Participants and Procedures
Participants were selected from a database of French-Canadian outpatients recruited during the intake procedure at a psychiatric outpatient clinic specialized for PD assessment and treatment, in Quebec City, Canada. All were referred to the treatment center for an initial evaluation of suitability in the outpatient treatment program, following a reference by a psychiatrist or general physician for a suspected PD. Patients were first asked to complete a computerized self-report battery of questionnaires, and then took part in a clinical interview led by one of the team's psychologist, who produced a detailed evaluation report. In line with the objectives of the present study, a total of 211 patients[1] (133 women, Mage = 33.66, SD = 10.97) were selected based on the presence of at least moderate borderline pathology according to cutoffs recently proposed by Kleindienst et al. [43] for the short version of the Borderline Symptom List ([44]; see below). According to the proposed grades of borderline symptom severity, 26.5% had a moderate level of pathology (score 0.7–1.7; n = 56), 40.8% had a high level of pathology (1.7–2.7; n = 86), 25.1% had a very high level of pathology (2.7–3.5; n = 53), and 6.6% had an extremely high level of pathology (3.5–4; n = 14). Almost all participants (97.2%) were of Caucasian-White ethnicity. Half (49.8%) were unemployed or on disability leave, while the others were full-time or part-time workers (34.6%), students (12.8%), or pensioners (2.8%). A majority (66.4%) were single, divorced, or widowed.
Measures
Self-reported Variables
The short version of the Borderline Symptom List (BSL-23 [44]; French validation by Nicastro et al. [45]) is a 23-item self-rating instrument assessing Borderline PD symptomatology. The instrument was used to guide participant selection, i.e., to identify prospective participants with at least moderate borderline pathology; it was also used as an external comparator between profiles. The BSL-23 covers DSM Section II BPD diagnostic criteria (e.g., affective instability, suicidality, transient psychotic symptoms) in addition to other affective experiences typical of borderline pathology (e.g., proneness to shame, self-criticism, loneliness, mistrustfulness, and helplessness), which should provide a large span of borderline pathology clinical presentations. The severity grades proposed by Kleindienst et al. [43] received robust empirical support from established assessments for psychopathology across three independent samples. Items are scored on a five-point Likert scale. The global score (MacDonald’s Omega [ω] = .91) was used in the present study.
The Self and Interpersonal Functioning Scale [32] is a 24-item self-report measure of the AMPD Criterion A. Items are rated on a five-point Likert scale (higher scores indicate higher dysfunction). It provides a global personality dysfunction score (ω = .80) and four subscale scores: Identity (ω =.60), Self-direction (ω =.68), Empathy (ω =.66), and Intimacy (ω =.68). Previous research on the SIFS using CFA yielded a second-order model, with four elements organized into a higher-order personality dysfunction factor [32]; meaningful patterns of associations with related psychological constructs were found for the four SIFS subscales. Content validity analysis of the SIFS items also showed promising results, and the severity level assessed by its items makes it very well suited to study populations with greater psychopathology [46]. In the present study, the four elements were used as LPA indicators, while the global score was used to contrast profiles.
The short form of the Personality Inventory for DSM-5 (PID-5-SF [47]; French validation by Roskam et al. [48]) is an abbreviated 100-item self-report version, based on item-response theory, of the original 220-item PID-5 [49]. It covers 25 pathological personality facets, hierarchically organized into five domains: Negative Affectivity (ω = .72), Detachment (ω = .84), Antagonism (ω = .91), Disinhibition (ω =.74), and Psychoticism (ω =.85). The official American Psychiatric Association scoring method (i.e., using only three facets per domain; see [50]) was used to determine domain scores. Items are rated on a four-point Likert scale. In the present study, the seven facets (ω range = .74 [Depressivity] to .89 [Impulsivity]) that define BPD in the AMPD model were used as latent indicators, whereas the other 18 facets (ω range = .66 [Irresponsibility] to .91 [Attention-seeking]) and the five domains were used in subsequent analyses to describe and contrast profiles.
Along with a sociodemographic questionnaire, other instruments were used to assess internalized as well as externalized pathological features commonly encountered in BPD, for profile characterization and comparison:
The 12-item short-form Buss-Perry Aggression Questionnaire (BPAQ-SF [51-52]; French validation by Genoud and Zimmerman [53]) covers four manifestations of aggression: Verbal (ω = .63), Physical (ω =.86), Anger (ω = .80), and Hostility (ω = .70). It also yields a global Trait Aggression score (ω = .86). Items are scored on a seven-point scale.
The 28-item Interpersonal Reactivity Index–French Version (IRI [54]; French validation by Gilet et al. [55]) measures empathy and its components. Two of its subscales were used in the present study: Perspective taking (the ability to adopt others’ point of view; ω = .83), which assesses the cognitive component, and Empathic concern (the motivation to care about others; ω = .80), which focuses on the affective component. Items are scored on a seven-point scale.
The Barratt Impulsiveness Scale (BIS-11 [56]; French validation by Baylé et al. [57]) is a 30-item questionnaire designed to assess three components of impulsiveness: Attentional (ω =.60), Motor (ω =.75), and Nonplanning (ω =.69). Items are scored on a four-point scale.
File-Rated Variables
Patient files were reviewed by two authors of the present study (DG, CS), to retrieve information pertaining to aggression, suicide attempts, and self-harm. Both raters have significant clinical experience with PD treatment (respectively 17 and 12 years). They both scored 20 randomly selected files for interrater agreement purposes (intra-class correlation [ICC] for aggression = 1.00; suicide attempts: ICC = .98, 95% CI [.94–.99]; self-harm: ICC = .84, 95% CI [.60–.94]). All remaining files were then scored by only one of the authors (DG). Most files included at least one detailed evaluation report with information pertaining to the three target clinical indicators. Detailed evaluation reports were missing from 43 files. The reason for this absence was either (a) that prospective patients only completed the first portion of the intake procedure from the clinic (i.e., self-report computerized questionnaires); or (b) the evaluation report had not been completed or archived yet by the psychologist. For each indicator, raters used a three-point scale to assess antecedents of aggression, suicide attempts, and self-harm, inspired by the Historical Clinical Risk Management-20, Version 3 (58): no prior aggression/suicide attempts/self-harm (0); possible (1), corresponding to rare or minor acts (i.e., one or two minor acts of violence or self-harm that did not cause/intend to cause serious injury or that did not result in a hospitalization); or confirmed (2), corresponding to repeated or severe acting outs (i.e., at least three occurrences of minor acts, or one severe act causing/intending to cause injuries, death, or that resulted in a hospitalization). Raters were allowed to score mid-points (i.e., 0.5 and 1.5) to maximize score discrimination.
Statistical Analyses
A Latent profile analysis was performed to determine the presence of distinct profiles of personality functioning using Mplus version 8.4[2] [59]. Latent profiles were evaluated using the four SIFS elements (Identity, Self-direction, Empathy, and Intimacy) and the seven AMPD BPD facets (Anxiousness, Depressivity, Emotional lability, Hostility, Impulsivity, Risk taking, Separation insecurity) as parameters. After data standardization, latent models for six different class solutions were evaluated. Optimal class solution was determined based on model entropy (with a score between .8 and 1.0 indicating adequate classification precision), Akaike (AIC) and Bayesian (BIC) Information Criterion, Sample-Size Adjusted-BIC (SABIC), and Lo-Mendell-Rubin Adjusted Likelihood Ratio Test (LMRT). Lower values for the AIC, BIC, and SABIC metrics are indicative of a better-fitting model, while a significant difference on the LMRT between consecutive class solutions (i.e., k vs. k −1) suggests that the k class solution shows a better fit than the k − 1 solution [60]. Interpretability of the solution also factored in the decision for model selection.
In a second step, latent profiles from the retained solution were contrasted on sociodemographic variables, comorbid AMPD personality disorders, and on multiple variables meaningful for BPD (PID-5 domains and non-borderline facets; BPAQ-SF scales; IRI Perspective taking and Empathic concern; BIS-11 impulsiveness scales; and clinician-rated aggression, suicidality, and self-harm), using Kruskal-Wallis tests considering the sample size differences and that normality assumption was not met for multiple variables. For contingency tables involving categorical data, Chi-square analyses were used.
[1] An additional participant was excluded from further analyses, as he was deemed uncooperative upon careful examination of his scores (with a highly implausible score of 0 on all four Criterion A elements).
[2] All other analyses were executed with SPSS 26.0, with the exception of ω computation, which was done with JASP 0.13.1.