1. Introduction

In the Alternative Model for Personality Disorders (AMPD) in the Diagnostic and Statistical Manual of Mental Disorders (DSM-5, section III), specific personality disorders (PD) are defined by typical impairments in personality functioning (Criterion A) and characteristic pathological personality traits (Criterion B) [1]. The AMDP retained only 6 of the 10 DSM-IV PDs and changed the Personality Disorder Not Otherwise Specified (PDNOS) category to a PD-Trait Specified diagnosis. The AMDP has not been accepted as the official diagnostic system for use in the DSM-5, partly because of the limited research support for its specific components. Since its publication, more than hundred publications investigated the validity of the AMDP in different languages [3, 7] and it needs further validation and examination whether it is generalizable to Hungarian language and culture. Although another study has already been published about the psychometric properties of the PID-5 in a Hungarian clinical and community sample [19], we found that this study had some psychometric and scientific limitations besides its strengths, and it is worth using another adaptation of the PID-5 mesure. First of all, our Hungarian translation of the PID-5 scale had been approved by the American Psychiatric Association (APA).

Moreover, we added some important extensions to the investigation of the psychometric properties of the PID-5 by including a 1-month test-retest reliability of the scales, by using SCID-II and SCL-90-R correlations as validity criteria, and by presenting three different models of the PID-5 internal structure in the confirmatory factor analysis.

Pathological personality traits (Criterion B), proposed by Krueger, Derringer, Markon, Watson, and Skodol [2] are organized into five broad domains: Negative Affectivity, Detachment, Antagonism, Disinhibition, and Psychoticism. Within the five broad trait domains, there are 25 specific trait facets [1]. APA (https://www.psychiatry.org/psychiatrists/practice/dsm/educational-resources/assessment-measures) provided different kinds of measures (self-report, informant, rating scale) for assessing maladaptive personality traits. The Personality Inventory is a self-report and other-informant measure for the DSM–5 (PID–5) [2]. Since its introduction, several studies examined various aspects of PID-5’s psychometric properties in non-clinical and clinical samples (see a review by Al-Dajani et al. [3]). We aim to further examine the psychometric properties of the Hungarian version of PID-5 in non-clinical and clinical samples. Our decision to use a non-clinical sample as well is based on the theory that traits that describe personality disorders are experienced at a less extreme level among healthy persons [4].

Internal consistencies of PID-5 domains and facets have been studied extensively, and mixed internal consistency scores were demonstrated. Coefficient alphas [5] for domain scales were between .75-.92 and .87-.96 and for facets ranged from .46-.77 to .94-.96, demonstrating some possible issues with scale reliability at the facet level (see Al-Dajani et al. [3]). Two versions of the domain scoring algorithm are used in the literature. The 25 facets version, which uses all the 25 facets in the definitions of the five domains, and the 15 facets version, which uses only 3 facets per domain [6]. Both versions derived from the study of Krueger et al. [2] which identified five higher order factors by explorative factor analysis of 25 facets. In order to eliminate cross-loadings of facets, three traits with the highest factor loadings were selected for each domain [2]. This scoring instruction was published in the APA copyright version of the PID-5 [6]. Further studies that investigated the five-factor exploratory factor model of the PID-5 facets also found facets that have cross-loadings on more than one domain and the number, magnitude and domain location of the loadings for crossloading facets varies across study samples [7].

In an attempt to address the variable nature of PID-5 factor analytic results, Watters and Bagby [7] conducted a meta-analysis of the PID-5 lower-order factor structure. They found that the level of crossloadings decreased when multiple samples were combined, and that the 15 facets scoring version of domains showed the lowest number of cross-loading facets (only anhedonia loaded on two domains substantially). A comparison of the factor structures of the three models is in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Fit indexes of CFA of factor structures of PID-5.

https://doi.org/10.1371/journal.pone.0266201.t001

One of our objective is to compare the factor structure of the 15 and 25 facets scoring version by confirmatory factor analysis. We also intend to quantify whether substantive differences in results occur because of using different domain scoring methods. In order to compare the different methods used in research examining the psychometric properties of PID-5 with exploratory or confirmatory factor analysis, Table 2 presents an overview of methods for such studies. This mini-review is also a means to compare the results of studies that analyze the psychometric properties of PID-5 in different languages.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Overview of methods for studies examining the psychometric properties of PID-5 with exploratory or confirmatory factor analysis.

https://doi.org/10.1371/journal.pone.0266201.t002

We found one study that examined the test-retest reliability of the PID–5. Wright et al. [8] found that across an average of 1.44 years, Cohen’s d ranged between .17 (Antagonism) and 0.00 (Negative Affectivity) for the domains whose range is in the little to no change area according to Cohen’s guidelines [9] and between little and small change was found for the facets (d = .02 (Depressivity) and -.30 (Submissiveness).

The current study aimed to examine the psychometric properties of the Hungarian PID-5 facet and domain scales among Hungarian-speaking adults from a non-clinical community sample as well as in a Hungarian clinical sample. Specifically, we analyzed its facets and domains’ internal consistency and test-retest reliability. In addition, we tested the assumption of whether the DSM-5 traits are indeed maladaptive personality traits by analyzing their correlations with general personality disorder severity measured by the sum of the SCID-II screening questionnaire and general distress measured by SCL-90-R GSI. We also analyzed the specific associations between PID-5 domains and facets with personality disorder dimensions measured by the SCID-II screening questionnaire and the nine symptom dimensions of SCL-90-R. Finally, we analyzed the associations between PID-5 traits and age, gender, and education.

2. Methods

2.1. Participants and procedure

471 non-clinical participants were recruited by 20 psychology students who received class credit for their work. Each student aimed to collect data by using snowball sampling techniques by reaching around 50 volunteers with no known psychiatric disorders. Data was also gathered from 314 clinical patients of Semmelweis University’s Department of Psychiatry and Psychotherapy. Participants were stratified according to gender and age range. The majority of the non-clinical participants were female (69,6%) and their mean age was 37.63 years (SD: 12.96, their ages were between 18 and 74 years). The majority of the clinical participants were female as well (69,7%) and their mean age was 37,41 years (SD: 13,34, their ages were between 18 and 74 years). Every participant was Caucasian.

Based on SCID II interview, 173 participants (55,1%) in the clinical sample had at least one diagnosis of personality disorder, 117 subjects (37,3%) had multiple diagnoses (two or more personality disorders). No information was available on the clinical history of 64 individuals. The distribution of the different types of personality disorder was as follows:

Avoidant 21,6% (54 subjects); Dependent 16,06% (40 subjects); Obsessive-compulsive 18,145% (45 subjects); Passive-aggressive 13,2% (33 subjects); Depressive 24,4% (61 subjects); Paranoid 14,8% (37 subjects); Schizotype 2,4% (6 subjects); Schizoid 1,2% (3 subjects); Histrionic 14% (35 subjects); Narcissistic 17,6% (44 subjects) and Borderline 30,8% (77 subjects).

Participants completed all questionnaires online in Google-docs, a secure online survey tool. Each participant gave written consent to participate by choosing the "agree with" option in response to an informed consent form.

In the clinical sample some of the data was missing, and these missing data were not replaced, therefore the sample sizes vary in each analysis. There were 92 individuals who did not mark an answer for at least one item of the PID-5 questionnaire. One participant was excluded from the analysis due to too much missing data (50%), for the others, the number of missing data did not exceed 10%. The sample size was the lowest (222) in the analyzes where all PID-5 items were included in the confirmatory factor analysis (CFA). In CFAs when the complete factor structure was analyzed, the sample sizes varied between 222–241. Sample sizes ranged from 295 to 307 for item-level analyzes and from 265 to 277 for domain-level analyzes. No data were missing in the analyzes on facet- and domain scales, as the scores on the scales were calculated by averaging.

In addition, 93 (62 female) non-clinical subjects participated in the test-retest reliability study. The age of the participants in the test-retest reliability study varied from 19 to 62 years (Mean: 40.32; SD: 13.08). The descriptive statistics of the samples can be found in Table 3. The local Ethical Committee of the ELTE PPK (2018/8) approved the study protocol. The Regional and Institutional Committee of Science and Research Ethics of Semmelweis University approved the research procedure.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Descriptive statistics of the samples.

https://doi.org/10.1371/journal.pone.0266201.t003

3. Results

3.1. Reliability

3.1.1. Internal consistency.

Internal consistency results are reported in Tables 4 and 5. Alpha coefficients for the Hungarian PID-5 domain scores ranged from .88 (Detachment) to .95 (Psychoticism) in the non-clinical sample, and from .90 (Antagonism) and .95 (Psychoticism) in the clinical sample. For the facets, scores ranged from .53 (Suspiciousness) to .95 (Eccentricity) in the non-clinical sample, and from .71 (Suspiciousness and Restricted Affectivity) to .95 (Eccentricity) in the clinical sample. The mean alpha for the domain scores was .90 (non-clinical) and .92 (clinical) and .81 (non-clinical) and .84 (clinical) for the facet scores. In summary, most scales have appropriate reliability, except for the Suspiciousness facet in the non-clinical sample, the reliability of which is low.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Internal consistency and item level CFA of Personality Inventory for DSM-5 facets and domains.

https://doi.org/10.1371/journal.pone.0266201.t004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Internal consistency and item level CFA of Personality Inventory for DSM-5 scales on clinical samples.

https://doi.org/10.1371/journal.pone.0266201.t005

3.1.2. Test-retest reliability and magnitudes of change.

As it is shown in Table 4, internal reliability coefficients of PID-5 dimensions on test-retest sample were between .64 for Irresponsibility, .96 for Eccentricity. Test-retest correlation coefficients (Pearson) were between .53 (Callousness) and .86 (Risk-taking), which reflects a moderate and strong connection between the two measurements. Utilizing Cohen’s d to establish magnitude of change within a one month interval, a small amount of change (i.e., d of .2–.4; Cohen, 1988) was observed in Negative Affectivity (d = –.30), and in Disinhibition (d = –.20) domains, and in Anxiousness (d = –.22), Separation Insecurity (d = –.21), Withdrawal (d = –.22), Restricted Affectivity (d = –.23), Distractibility (d = –.33). In other domains and facets, we found a smaller d than–.20, which reflects little to no change according to Cohen’s guidelines (Cohen, 1988).

3.1.3. Confirmatory factor analysis.

We ran a Confirmatory Factor Analysis in order to explore the factor structure of PID-5 traits; results are reported in Tables 4 and 5. We analyzed the fitting of all the three factor structures found in the literature (the APA model [1], the early Krueger model [2] and the Watters-Bagby model [7]). The results showed that the RMSE and SRMR fit indices reach the good or acceptable level in both samples in case of all the three models, while CFI scores indicate a poor fit in all samples and models in almost all domains. Based on the APA model, the Disinhibition scale fits well in both samples, and Detachment fits well in the clinical sample. According to the results of item-level CFA, CFI of the PID-5 facet scales ranged from .81 to .99 in the non-clinical sample, and from .63 to .99 in the clinical sample. The values of RMSEA ranged from .01 to .22, and the SRMR ranged from .01 to .13 in the non-clinical sample, and in the clinical sample RMSEA ranged from .04 to .15, while the SRMR ranged from .01 to .15. (For more detail, see Tables 4 and 5.) These results indicate poor fit of 11 out of the 25 PID-5 facets in the non-clinical sample, and of 13 out of the 25 facets in the clinical sample. The scales that are found to fit the least in both samples are Emotional Lability, Unusual Beliefs, Depressivity, Perseveration, Restricted Affectivity and Risk Taking.

3.2. Validity

3.2.1. Correlations between PID-5 traits and SCID-II as well as SCL-90-R scores.

Correlations between PID-5 traits and SCID-II personality disorder scores and SCL-90-R scores are reported in Tables 6 and 7. We found that correlations of PID-5 domains and SCID-II total PD score are significant in both the clinical and the non-clinical groups with a Bonferroni adjusted p-value. Effect sizes are moderate or strong. Correlations between the traits and SCID-II total score are stronger in the clinical sample, where correlations are moderate to strong, except for weak correlations with Intimacy Avoidance, Submissiveness and Risk Taking. In the non-clinical sample, correlations between the traits and SCID-II total score are mostly moderate, with the exceptions of avoidance, Submissiveness, Risk Taking and Rigid Perfectionism, Restricted Affectivity and Attention Seeking, which are also weakly correlated.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Pearson correlations between PID-5 traits, SCID-II total personality disorder score and DSM-IV personality disorders proposed to be retained in DSM-5 measured by SCID-II.

https://doi.org/10.1371/journal.pone.0266201.t006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Pearson correlations between PID-5 traits and SCL90-R scales.

https://doi.org/10.1371/journal.pone.0266201.t007

Different personality disorder scores of SCID-II-PQ are significantly correlated to the expected PID-5 domains and traits with moderate or strong effect size, however, the correlations are stronger in the clinical sample. There are only a few exceptions: Intimacy Avoidance has only a weak relationship with Avoidant- and Obsessive-Compulsive PD. Restricted Affectivity is weakly correlated to Schizotypal- and Obsessive-Compulsive PD and the correlation between Borderline PD and Risk Taking is also weak.

Total score of SCL-90-R has significant (Bonferroni corrected) correlations with all PID-5 scales, except for Risk Taking in the clinical sample. Correlations are moderate to strong, except for Attention Seeking, Callousness, Grandiosity, Impulsivity, Irresponsibility, Restricted Affectivity in both samples as well as Deceitfulness and Submissiveness in clinical sample. The effect sizes of these latter correlations are weak.

At domain level, the PID-5 negative affectivity, detachment, disinhibition and psychoticism show significant, moderate or strong correlations with all the SCL90-R scales. PID-5 antagonism has weak relationship to SCL-90 R scales.

With regards to the facet level, as expected, the strongest relations were found between the PID-5 and the SCL-90-R counterparts. PID-5 anxiousness shows moderate correlation with the SCL-90-R anxiety scale and the PID-5 depressivity displays high correlation with the SCL-90-R depression scale. Furthermore, PID-5 hostility is correlated with SCL-90-R hostility, while PID-5 suspiciousness correlates with the SCL-90-R paranoid ideation, all of them with moderate effect sizes.

However, there are several PID-5 traits that do not correlate with some of the SCL-90-R dimensions in the clinical sample, these are: callousness, deceitfulness, grandiosity, intimacy avoidance, manipulativeness, risk taking and submissiveness.

3.2.2. Differences between clinical and non-clinical samples, gender, age and education.

Associations between PID-5 domains and facets comparing clinical and non-clinical groups, gender differences, as well as correlations with age and education are reported in Table 8.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 8. Descriptives of PID-5, differences between clinical and non-clinical samples, gender differences, correlation with age and educational level.

https://doi.org/10.1371/journal.pone.0266201.t008

3.2.2.1. Clinical and non-clinical differences. MANOVA was run to analyze clinical and non-clinical differences along with the 25 personality disorder traits and the 5 domains. Based on the 25 traits we found significant differences with a large effect size (Wilks’s l = .449; F (25, 759) = 37.31; p < .001; h² = .551). More specifically, the Manipulativness, Deceitfulness and Risk Taking traits did not show significant differences between the clinical and non-clinical groups; for all the other traits, scores were significantly higher in the clinical group. At the domain level, the difference was also significant with a large effect size (Wilks’s l = .564; F (25, 779) = 120.606; p < .001; h² = .436).

3.2.2.2. Gender differences. According to our results, the male and female profiles of personality disorder traits were significantly different and the difference had a large effect size in both the clinical (Wilks’s l = .755; F (25, 288) = 3.735; p < .001; h² = .245) and the non-clinical sample (Wilks’s l = .70; F (25, 445) = 7.64; p < .001; h² = .30). Significant difference between males and females was found at the domain level as well, and the difference had a large effect size in both the clinical (Wilks’s l = .896; F (5, 308) = 7.14; p < .001; h² = .104) and the non-clinical (Wilks’s l = .857; F (5, 465) 15.564; p < .001; h² = .0143) sample. Along the higher order domains, in the non-clinical sample we found significant differences between women and men in every domain except for Disinhibition; females scored significantly higher in the Negative Affectivity domain and lower in the other domains. However, based on the effect sizes, difference in the Negative Affectivity domain is trivial, in Detachment and Psychoticism difference is small, and in Antagonism the difference is moderate. In the clinical group at the domain level we found significant differences only in the Negative Affectivity domain, where females scored higher, and in the Antagonism domain, where males scored higher. On the facet level, there are significant gender differences along 17 traits in the non-clinical group; effect sizes are between .011 and .12.

3.2.2.3. Age and education. There are several traits and domains showing significant correlations with age, but the correlation coefficients are very weak, except for the Deceitfulness trait and the Antagonism domain related to it, where we found a weak negative correlation, mostly in the clinical sample.

We also found some significant but very weak correlations with education on both the domain and the facet levels. There is a weak negative correlation between Suspiciousness and education in the clinical sample (r = 28).

4. Discussion

The results of the present study supported the hypothesis that the Hungarian translation of the PID-5 is a reliable and valid measure of the proposed DSM-5 maladaptive traits, and our findings provide valuable information for assessing personality pathology in non-clinical and clinical populations. Specifically, we found that

1. based on the Cronbach alphas, most of the scales have appropriate reliability in both samples, except for the Suspiciousness facet in the non-clinical sample; however, item-level CFA resulted in several poorly fitting trait scales
2. test-retest reliability estimates and effect sizes reflect small to no changes;
3. we did not find significant differences among the three different models in terms of the factor structure of PID-5 traits, however, the APA model was found to fit slightly better than the others
4. validity of the PID-5 instrument is acceptable in both the clinical and the non-clinical sample, as PID-5 scales show a significant correlations with the equivalent scales of SCID-II and SCL-90-R
5. there are significant differences between the personality disorder traits of clinical and non-clinical populations
6. male and female profiles of personality disorder traits were significantly different in both samples, but in the non-clinical sample there are more significant differences at the facet and the domain level as well;
7. there are significant negative correlations between age and PID-5 domains in the non-clinical sample, except for the Detachment domain; and in the clinical sample, Antagonism and Disinhibition is significantly and negatively correlated with age
8. we did not find strong significant associations between education and PID-5 traits and domains

Based on our results our conclusion is that we recommend the Hungarian version of the PID-5 for future clinical research and practice. In the following, we highlight selected findings that contribute to the current literature on the PID-5 and the new DSM-5 trait model in general.

First, according to the analysis of the internal consistency index (Cronbach alpha) and the item-level, CFA fit indexes of PID-5 facets, six scales of the Hungarian PID-5 were poorly fitting based on several fit indices in both samples. These are Emotional Lability, Unusual Beliefs, Depressivity, Perseverance, Risk Taking, and Restricted Affectivity. Some of the fit indices of Emotional Lability and Perseverance did not meet minimal standards in other studies either [18]. Restricted Affectivity and Risk Taking also showed inadequate model fit in another research analyzing the psychometric properties of another translation of the Hungarian PID-5 [3, 19]. In our research, the only scale with an unacceptable Cronbach alpha level was Suspiciousness (.54) in the non-clinical sample. The reliability of the Suspiciousness scale was weak in several other psychometric analysis of PID-5 [3], suggesting that this is not a specific characteristic of the Hungarian version of PID-5. We analyzed the factor loadings of the items of the 6 problematic scales and some of them were very low, under .2 (Emotional Lability: item 18, 138, 181, Unusual Beliefs: item 143, Depressivity: item 86, Perseverance: item 46, Restricted Affectivity: item 45, Risk taking: items 98, 164). It might be necessary to revise and rewrite these problematic items in order to increase reliability.

Second, temporal consistency of the Hungarian PID-5 has been confirmed by a one-month test-retest reliability analysis. Based on Cohen’s d values, small to no change was observed along with PID-5 facet scales within the one-month interval. The small amount of change observed in the Negative Affectivity (d = –.30), and in the Disinhibition (d = –.20) domains is similar to the findings of Wright and colleagues [8] who observed a small degree change in 6 out of the 25 traits after 1,5 years. Wright considers that the small negative change might be explained by some amelioration of symptomatology and shows that although PD traits are highly stable, some traits might be dynamically associated with the change of psychosocial functioning over time. In our study, test-retest correlation coefficients were between .53 (Callousness) and .86 (Risk-taking), which reflects a moderate to strong connection between the two measurements. In comparison with other self-report measures of psychopathology in other studies [20], PID-5 scales have similar stability coefficients.

Third, we did not find significant differences among the three different models in terms of the factor structure of PID-5 traits. According to our analysis, those models that include all the 25 traits fit weakly in every domain based on the CFI scores. However, the APA model—that builds up the five domains out of 15 facets—, was found to fit slightly better than the others. The Disinhibition domain proved to be a good fit in both samples, and the Detachment domain was fitting well in the clinical sample based on every fit index. The weak CFI scores could be explained by that many items of the scales load onto several factors simultaneously, however, this does not mean that the scales are inconsistent because they have acceptable Cronbach alpha levels.

Fourth, validity of the PID-5 instrument is acceptable in the clinical and the non-clinical sample as well. PID-5 scales are significantly correlated with SCID-II total PD score and SCL-90-R global severity index. Furthermore, different personality disorder scores of SCID-II-PQ are significantly correlated with the expected PID-5 domains and traits with moderate or strong effect size. Moreover, PID-5 domains show significant correlations with all the SCL90-R scales, and as expected we found strong relations between the PID-5 facets and their SCL-90-R counterparts. These results support the hypotheses that the Hungarian PID-5 is a valid measurement of personality pathology.

However we found some unexpected results as well. PID-5 Intimacy avoidance shows only a weak, almost negligible relationship with Avoidant PD, and has no significant correlations with 7 out of the 9 SCL-90-R dimensions, but only in the clinical sample. According to these results, further studies are needed to understand how PID-5 Intimacy avoidance scale can describe the pathological avoidance behavior. We also found that PID-5 Risk taking scale does not correlate significantly with Global Severity Index of SCL-90-R, and 8 of the 9 SCL-90-R subscales in the clinical sample, and although its relationship is significant with Borderline PD, it is only with weak effect size. Furthermore, Risk Taking is one of the four PID-5 traits that does not differentiate significantly between clinical and non-clinical samples. Our results support that Risk taking is a reliable scale of PID-5, but the question arises as to how pathological it can be considered. There are two other PID-5 traits, Deceitfulness and Manipulativeness, which are not correlated significantly with many SCL-90-R dimensions. It suggests that these traits might be independent from psychological distress.

Fifth, we found significant differences among clinical and non-clinical participants in four domains (no difference in Antagonism) and in 22 out of 25 facets, with mostly medium effect sizes. (See Table 8.) In the majority of comparisons, clinical participants received significantly higher scores than non-clinical participants. No significant differences were observed for the facets of Deceitfulness, Manipulativeness and Risk Taking between the two samples; these facets all belong to the Antagonism domain. This finding is only partially in line with Bach and colleagues’ findings [21], where they observed no significant differences between clinical and non-clinical groups for the facets of Manipulativeness and Attention Seeking. Scores in the Antagonism domain are relatively low in both samples, meaning that antisocial personality functioning was not characteristic in our clinical sample either. The low level of Antagonism domain in the clinical sample might be explained by the fact that our clinical sample consisted of the participants of the four weeks long psychotherapy program of our department and severe antisocial traits and the diagnosis of antisocial personality disorder with traits of primary psychopathy are exclusion criteria for our treatment program.

Sixth, we found significant differences between the profiles of females and males in PID-5 domains and facets in both samples. Females presented higher scores in the Negative Affectivity domain in both samples, while men score significantly higher in the Antagonism domain in both samples. However, based on effect sizes, the difference between genders in Negative Affectivity is small. In this domain, Emotional Lability scores are higher in females and the difference has medium effect size, but with Anxiousness and Separation Anxiety differences are negligible.

These results are partially in line with the findings of Bastiaens et al. [22] on a clinical sample, where females scored higher on Negative Affectivity and lower on Antagonism than males. However, Bastiaens and colleagues found that females scored higher on Disinhibition, which domain showed no difference in neither of our samples. Moreover, in our non-clinical sample at the facet level males scored higher on the Irresponsibility, Distractibility, Manipulativeness and Risk Taking scales, which is in accordance with the results of Fossati et al. [23], and with reports of higher scores on Conscientiousness in females [24, 25].

Seventh, regarding age, a significant negative relation was found with Negative Affectivity, Antagonism, Disinhibition, and Psychoticism in the non-clinical sample. In the clinical sample, only Antagonism and Disinhibition showed significant negative relation with age. In previous PID-5 studies, inverted relations were found between age and the facets of Antagonism, Disinhibition and Psychoticism [23]. In a different study, similar associations were found in a clinical sample [22]. In a non-clinical sample, again, an inverted relation was found between age and Antagonism and Disinhibition [26]. Our results are in line with the results of longitudinal studies in the US and cross-sectional studies from many cultures that were using self-report measures of personality traits, namely that the broad factors of Neuroticism, Extraversion, and Openness to Experience decline from adolescence to adulthood, whereas Agreeableness and Conscientiousness increase [27].

Eighth, we did not find strong significant associations between education and PID-5 traits and domains. This is contradicted by some earlier findings, for example the study of Bastiaens et al. [22] found that Disinhibition was significantly negatively related to education.

Our results are limited by that our measure of personality disorders (SCID-II-PQ) was based only on self-reports. We did not measure the symptoms of Antisocial personality disorder since our ethical committee did not allow us to use the antisocial modul of the SCID-II-PQ for legal and ethical reasons. There was many data missing in the clinical sample, which led to a smaller sample size than originally planned.

Even in light of these limitations, our study shed new light on some aspects of PID5, as well as replicating some previous results [12, 28] in a linguistically and culturally different non-clinical and clinical sample.