The construct validity of general and specific dimensions of personality pathology.

Abstract

Modern personality disorder (PD) theory and research attempt to distinguish transdiagnostic impairments common to all PDs from constructs that explain varied PD expression. Bifactor modeling tests such distinctions; however, the only published PD criteria bifactor analysis focused on only 6 PDs and did not examine the model's construct validity. We examined the structure and construct validity of competing PD criteria models using confirmatory and exploratory factor analytic methods in 628 patients who completed structured diagnostic interviews and self-reports of personality traits and impairment. Relative to alternative models, two bifactor models - one confirmatory model with 10 specific factors for each PD (acceptable fit) and one exploratory model with four specific factors resembling broad personality domains (excellent fit) - fit best and were compared via connections with external criteria. General and specific factors related meaningfully and differentially to personality traits, internalizing symptoms, substance use, and multiple indices of psychosocial impairment. As hypothesized, the general factor predicted interpersonal dysfunction above and beyond other psychopathology. The general factor also correlated strongly with many pathological personality traits. The present study supported the validity of a model with both a general PD impairment dimension and separate individual difference dimensions; however, it also indicated that currently prominent models, which assume general PD impairments and personality traits are non-overlapping, may be misspecified.