Investigating Models of Personality Disorder: Factor Structure, Predictive Validity and Model Translatability of Item-Level Personality Disorder Assessments

There is no consensus on an ideal classification system for diagnosing Personality Disorders (PDs). The DSM-5-TR and the ICD-11 currently feature two dissimilar, competing models as official diagnostic systems for use in clinical practice. This thesis focuses primarily on the PD criteria that make up the official DSM-5 Section II model consisting of ten specific PDs that were retained from the previous DSM-IV. To this end, factor analytical and machine learning methods were applied to item-level assessments of DSM-5 Section II PD criteria and item- level PD severity measures to address prevailing research gaps in the literature. The four presented manuscripts are concerned with a) the factor structure of DSM-5 PD criteria and the structural validity of the DSM-5 Section II model, b) the predictive validity of individual item-level PD criteria compared to scale-level scores, c) the translatability of assessments between the DSM-5 Section II criteria and the DSM-5 Section III and ICD-11 alternative trait models, and d) the translatability of assessments of general PD severity. Manuscript 1 reported on the most comprehensive factor analysis of DSM-5 PD criteria to date and observed structural differences between clinical interviews and self-report measures. A ten PD factor model was a better representation in interviews than in self-reports. Manuscript 2 demonstrated that the predictive validity of individual item-level PD criteria is similar to the predictive validity of aggregated PD scores. Manuscript 3 rearranged items from the SCID-II- SQ into new proxy scales of the DSM-5 Section III and ICD-11 traits. Proxy scales mostly showed acceptable convergent validity, but reliability was less than acceptable for all but the negative affectivity proxy scale. Manuscript 4 demonstrated that six new and established self- report instruments of PD severity can be measured on a common latent severity dimension. The thesis discusses the findings within the context of the ongoing change to a novel trait- based diagnostic system of PDs in the ICD-11 and the prior criticism on the ICD-10/DSM-5 Section II classification systems.