Un ensemble de recherches récentes suggèrent que les troubles de la personnalité décrits par le DSM-IV peuvent être évalués à partir d’un modèle général de description de la personnalité : le modèle à cinq facteurs (five-factor model: FFM). Dans cet article, nous présentons une méthode permettant de calculer des scores pour les différents troubles de la personnalité à partir du FFM. Sur la base de profils de personnalité prototypiques fournis par des experts pour chacun des troubles décrits par le DSM-IV, il est possible d’identifier, puis d’additionner les scores obtenus pour les traits de personnalité caractéristiques de chacun des troubles et d’obtenir un score de type dimensionnel pour chacun d’eux. Les scores ainsi obtenus pour les différents troubles de la personnalité sont comparés (corrélations et efficacité diagnostique) à des évaluations plus explicites et directes de symptômes décrits par le DSM-IV, réalisées sur deux échantillons de patients. Pour la plupart des troubles de la personnalité, à l’exception des troubles de la personnalité obsessionnelle–compulsive et histrionique, la validité convergente, la spécificité, la sensibilité et le potentiel de prédictivité négative (NPP) varient de « modérés » à « satisfaisants ». La validité discriminante, qui permet l’examen de la comorbidité entre les différents troubles, est équivalente à celle obtenue par des méthodes d’évaluation explicites et directes. Nous envisageons finalement les avantages de cette méthode en pratique clinique : facilité de l’évaluation de la personnalité, facilité de la cotation, existence de données de référence normatives.

Widespread and long-standing dissatisfaction with the current DSM-IV diagnostic system for the personality disorders (PDs) revolves around five specific issues: use of a dichotomous, categorical model; extremely high rates of comorbidity within the PDs, as well as with Axis I disorders; excessive heterogeneity within the PDs; nonempirically derived diagnostic cut-offs and limited coverage of the personality pathology seen by clinicians. Many critics have suggested that the personality pathology might be better characterized using a dimensional trait model of general or pathological personality. Much research documents support relations between the PDs and the five-factor model of personality (FFM), a prominent model of general personality functioning. Recent research has examined the ability of FFM trait configurations to assess the DSM-IV PDs. Initial work focused on matching individual FFM profiles to prototypical PD profiles derived from expert ratings. Although this initial work showed that these FFM-assessed PDs performed like explicit PD assessments, the prototype matching approach was deemed cumbersome and a simpler, alternative count technique derived from the expert profiles was developed. The FFM PD counts sum the facets rated as being particularly prototypic (high or low) of a PD.

The current study tests the FFM PD counts by examining their convergent and discriminant validity correlations with explicit measures of DSM-IV PD symptomatology in two clinical samples (one French; one Belgian). In addition, “receiver operator characteristics” (ROC) analyses are used to provide information on the clinical utility of the FFM PD counts.

The French clinical sample consisted of 100 female inpatients hospitalized for treatment of an eating disorder, whereas the Belgian clinical sample consisted of 130 psychiatric inpatients (47% women) with diverse presenting problems. Translated versions of the NEO PI-R were used to assess the FFM in each sample and to generate the FFM PD counts. DSM-IV PD symptom counts and diagnoses were obtained via structured interview in the French sample and via self-reports in the Belgian sample.

In the French sample, convergent validity correlations between the FFM PD counts and DSM symptom counts ranged from 0.20 to 0.65 with a median r of 0.47. In the Belgian sample, these correlations ranged from 0.17 to 0.61 with a median r of 0.45. For the French sample, the median discriminant correlations for each FFM PD ranged from −0.23 to 0.35; for the Belgian sample they ranged from −0.13 to 0.28. Finally, ROC analyses were conducted for the PDs in which 10 or more individuals met the DSM diagnosis. ROC analyses provide “areas under the curve” (AUC), as well as other diagnostic efficiency statistics. An AUC of 0.50 indicates that the FFM PD count is no better at distinguishing between those with and without a given PD diagnosis than chance. AUCs were significant for all PDs examined except for the OCPD, ranging from 0.57 to 0.94 with a mean of 0.75. Average diagnostic efficiency statistics were also generally good.

The FFM PD counts adequately assess the majority of extant DSM PDs. With the recent development of normative data and scoring sheets, these FFM counts can be applied in most clinical settings. Importantly, FFM data can be used flexibly with a focus either on the role of specific domains and facets or on specific trait configurations. Additionally, there are theoretical benefits to conceiving of PDs as constellations of FFM trait. Comorbidity is to be expected to the degree that the PDs include the same traits. The FFM PD counts are inherently dimensional, representing approximations to a prototype; thus, concerns surrounding the PDs as categories are rendered inert. Using the FFM to conceptualize the PDs allows basic research on personality to inform research on the etiology, course and treatment of the PDs.