61	.87	.16
PSY	Unusual beliefs & exper	0.79	0.68	.83	0.61	0.60	.83	.30
PSY	Eccentricity	1.18	0.79	.95	1.00	0.75	.95	.24
PSY	Cognitive & percept dysreg	0.79	0.60	.87	0.55	0.48	.84	.50

Note. N = 165 (Patients), 214 (High-risk community). PID-5 = Personality Inventory for DSM-5. α = coefficient alpha. d = Cohen’s d. NA = Negative Affectivity; -NA = low NA; DET = Detachment; ANT = Antagonism; DIS = Disinhibition; −DIS = low DIS; PSY = Psychoticism; exper = experiences; percept dysreg = perceptual dysregulation

^aDifferences are significant (p < .01; d < .19) for all but six scales, italicized.

Table 2.

Descriptive Statistics (T-scores) for the SNAP-2 Scales in Patient and High-risk Subsamples

	Patients			High-risk
SNAP-2 Scale	M	SD	α	M	SD	α	d
	Validity Scales
VRIN	51.9	12.0	---	51.4	9.4	---	.05
TRIN	57.3	13.3	---	53.6	11.7	---	.32
DRIN	48.6	10.5	---	49.1	8.9	---	.06
Rare Virtues	56.1	9.9	---	54.6	8.9	---	.17
Deviance	57.5	13.3	---	53.6	10.5	---	.37
Back Deviance	63.4	16.7	---	56.0	11.1	---	.67
Index of Invalidity	56.2	10.0	---	53.4	9.7	---	.29
	Trait Scales
Negative Temperament	59.2	10.1	.92	58.5	9.9	.91	.07
Mistrust	62.4	13.6	.90	60.7	11.1	.83	.15
Aggression	55.5	14.2	.89	54.2	12.0	.85	.11
Manipulativeness	52.0	12.5	.84	49.9	9.4	.73	.22
Self-harm	66.8	16.6	.87	58.3	12.4	.80	.69
Eccentric Perceptions	54.1	12.3	.85	53.8	10.6	.79	.03
Dependency	55.0	12.2	.81	53.3	11.1	.79	.15
Positive Temperament	44.3	12.3	.90	45.1	12.2	.90	.07
Exhibitionism	47.7	10.4	.82	45.6	9.5	.80	.22
Entitlement	49.1	12.4	.86	47.7	10.8	.80	.13
Detachment	55.6	11.5	.87	55.8	10.9	.85	.02
Disinhibition	52.5	11.0	.86	50.1	9.4	.82	.26
Impulsivity	51.8	9.4	.75	50.0	9.8	.80	.17
Propriety	54.6	8.7	.81	55.5	7.1	.71	.13
Workaholism	54.3	10.6	.80	54.6	10.6	.78	.03
	Subscales
Disinhibition-Pure	58.8	13.6	.77	54.7	11.1	.68	.37
Low Self-Esteem	66.6	16.2	.84	61.7	14.1	.77	.35
Suicide Proneness	63.3	17.0	.82	53.3	11.6	.73	.86

Note. N = 165 (Patients), 214 (High-risk). M = mean; SD = standard deviation. α = coefficient alpha. d = Cohen’s d (effect size). VRIN = Variable Response INconsistency. TRIN = True Response INconsistency. DRIN = Desirable Response INconsistency. Larger value is bolded if the effect size is ≥ .50. Effect sizes associated with non-significant t-test results (p > .01; d < .19) are italicized.

Table 3.

Descriptive Statistics for the Personality Functioning Scales in Patient and High-risk Groups

		Patients			High-risk
Measure	Scale	M	SD	α	M	SD	α	d
GAPD	Self-pathology	1.45	.86	.92	1.17	.71	.90	.39
GAPD	Interpersonal Pathology	1.07	.71	.88	0.87	.64	.89	.31
MDPF	Non-coping	1.36	.71	.90	1.16	.63	.89	.32
MDPF	Non-cooperativeness	0.72	.50	.88	0.60	.40	.86	.30
SIPP	Identity	1.10	.74	.91	0.88	.63	.88	.35
SIPP	Relations	1.22	.71	.86	1.14	.66	.84	.12

Note. N = 165 (Patients, except GAPD, n = 163), 214 (High-risk). M = mean; SD = standard deviation. α = coefficient alpha. d = Cohen’s d (effect size). GAPD = General Assessment of Personality Dysfunction. MDPF = Measure of Disordered Personality Functioning. SIPP = Severity Indices of Personality Pathology. Effect sizes associated with non-significant t-test results (p > .01) are italicized.

Across all scales, however, effect sizes were small to moderate, mostly in the .20-.40 range, with only six (of 56) scales having an effect size above .40. Differences of medium effect sizes (i.e., > .50) were even fewer (three scales)—one PID-5 scale (Cognitive and Perceptual Dysregulation), one SNAP-2 trait scale (Self-harm, and its subscale, Suicide Potential) and one SNAP-2 validity scale (Back Deviance, similar to MMPI Back Deviance)—and there were no large effect sizes. Therefore, we analyzed the complete sample, referencing the two subgroups when relevant or of interest.

PD-type diagnostic prevalence

Using the operational definition described earlier, about one-third (30.0%) of the sample (n = 114) was above threshold on Criterion A, with a significantly greater proportion in the patient (40.6%) than the high-risk (22.0%) subsample (p < .0001). Table 4 presents diagnostic prevalences for the six DSM-5-III PD types, and for any PD type, assessed using the PID-5. The first column shows the prevalences based on meeting both Criterion A and Criterion B (i.e., each PD type’s specific pattern of pathological traits), whereas the second shows the prevalence without considering Criterion A (i.e., using only Criterion B, pathological traits). The last column shows the prevalence of each PD type for those with one and only one PD-type diagnosis. We address this last set of results in a later section.

Table 4.

Prevalence of Specific DSM-5-III Personality Disorder Diagnoses With and Without Criterion A and Pure Cases^a

	Criterion A
Specific PD diagnosis	% With (n = 81)	% Without (n = 127)	Pure
Antisocial PD	1.3	1.6	0.0
Avoidant PD	13.5	16.6	4.0
Borderline PD	13.7	19.8	3.2
Narcissistic PD	2.1	5.3	0.0
Obsessive-Compulsive PD	6.3	8.4	1.3
Schizotypal PD	6.6	7.4	1.1
Any specific PD (of 6)	21.4	33.5	9.5

Note. N = 379. Pure = Meets criteria only for one specific PD.

^aBased on self-report questionnaires; see text for details.

Consistent with the proportion of the sample meeting Criterion A, overall prevalence was over one-third lower when Criterion A was required (21.4%) compared to when diagnosed using only Criterion B (33.5%). Thus, it does appear that Criterion A serves to exclude from a PD diagnosis individuals with extreme traits who nonetheless are not also reporting personality-functioning impairment. Avoidant and borderline PD were the most prevalent PD types at 13.5% and 13.7%, respectively, whereas antisocial and narcissistic PD were the least prevalent at 1.3% and 2.1%, respectively. Obsessive-compulsive and schizotypal PD were in between at 6.3% and 6.6% prevalence, respectively.

Parallel data on the high-risk and patient subsamples separately are provided in Supplemental Table 2, and we summarize them here: schizotypal, borderline, and avoidant PD were significantly more common among patients, regardless of whether Criterion A was required. Moreover, significantly more patients met criteria for any PD-type diagnosis: 12.6% versus 32.7% when Criterion A was required, and 22.0% versus 48.5% when only Criterion B was considered. Thus, taken together, the data indicate that the patient sample exhibited somewhat more personality pathology than the high-risk sample which, in turn, appeared to exhibit somewhat more personality pathology than would an unscreened community-adult sample, in which PD prevalence, even considering all 10 DSM-5-II PDs, is typically found to be approximately 10–13% (Sansone & Sansone, 2011).

Primary Analyses

All individuals meeting Criterion A (n = 114) had at least one pathological trait (i.e., the trait score was higher than the cut point); in fact, the minimum was two, ranging up to 24. Thus, all 114 were “eligible” for a PD diagnosis, either a PD type or PD-TS. The top portion of Table 5 shows, separately for those with and without PD, the mean and standard deviation of the number of pathological traits, the severity (elevation) of those traits, and the severity of personality-functioning impairment (i.e., Criterion A). The average number of pathological traits in those with PD was 9.9, significantly more than the 3.8 in those without a PD diagnosis. Trait severity and level of personality-functioning impairment also were significantly higher in the PD than the no-PD group (with the latter being the primary basis for making a PD diagnosis). Interestingly, the number of pathological traits in the no-PD group ranged from 0 to 14, clearly indicating that some individuals have pathological-level traits without also having the personality dysfunction required by Criterion A. Nonetheless, 75% of those not meeting Criterion A had 5 or fewer pathological traits, and only 6.4% had 10 or more, so it was rare to have a considerable number of pathological traits absent personality-functioning impairment.

Table 5.

Prevalence and Severity of Personality Pathology in those with and without PD, and in Four PD Subgroups, including Three Definitions of Personality Disorder-Trait Specified

			Pathological Traits					Personality Impairment Severity1
PD Group	n	Freq. of PD	Number			Severity
			M	SD	range	M	SD	M	SD
No PD	265	69.9	3.8	3.1	0–14	1.89	.21	−0.40	.53
Any PD	114	30.1	9.9	4.3	3–24	2.02	.22	0.93	.41
			Participants With PD
Simple PD-TS	33	29.0	5.9a	2.2	2–10	1.90a	.21	0.71a	.33
Mixed PD-TS	31	27.2	10.9b	2.2	8–15	2.07b	.23	0.86ab	.37
Complex PD-TS	26	22.8	15.2c	3.7	11–24	2.12b	.21	1.22c	.41
Basic PD	24	21.1	8.7d	2.7	3–14	2.01ab	.13	0.98b	.39

Note. PD = Personality Disorder. TS = Trait Specified. Basic PD = met criteria for one or two specific PDs and had fewer than five additional pathological traits.

¹Z scores.

For No PD versus Any PD comparisons, bolded values are significantly higher, p < .0001.

For PD-group comparisons, means within a column with different superscripts are significantly different, Bonferroni-corrected p < .05.

Personality Disorder-Trait Specified

Coverage and prevalence

We then applied our three PD-TS definitions to all individuals meeting Criterion A. The lower portion of Table 5 shows the prevalence of each PD-TS type in this PD-eligible group: 29% met the definition of Simple PD-TS (i.e., no PD type and one or more pathological traits), 27.2% met the definition of Mixed PD-TS (i.e., one or two PD-type diagnoses and at least five additional pathological traits), and 22.8% met the definition of Complex PD-TS (i.e., three or more PD types). The remaining 21.1%, which we term “Basic PD,” all met the criteria for one or two PD types and had four or fewer additional pathological traits. Importantly, these four groups were non-overlapping and fully accounted for all 114 PD-eligible individuals in the sample.

Thus, 29% of those with PD (8.7% of the whole sample) would meet criteria for PD-TS using a narrow interpretation of DSM-5-III (i.e., meet the general criteria for PD and have at least one pathological trait, but do not meet the criteria for any PD type), with the remaining 71% meeting criteria for at least one PD type. In contrast, using our additional definitions of PD-TS (i.e., Mixed and Complex PD-TS), the converse emerges: Nearly 80% of the sample meets one of our three PD-TS definitions, whereas barely one-fifth of those with any PD are well characterized using the set of six DSM-5-III PD types, in the sense that their trait profiles match those of only one or two PD types with few additional traits to complicate the clinical picture. The rest—the large majority—are better characterized by their complete trait profile, either because they do not meet criteria for any PD type (29%) or because their trait profile is mixed (27.2%) or complex (22.8%). Of course, those with basic PD also can be characterized by their trait profiles, so PD-TS could be used to characterize all individuals with PD in our sample.

Severity levels of Basic PD and PD-TS

PID-5 trait and functioning impairment levels

Table 5 also shows the mean and standard deviation of the number of pathological traits, the severity (elevation) of those traits, and the severity of personality-functioning impairment (i.e., Criterion A) in each of the defined subgroups. As can be seen in the table, the average number of pathological traits in the Simple PD-TS group was almost 6 (range = 2 to 10), indicating a general level of pathological traits sufficient to warrant a PD diagnosis. Nonetheless, individuals in this group appeared to be somewhat less pathological than those with a Basic-PD diagnosis (i.e., meeting criteria for one or two PD types with four or fewer additional pathological traits). Specifically, those in the Basic PD group were characterized by significantly more pathological traits and more severe personality impairment than those with Simple PD-TS; however, the two groups did not differ in their pathological-trait severity/ elevation.

Individuals in the Mixed PD-TS group, in turn, showed slightly more personality pathology than those with Basic PD, in that those in the Mixed PD-TS group had a significantly greater number of pathological traits. However, there was no difference in the severity level of their traits or in the level of personality-functioning impairment. In contrast, individuals in the Complex PD-TS group were notably more pathological in two respects. Specifically, they had significantly more pathological traits and more severe personality-functioning impairment than any of the other three groups. Interestingly, however, the severity of pathological traits differed only from that of the Simple PD-TS group.

SNAP-2 trait levels

These results are all confounded, however, because they are based on the same indices that were used to categorize the individuals. Thus, examining pathological traits and functional impairment in independent measures is important. Figure 1 shows the SNAP-2 personality profiles for individuals with no versus any PD, and Figure 2 shows the SNAP-2 profiles for our four PD subgroups (i.e., Basic PD and the three PD-TS groups). Table 6 provides the corresponding means and standard deviations. As can be seen in Figure 1 and as documented in Table 6, there was a clear overall level/ severity difference between the no-PD and any-PD groups, with significant differences (p < .01) on all scales except Propriety and Workaholism. Effect sizes were large (M = .89, range = .56 to 1.41), except for Exhibitionism (ES = .31) and Entitlement (ES = .26). Moreover, in the any-PD group, six of the scales’ mean scores (Negative Temperament, Mistrust, Anger, Self-harm, [low] Positive Temperament, and Detachment) were in the pathological range (≥ |1 SD|) and two others (Eccentric Perceptions and Dependency) were close to this range. In contrast, mean scores in the no-PD group were all within normal range. Thus, the DSM-5-III model clearly differentiates between individuals with and without personality pathology, even when using a self-report questionnaire that is independent of the diagnostic criteria themselves. Finally, the mean scores for three traits—Negative Temperament, Mistrust, and Self-harm—were particularly elevated, suggesting that these traits may characterize PD personality-trait pathology in general.

Figure 1. SNAP-2 Trait Profile for Individuals with No and Any DSM-5-III PD.

Note. SNAP-2 = Schedule for Nonadaptive and Adaptive Personality-2^nd Edition. PD = Personality Disorder. TS = Trait Specified. NT = Negative Temperament. MIS = Mistrust. MANIP = Manipulativeness. AGG = Aggression. SLFH = Self-harm. ECCP = Eccentric perceptions. DEP = Dependency. PT = Positive Temperament. EXH = Exhibitionism. ENT = Entitlement. DET = Detachment. DIS = Disinhibition. IMP = Impulsivity. PRO = Propriety. WRK = Workaholism. LOSE = Low self-esteem. SUICP = Suicide proneness. Cut points for pathological extremes on the SNAP-2: < 40 and > 60.

Figure 2. SNAP-2 Trait Profiles for Basic PD and Three PD-TS Groups.

Note. SNAP-2 = Schedule for Nonadaptive and Adaptive Personality-2^nd Edition. PD = Personality Disorder. TS = Trait Specified. NT = Negative Temperament. MIS = Mistrust. MANIP = Manipulativeness. AGG = Aggression. SLFH = Self-harm. ECCP = Eccentric perceptions. DEP = Dependency. PT = Positive Temperament. EXH = Exhibitionism. ENT = Entitlement. DET = Detachment. DIS = Disinhibition. IMP = Impulsivity. PRO = Propriety. WRK = Workaholism. LOSE = Low self-esteem. SUICP = Suicide proneness. Cut points for pathological extremes on the SNAP-2: < 40 and > 60.

Table 6.

SNAP-2 T-Scores in Six Groups: No PD, Any PD, Basic PD, and Three PD-TS Subtypes

							PD-Trait Specified Subtypes
	No PD n = 265		Any PD n = 114		Basic PD n = 24		Simple n = 33		Mixed n = 31		Complex n = 26
	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD
Negative Temperament	56.1l	9.7	65.0	7.7	65.6	8.3	62.4	8.0	65.3	7.5	67.4	6.1
Mistrust	57.8l	11.4	70.0	9.5	70.1ab	8.7	66.1a	10.9	69.9ab	9.2	75.0b	6.2
Manipulativeness	48.8l	9.5	55.3	12.5	50.9a	12.6	55.5ab	10.9	52.8a	10.7	61.9b	14.2
Aggression	52.3l	10.8	74.0	13.9	61.8ab	14.2	57.9a	14.6	56.0a	14.4	68.7b	16.7
Self-harm	56.8l	12.2	58.0	12.5	75.1ab	12.6	68.4a	12.5	73.2ab	13.3	81.2b	14.8
Eccentric Perceptions	52.1l	10.4	58.7	12.8	55.2ab	12.3	52.6a	11.1	60.9a	11.4	63.8b	12.5
Dependency	52.0l	10.5	60.6	15.5	59.5	14.5	59.8	14.0	57.9	11.6	57.3	11.3
Positive Temperament	47.6l	11.1	38.3	12.3	35.2	11.9	38.9	13.7	38.1	11.5	40.6	11.8
Exhibitionism	47.4m	9.6	44.3	10.3	39.8	7.8	46.6	10.3	43.3	9.9	46.5	11.7
Entitlement	49.3s	11.2	46.1	12.0	40.7a	11.6	44.2ab	9.7	49.2b	11.9	49.7b	13.2
Detachment	52.6l	10.2	63.1	9.8	66.4a	8.6	59.5a	9.1	61.5ab	9.8	66.7b	9.9
Disinhibition	49.0l	8.9	55.9	11.3	54.0	12.1	54.8	9.6	54.7	8.8	60.5	14.4
Impulsivity	48.9l	8.4	55.2	10.8	56.2	14.2	54.4	9.8	53.8	8.7	56.9	11.1
Propriety	55.2ns	7.6	55.1	8.4	54.8	8.2	54.3	8.6	54.6	8.9	56.8	7.8
Workaholism	54.4ns	10.7	54.6	10.3	53.0	10.9	53.1	10.0	54.1	8.9	58.7	11.3

Notes. PD = Personality Disorder. TS = Trait Specified. Temperament-trait scales are shown in boldface. Superscripts after the No-PD means indicate effect sizes in comparison with Any PD. Effect size abbreviations: l = large, m = medium, s = small, ns = not significant. For the four more specific PD groups, means with different superscripts are significantly different, Bonferroni-corrected p < .05. Variances are significantly different between the PD and no-PD groups: *p< .01, †p < .03.

To compare the SNAP-2 profiles of the four PD groups (i.e., Basic PD and the three PD-TS groups shown in Figure 2) quantitatively, we ran a generalized linear model ANOVA with Bonferroni-corrected follow-up contrasts, the results of which are also shown in Table 6. Interestingly, differences in trait level/ severity across the four PD groups were relatively small and the overall pattern or shape of the profiles was generally quite similar. Five of the 15 scales (Mistrust, Manipulativeness, Self-harm, Eccentric Perceptions, and Detachment) yielded a significant ANOVA at p < .01, whereas Entitlement showed significant differences at p < .02. Bonferroni-corrected follow-up contrasts yielded complex patterns in which, for example, Simple and Complex PD-TS were different from each other, but neither was significantly different from Mixed PD-TS or Basic PD. Close examination of these patterns revealed two constants: For the six scales with significant between-group differences listed above, (1) the mean SNAP-2 scores of the Complex PD-TS group were always significantly higher than at least one other group’s, and (2) the other groups’ means were not different from each other, except on Entitlement, for which Basic PD was significantly lower than Mixed PD-TS. However, which group or groups differed from Complex PD-TS varied across scales, with no discernable pattern.

These results suggest that differences among the four PD groups (Basic PD and three definitions of PD-TS) are largely quantitative, not qualitative in nature. That is, (1) there is a similar admixture of pathological personality traits in the four groups, and (2) the groups differ primarily in the number and severity of pathological traits, not the particular traits manifested. Moreover, recalling that certain scales—specifically, Negative Temperament, Mistrust, and Self-harm—were elevated in all four PD groups, the fact that two of these—Negative Temperament and Mistrust—also have among the smallest standard deviations across all the subgroups, further suggests that these traits are generally characteristic of this PD sample. It will be important to determine whether this is sample specific or generalizable to other PD samples. We turn now to an examination of the effects that taking this broader view of PD-TS has on various well-known problems with categorical PD systems.

Comorbidity

As can be inferred from the data in Table 4 and consistent with the literature, there is considerable PD comorbidity in the sample. The percentages of the six PD types sum to 43.5%, whereas the overall prevalence of individuals with any PD type is only 21.4%, indicating an average of 2 (2.03) PD-type diagnoses in those with any PD-type diagnosis. The last column in Table 4 shows the prevalence of “pure” cases of PD types (n = 36), that is, those meeting criteria for one and only one PD diagnosis. Although the relative prevalence of the specific diagnoses is maintained (e.g., avoidant and borderline PD remain the most common), the absolute prevalence of all six PD types is markedly reduced, in two cases to zero percent, when comorbid cases are excluded.

Table 7 provides a more detailed examination of the sample’s comorbidity, showing the frequency of individuals meeting criteria for zero through six DSM-5-III PD types in the PD subsample, first excluding Simple PD-TS, which is how comorbidity is usually calculated, and then including Simple PD-TS, a more inclusive standard. Again consistent with the DSM PD literature on comorbidity reviewed earlier, of those meeting criteria for any PD type, over half (55.6%) met criteria for two or more PD types and almost a third met criteria for three or more. Even including those with simple PD-TS, the prevalence of comorbid PD is 39.5%.

Table 7.

Frequency of Levels of PD-Type Comorbidity Overall and With and Without Including PD-TS

		PD Sample With and Without Simple PD-TS
Number of PD-type Diagnoses	n	% Without (n = 81)	% With (n = 114)
None	265	0.0	0.0
Simple PD-TS	33	0.0	29.0
1	36	44.4	31.6
2	19	23.5	16.7
3	18	22.2	15.8
4–6	8	9.9	7.0
Total Comorbid	45	55.6	39.5

In contrast, if we include all three definitions of PD-TS and examine comorbidity among the 24 individuals with Basic PD, the prevalence of comorbidity drops to 9.6% of those with any PD. That is, only 11 individuals in the entire sample met criteria for two PD types and had four or fewer additional traits. Eight more individuals (7.0%) met criteria for only two PD types, but they also have five to eight additional traits that complicate the clinical picture, thus yielding a Mixed PD-TS diagnosis. Finally, as noted earlier, 22.8% met diagnosis for three or more PD types—and all had three to six additional traits as well—so these individuals are arguably better characterized as having Complex PD-TS than three to six PD types plus additional pathological traits. Thus, defining PD-TS broadly reduces specific diagnostic comorbidity markedly by allowing individuals with complex personality pathology to be characterized by their trait profiles rather than by a list of two to six PD diagnoses that only rarely, even collectively, provide a complete clinical picture of the individuals’ trait pathology.

Severity

The effect of comorbidity can be considered also by examining the severity of Criterion A, personality-functioning impairment. Due to the small numbers, we combined those with four-to-six comorbid PDs and again ran a generalized-linear-model ANOVA with Bonferroni-corrected follow-up contrasts, with the number of PD types as the independent variable. The average personality-functioning impairment (an average of the self- and interpersonal-impairment Z-scores) was lowest in the non-comorbid and single-PD groups (M Z scores = .71 and .82, respectively), which did not differ from each other, and highest in the three comorbid groups, which also did not differ from each other (M Z scores = 1.11, 1.19, and 1.29, in those with two, three, and four-to-six PD-type diagnoses respectively). These results are similar to those based on the different PD-TS definitions in which Simple PD-TS had the lowest level of impairment, Complex PD-TS the highest, with Basic PD and Mixed PD-TS in between.

In sum, as specific comorbidity increases, severity increases in two ways: First, the extent of trait pathology necessarily increases, given that the number of pathological traits increases with number of diagnostic criteria met (indeed, the number of PD diagnoses and of pathological traits correlated .86 in the PD sample and .75 in the full sample). Second, the level of personality-functioning impairment increases with comorbidity and its associated profile complexity, the latter of which is captured more directly using the PD-TS diagnosis.

Within-PD heterogeneity

Standard deviations

We noted earlier that Negative Temperament and Mistrust were both elevated (> 65 T-scores) and had lower standard deviations in the overall PD group (7.7 and 9.5, respectively) as well as in the separate Basic PD and three PD-TS subcategories (range = 6.1 to 10.9), suggesting they are common, core PD traits. Beyond this similarity, however, the concept of PD-TS emphasizes individuals’ profiles, so the trait scales’ standard deviations are of significant interest. Turning again to Table 6, it is noteworthy that the standard deviations of the any-PD group compared to those of the no-PD group are significantly higher (p < .01) for eight of the scales, somewhat higher (p < .03) for three more scales (Mistrust, Dependency, and Entitlement), and not different for only three scales (Detachment, Propriety, and Workaholism).

These results indicate that on most traits there is more within-group heterogeneity in the any-PD than in the no-PD group, which is consistent with a dimensionally based approach to personality pathology, in which individuals’ profiles would be expected to vary. The standard deviations of two traits—Aggression (15.5) and Self-harm (13.9)—were quite high (recall that with T-scores 10 points = 1 SD in a normative sample), indicating particularly marked heterogeneity on these traits. Notably, the standard deviations for these scales were high not only in the whole PD sample, but also within each of the PD subgroups, indicating marked heterogeneity warranting consideration of individuals’ profiles in every case.

Interestingly, despite the lower standard deviations of Negative Temperament and Mistrust, heterogeneity (i.e., SD) was generally higher among traits associated with Negative Affectivity (12.1 for the first seven scales in Table 6, 13.4 excluding Negative Temperament and Mistrust) than those associated with either Positive Affectivity (11.1 for the next four scales) or Disinhibition (10.2 for the final four scales). Determination of whether this is sample specific or a general characteristic of PD samples requires replication.

Extra-diagnostic traits

In our discussion of comorbidity, we touched upon the issue of extra-diagnostic traits in those who meet criteria for PD-type diagnoses. Consideration of the extent of extra-diagnostic traits is another way to approach the issue of heterogeneity. In particular, if the number of such extra-diagnostic traits is small, then it is reasonable to specify them along with the diagnosis (e.g., narcissistic PD plus callousness, or avoidant and obsessive-compulsive PD plus depressivity and distractibility), whereas if the number of extra-diagnostic traits is large, it may make more sense to consider individuals as manifesting PD-TS and to specify their trait profiles.

Table 8 shows the frequency of extra-diagnostic pathological traits per number of PD-type diagnoses, for any PD type, and for any PD (i.e., including Simple PD-TS). First, it is notable that over all individuals with PD, the number of pathological traits used in diagnosing the PD types is only slightly more than the number of extra-diagnostic traits (i.e., 5.3 vs. 4.6). Focusing only on those who meet criteria for one or more PD types (i.e., removing those with Simple PD-TS, who have no PD-type diagnosis), this gap, of course, widens (7.5 vs. 4.1), but for individuals with only one PD type, there are actually more extra-diagnostic traits, on average (4.9) than diagnostic traits (4.0). Thus, the number of pathological traits not captured by the six PD-type diagnoses is considerable.

Table 8.

Number of “Extra” (Non-Diagnostic) and Diagnostic Pathological Traits by the Number of PD Types, Any PD Type, and Any PD (Including PD-TS)

Number of PD types		Number of “Extra” Traits			Number of PD-type Diagnostic Traits
	n	M	SD	Range	M	SD	Range
1	36	4.9	2.1	0–8	4.0	1.2	3–7
2	19	4.2	1.8	2–8	7.6	1.3	6–11
3	18	3.7	1.1	1–6	10.1	1.3	8–12
4–6	8	1.6	1.0	0–3	16.6	4.3	11–22
Any PD Type	81	4.1	2.0	0–8	7.5	4.3	3–22
Any PD	114	4.6	2.2	0–10	5.3	5.0	0–22

Note. PD = Personality Disorder; TS - Trait Specified

Second, as noted earlier (and shown in Table 4) only 36 individuals met criteria for a single PD type. If we take a more extreme stance and consider “pure” cases to be those not only with one and only one PD type, but also having few (i.e., ≤ 4) extra traits, this number is reduced to only 13 individuals (11.4% of the PD subsample), and only one person had a single PD with no extra traits. Thus, almost two-thirds of those meeting criteria for only one PD type had five or more additional traits, which is more than the average number of traits required for the PD-type diagnoses (M = 3.7, range = 2 to 6). This raises the question of the value of the complex infrastructure subserving the DSM-5-III PD types, which cleanly characterizes only a small minority of individuals with PD. In contrast, a far simpler system, consisting of only personality-functioning impairment (Criterion A) and the Criterion B trait set, can be used to characterize all PD comprehensively, using the PD-TS diagnosis.

Third, Table 8 also shows that many individuals meeting criteria for two to three PD-type diagnoses also have a considerable number of pathological traits above and beyond those contributing to their diagnoses, with means of 4.2 and 3.7, respectively, ranging up to 8. Only six of 37 individuals with two or three PD types had just one or two extra traits; the remaining 31 had three or more, with a modal number of three for those with two PD types and four for those with three PD types. It would be somewhat cumbersome to specify even three or four additional traits (let alone 8) in addition to two or three diagnoses, again calling into question the clinical utility of the DSM-5-III hybrid model. It arguably would be more clinically useful simply to diagnose PD-TS and provide an overall trait profile than to diagnose a person with, for example, avoidant, borderline, and narcissistic PD with pathological perseveration, rigid perfectionism, distractibility, and eccentricity.

The PID-5 and SNAP-2 trait profiles of the individual in our sample who met criteria for these PD-type diagnoses and pathological traits is shown in Figure 3. The particular combination of elevations on almost all the PID-5 Negative Affectivity-domain traits, echoed in SNAP-2 scale elevations on Negative Temperament, Mistrust, and also PID-5 Withdrawal/ SNAP-2 Detachment, PID-5 Anhedonia/ SNAP-2 low Positive Temperament, PID-5 Hostility/ SNAP-2 Aggression, PID-5 Grandiosity/ SNAP-2 Entitlement, PID-5 Attention seeking/ SNAP-2 Exhibitionism, and PID-5 Distractibility and Eccentricity (which have no direct SNAP-2 counterparts), paint a picture of a distressed patient with a complex mixture of attitudes and behaviors towards others that suggests difficulty in therapeutic engagement. However, the individual also has some low-normal range scores that indicate psychological strengths from which to begin building a working alliance, such as low PID-5 Callousness, Deceitfulness, Irresponsibility and Impulsivity (cf. low-normal SNAP-2 Manipulativeness, Disinhibition, and Impulsivity), and low scores on the other two PID-5 Psychoticism scales (cf. SNAP-2 Eccentric Perceptions), suggesting intact cognitive functioning despite difficulty in concentrating and goal pursuit (i.e., PID-5 Distractibility).

Figure 3. PID-5 and SNAP-2 Trait Profiles for an Individual Meeting Criteria for DSM-5-III Avoidant, Borderline, and Narcissistic PDs.

Note. PID-5 = Personality Disorder Inventory for DSM-5. SNAP-2 = Schedule for Nonadaptive and Adaptive Personality-2^nd Edition. PD = Personality Disorder. Anx = Anxiousness, Emolab = Emotional lability. SepIn = Separation insecurity. Submis = Submissiveness. Persv = Perseveration. RgdPerf = Rigid perfectionism. Suspic = Suspiciousness. Deprsv = Depressivity. Wthdrl = Withdrawal. IntAvd = Intimacy avoidance. RstAff = Restricted affectivity. Hstl = Hostility. Manipl = Manipulativeness. Deceit = Deceitfulness. Grndos = Grandiosity. Attnsk = Attention seeking. Irresp = Irresponsibility. Impulsv = Impulsivity. Distrac = Distractibility. RskTkng = Risk taking. UnusBlEx = Unusual beliefs and experiences. Eccn = Eccentricity. CgPcDys = Cognitive and perceptual dysfunction.

NT = Negative Temperament. MIS = Mistrust. MANIP = Manipulativeness. AGG = Aggression. SLFH = Self-harm. ECCP = Eccentric perceptions. DEP = Dependency. PT = Positive Temperament. EXH = Exhibitionism. ENT = Entitlement. DET = Detachment. DIS = Disinhibition. IMP = Impulsivity. PRO = Propriety. WRK = Workaholism. LOSE = Low self-esteem. SUICP = Suicide potential. Cut point for pathological elevation on the PID-5 = 1.5 Cut point for pathological extremes on the SNAP-2: < 40 and > 60.

Finally, and not surprisingly, the number of extra-diagnostic traits is fewer in individuals meeting criteria for four or more PD types, because so many pathological traits are used in making the multiple diagnoses (three individuals had 0–1 and five had 2–3 extra-diagnostic pathological traits). Nevertheless, these data again raise the question of whether it is more useful to indicate that an individual, for example, meets criteria for avoidant, borderline, obsessive-compulsive, and schizotypal PD (the most common combination of those meeting criteria for four or more PD types in our sample), plus up to three additional traits, or simply to list their pathological traits and provide a trait profile, indicating PD-TS.

Discussion

Our results suggest an important main finding: Although the hybrid model developed for DSM-5-III may be useful as a transitional object for those who do not have previous experience with trait-dimensional models, it does not clearly and fully represent the highly varied empirical reality of PD. Moreover, Wakefield (2013) dubbed the diagnostic criteria for the PD types “theoretically questionable and clinically less useful” (p. 182), whereas Livesley (2012) criticized the hybrid model on several points, including that it had an “inconsistent conceptual structure” (p. 87) and stating, “The use of different definitions of the same trait is a puzzling violation of the basic principles of trait psychology” (p. 86).

In contrast, a broadly defined PD-TS diagnosis efficiently provides comprehensive information about individuals’ personality pathology, thus maximizing the clinical utility of a PD diagnosis. We developed our three definitions of PD-TS based on the DSM-5-III system, that is, using Criterion A, the PD-type diagnoses, their comorbidity, and additional pathological traits. Importantly, these three definitions, together with a Basic-PD group consisting of those with fairly simple and straightforward DSM-5-III PD-type diagnoses, fully account for all “PD-eligible” individuals (i.e., all those above threshold on Criterion A), with no overlap among groups. Moreover, if we lift the restriction of not meeting any PD-type diagnosis, then all four groups can be subsumed under the definition of Simple PD-TS: Above threshold on Criterion A and having one or more pathological traits. These results, by demonstrating that “all roads lead to PD-TS,” lead us to question whether there is added value in the complex DSM-5-III model, other than as a transitional object to aid those who are unfamiliar with a functioning-and-trait dimensional approach to PD diagnosis.

Coverage

About 30% of our sample was “PD-eligible”; that is, met or exceeded the Criterion A threshold for moderate or greater impairment in personality functioning. Of these, only just over 20% had a Basic PD, which represents a relatively clear, straightforward clinical picture using the DSM-5-III PD-types. In other words, the DSM-5-III hybrid PD system was able to characterize only one-sixteenth of our full sample easily and clearly. In contrast to the ~20% covered by Basic PD, almost 30% of our PD sample had a simpler clinical picture, well characterized using only Criterion A and the Criterion B trait set (i.e., Simple PD-TS), whereas exactly 50% showed a more complex clinical picture (i.e., Mixed or Complex PD-TS) that we argue is better represented by using the Criterion B trait set to provide a complete trait profile than by a combination of two or more PD types and up to eight additional traits. In other words, PD-TS seems more useful than the hybrid model for diagnosing 80% of our PD sample. And, of course, even the subset of individuals with Basic PD can be characterized by PD-TS, which would provide 100% coverage and render the PD types unnecessary from this perspective.

Prevalence

Some researchers (e.g., Wakefield, 2013) have suggested that using the DSM-5-III definition of PD-TS would lead to elevated rates of PD diagnoses, claiming that the requirement of only one pathological trait is too lenient. If Criterion A were not required, we are inclined to agree: Only 38 individuals in the entire sample had no pathological traits. If Criterion A were not required for either a PD-type diagnosis or PD-TS, then 33.5% of the sample would meet Criterion B for a PD type and 56.5% would receive a PD-TS diagnosis, clearly warranting Wakefield’s “massive false positives” designation. However, by requiring moderate or greater impairment in personality functioning for diagnosis, PD prevalence was comparable to that reported in the literature: 21.4% for any PD-type (12.6% and 32.7% in the high-risk community and patient subsamples, respectively), and DSM-5-III defined PD-TS fell to 8.7% (9.3% and 7.9% in the high-risk and patient subsamples, respectively).

Thus, even though the Criterion B requirement of only one elevated trait may seem low, the addition of Criterion A appears to address the concern expressed by Wakefield (2013), and likely others, of a high false-positive rate of PD diagnoses in the DSM-5-III model. In this context, it is also worth noting that our data provide some support for the Personality and PD Work Group’s decision not to require impairment in both self and interpersonal functioning. Specifically, when we required that individuals be above threshold on both the self- and interpersonal-pathology scales, PD prevalence was only 10.6% (6.5% and 15.8% in the high-risk and patient subsamples, respectively), which is considerably lower than rates using the DSM-5-II PD diagnoses.

Of course, these findings need replication, but our results suggest that personality-functioning impairment is a useful construct for distinguishing individuals who have pathological-level traits, but not PD per se, from those whose personality-functioning impairment together with their pathological-level traits warrants a PD diagnosis. The former individuals may be described as having Personality Difficulty, a designation being considered for the International Classification of Diseases, 11th Edition, to indicate subthreshold manifestations of personality disorder. Future research will need to examine further the effect of Criterion A on PD prevalence and its relations to other important clinical outcomes.

A second important point is that the overall prevalence of PD does not change depending on whether one uses the DSM-5-III hybrid system with its PD types and definition of PD-TS, or defines all PD using our PD-TS definitions collectively because, in our data, overall prevalence is determined by Criterion A which always was accompanied by the presence of at least one pathological trait. What does change, of course, is the prevalence of the six PD types, which was 21.5% considering any PD type, fell to 6.3% when our three PD-TS groups were included, and would fall to 0% if PD-TS were the only diagnosis. Some might argue that this loss is catastrophic because it would nullify the considerable research on the DSM-5-II PDs. Others might argue this loss occurs simply by adopting the hybrid model, but we are convinced by data showing that the hybrid model preserves the essential features of the DSM-5-II PDs (e.g., Morey & Skodol, 2013).

In any case, the seriousness of the loss of the six PD types rests, at least in part, on whether there is important information carried in the configural arrangement of the traits that comprise the types. Although some may take this as axiomatic, we are unaware of any research that demonstrates that a PD-type diagnosis adds significant predictive power beyond that afforded by its component traits and, indeed, there is considerable evidence to the contrary (e.g., Morey et al., 2012). Claiming that losing the six PD types is problematic also appears to ignore the fact that most individuals diagnosed with one of these types have either a considerable number of extra-diagnostic traits and/or more than one PD-type diagnosis, a point we return to subsequently. Finally, the loss of the six PD types must be weighed against the clarity gained by regarding individuals’ clinical pictures in terms of their complete trait profiles, and we argue that the latter provides the more clinically useful information.

Comorbidity

Stated bluntly, a PD diagnostic system defined by only PD-TS would make comorbidity irrelevant. Our data replicate previous findings of high comorbidity rates among PD-type diagnoses, one of the major disadvantages of the DSM-5-II PD system, as well as the PD types in the hybrid DSM-5-III model. Specifically, of the individuals who met criteria for a PD type, over half had multiple comorbid PDs (see Table 5). Additionally, the rates of “pure” PD diagnoses reported in Table 4 are strikingly low. Only 9.5% of the full sample and less than a third (31.5%) of the PD sample, had a single “pure” PD diagnosis, meaning they met criteria for only one of the six PD types. Further, if we consider a PD to be “pure” only if the number of extra-diagnostic pathological traits is small, this percentage drops to 9.6% of the PD sample, less than 3% of the full sample. These data demonstrate clearly that the overwhelming majority of people with pathological personalities do not fit prototypical PD-type descriptions and that their personalities are typically more complex than a categorical PD system implies, even when they meet criteria for a single PD-type diagnosis.

Heterogeneity

There are two ways in which individuals diagnosed with the same PD type(s) can present different clinical pictures: Different configuration of their diagnostic traits and differences in their extra-diagnostic traits.

Within-PD heterogeneity

The within-PD heterogeneity created by the polythetic system used throughout DSM-5 is well-known, and the DSM-5-III hybrid system takes some steps to address this problem by making pathological traits themselves the criteria. This is contrast to DSM-5-II, in which the component traits are listed in defining each of the PDs, but not directly used to diagnose individuals. For example, NPD is defined as “a pervasive pattern of grandiosity (in fantasy or behavior), need for admiration, and lack of empathy” (p. 669), but one can receive an NPD diagnosis by meeting, for example, the criteria for just grandiosity and a lack of empathy (and not those for requiring excessive admiration), or the criteria for just needing excessive admiration and a lack of empathy (but not those for a grandiose sense of self-importance). The DSM-5-III model addresses this problem to some extent by making the traits themselves the criteria and leaving open the ways in which the traits are manifest. However, it perpetuates the problem by allowing different combinations of traits to receive the same PD-type diagnosis.

For example, Figure 4 shows the SNAP-2 and PID-5 trait profiles for two individuals, both of whom met DSM-5-III criteria for antisocial, borderline and narcissistic PD, but who met different subsets of borderline PD criteria. Both participants endorsed the borderline-PD criteria emotional lability, hostility, impulsivity, and risk taking, the last three of which are also antisocial PD criteria; additionally, they both endorsed antisocial-PD criteria callousness, manipulativeness, deceitfulness, and irresponsibility, and narcissistic-PD criteria grandiosity and attention seeking. However, the high-risk subsample participant additionally endorsed pathological-level borderline-PD anxiousness, separation insecurity and depressivity, whereas the patient participant did not. In terms of which diagnostic criteria were met, therefore, the “flavor” of the borderline PD in these two participants is rather different. Thus, to the extent that PD-type diagnoses imply that the traits defining them are present, these diagnostic labels can be misleading.

Figure 4. PID-5 and SNAP-2 Trait Profiles of Two Individuals Both Meeting DSM-5-III Criteria for Antisocial, Borderline, and Narcissistic PD.

Note. PID-5 = Personality Disorder Inventory for DSM-5. HR = high-risk. PT = patient. SNAP-2 = Schedule for Nonadaptive and Adaptive Personality-2^nd Edition. Anx = Anxiousness, Emolab = Emotional lability. SepIn = Separation insecurity. Submis = Submissiveness. Persv = Perseveration. RgdPerf = Rigid perfectionism. Suspic = Suspiciousness. Deprsv = Depressivity. Wthdrl = Withdrawal. IntAvd = Intimacy avoidance. RstAff = Restricted affectivity. Hstl = Hostility. Manipl = Manipulativeness. Deceit = Deceitfulness. Grndos = Grandiosity. Attnsk = Attention seeking. Irresp = Irresponsibility. Impulsv = Impulsivity. Distrac = Distractibility. RskTkng = Risk taking. UnusBlEx = Unusual beliefs and experiences. Eccn = Eccentricity. CgPcDys = Cognitive and perceptual dysfunction.

NT = Negative Temperament. MIS = Mistrust. MANIP = Manipulativeness. AGG = Aggression. SLFH = Self-harm. ECCP = Eccentric perceptions. DEP = Dependency. PT = Positive Temperament. EXH = Exhibitionism. ENT = Entitlement. DET = Detachment. DIS = Disinhibition. IMP = Impulsivity. PRO = Propriety. WRK = Workaholism. LOSE = Low self-esteem. SUICP = Suicide potential. Cut point for pathological elevation on the PID-5 = 1.5. Cut point for pathological extremes on the SNAP: < 40 and > 60.

Extra-diagnostic traits

The ability to list extra-diagnostic traits is clearly a strength of the DSM-5-III model. In DSM-IV-TR, extra-diagnostic traits that did not reach threshold for a PD diagnosis could be specified on Axis II, but this was rarely done in practice and, along with the multi-axial system, appears to have been eliminated in DSM-5-II. However, our data indicate that there is considerable personality trait pathology that is not captured by the PD types. That is, those diagnosed with one or more PD types had an average of 4.1 extra-diagnostic traits (see Table 8), and over half of those with four or more PD-type diagnoses had 2–3 extra-diagnostic traits. These results reveal not only the incomplete coverage of the PD types, but also point to heterogeneity that is above and beyond that engendered by different criterion combinations.

Turning again to Figure 4, both participants endorsed “extra” traits Suspiciousness and Eccentricity, but the high-risk participant also endorsed Submissiveness, Perseveration, Distractibility, Unusual beliefs and experiences, and Cognitive and perceptual dysfunction, whereas the patient participant endorsed additionally only Intimacy avoidance. Thus, stating that these participants share three PD-type diagnoses masks substantial differences in their DSM-5-III trait profiles.

Their SNAP profiles are similarly divergent: Consistent with their PD-type diagnoses, they share pathological Mistrust (cf. PID-5 Suspiciousness), Manipulativeness (cf. PID-5 Manipulativeness), Aggression (cf. PID-5 Callousness and Hostility), Exhibitionism (cf. PID-5 Attention seeking), Entitlement (cf. PID-5 Grandiosity), Disinhibition (cf. PID-5 Irresponsibility and Deceitfulness) and Impulsivity (cf. PID-5 Impulsivity and Risk taking) but, in addition, the high-risk participant reported pathological levels of Negative Temperament (cf. PID-5 Anxiousness), Eccentric perceptions (cf. PID-5 Psychoticism), Dependency (cf. PID-5 Submissiveness and Separation insecurity), Workaholism (cf. PID-5 Perseveration), and Low self-esteem (cf. ID-5 Depressivity), as well as notably elevated Positive Temperament and low Detachment (cf. low PID-5 Withdrawal). In contrast, the patient participant endorsed Self-harm and its subscale Suicide potential and low Positive Temperament (cf. PID-5 Depressivity).

The DSM-5-III PD-dimensional model has the notable advantage over that in Section II of providing a framework for specifying these extra-diagnostic traits. However, listing them is optional in Section II, and if the clinician does not choose to do so, or lists only those that s/he considers to have clinical relevance for that particular patient, considerable information is lost. Moreover, even if “extra” traits are listed, the DSM-5-III model does not address the problem of within-PD heterogeneity described earlier. Thus, it appears that only a full dimensional PD-TS system is capable of comprehensively representing individuals’ personality pathology.

PD-NOS

As mentioned previously, PD-NOS has been replaced in DSM-5-II by other specified or unspecified personality disorder. At first it might appear that unspecified PD corresponds best to PD-NOS, as neither requires clinicians to clarify why the diagnosis was chosen. However, despite its label, other specified PD also does not actually require clinicians to specify the particular features comprising the diagnosis, but only “to communicate the specific reason that the presentation does not meet the criteria for any specific personality disorder” (p. 648; APA, 2013). One may argue that specifying the pathological features is implied, but it is not explicitly stated. In any case, important information regarding an individual’s pathological trait profile likely will be lost with the use of the PD-NOS replacements, just as it was with PD-NOS. In contrast, PD-TS resolves the problems associated with these “leftover” categories by providing a means to account for all pathological personality traits in any individual, thus providing maximum information.

Summary and Conclusion

The most useful theoretical, empirical, and statistical models achieve the best combination of maximum information and parsimony. How does the DSM-5-III PD hybrid system fare against using only PD-TS in this regard? First, regarding the amount of information provided, the two systems initially appear equivalent, as long as all additional pathological traits are listed along with the PD types. However, as described earlier, there is still the possibility in the hybrid model of two individuals receiving the same PD-type diagnosis through different sets of pathological traits (e.g., schizotypal PD requires elevation on four of a possible six traits). Thus, although the hybrid model is an improvement over the DSM-5-II PD system because it has a higher ratio of number of traits required to the total number of traits in each diagnostic set (72% vs. 56% for the DSM-5-II PDs), it still suffers from some lack of specificity in diagnostic profiles, whereas a PD-TS diagnosis always provides complete trait information.

Regarding parsimony, the DSM-5-III hybrid model has separate definitions for each of the four Criterion A subdomains (thus, 24 in total) and six algorithms for trait sets ranging from two to seven in number, some of which are complex (e.g., BPD requires four of seven traits, including at least one of a particular set of three traits). In contrast, having PD-TS as the only PD diagnosis is indisputably more parsimonious. Moreover, PD-TS eliminates the problems with categorical types in both DSM-5-II and DSM-5-III, specifically mitigating the effects of comorbidity, heterogeneity, and the vagueness of PD-NOS/Unspecified PD.

Finally, it is important to note that although defining Simple, Mixed, and Complex subtypes of PD-TS was useful as a heuristic to demonstrate how changing the operationalization of PD-TS might affect coverage, prevalence, and comorbidity, in the end, it appears that even these definitions are not necessary, because they lead us to the same conclusion: Criterion A and B together simply, clearly, and comprehensively characterize all individuals with PD via a diagnosis of PD-TS.