An Overview of Issues Related to Categorical and Dimensional Models of Personality Disorder Assessment

Abstract

Despite long-standing efforts to improve the current diagnostic system for Axis II, problems remain with the categorical conceptualization of personality disorders (PDs). Due in part to these problems, interest has developed in dimensional models of PD classification. In this article, we discuss four issues relevant to categorical vs. dimensional assessment of PDs: (a) problems with self-reports in PD patients, (b) methodological issues in behavioral and clinician assessment of PDs, (c) challenges that arise when dimensional models are applied to patient and nonpatient samples, and (d) clinical implications of categorical and dimensional PD models. We suggest that researchers and clinicians address these concerns to avoid implementing a new PD assessment model that—although different from the current system—would otherwise remain fraught with difficulties.

Notes

¹Outside PD assessment, it is well documented that there are problems in self–other agreement when assessing an individual's psychopathology. Thus, it should not be surprising that such problems occur when assessing PDs. For instance, in a large-scale meta-analytic review, Meyer et al. (2001) reported that convergence among raters (e.g., parents, peers, teachers) in studies with children and adolescents yielded rs of .14 to .34. When diagnostic agreement was reported, kappa values ranged between .13 and .39. In studies of adults, rs ranged between .04 and .44, and kappa values ranged between .12 and .34. A more recent meta-analysis by Achenbach, Kruskowski, Dumenci, and Ivanova (2005) found that the average self-informant correlation among parallel instruments were .43 and .44, respectively, for internalizing and externalizing disorders. When different instruments were used, the mean cross-informant rating was .30.

²The LOESS procedure used by O'Connor (2005) is complex, utilizing a series of local regression analyses that permits the form of a curve to vary across the variable continua … ideal for revealing potentially complex, unanticipated patterns of association between variables. The procedure produces a smoothed, nonlinear curve fit to the data that is analogous to the moving averages that are computed in time series analyses … [and yields an] unbiased depiction of the patterns in the data. (p. 291)

³This conclusion does not imply that clinician or observer ratings of patients should take priority over self-reports. Training and experience notwithstanding, mental health professionals have attributional biases that affect their perception of patients. When Markham and Trower (2003) investigated psychiatric nurses' causal attributions about patient behavior, they found that patients diagnosed with borderline PD were rated as having more personal control over behavior than were patients diagnosed with major depression or schizophrenia. Numerous studies have documented the impact of patient gender on clinicians' interpretation of PD-relevant behavior and on their willingness to assign PD diagnoses (e.g., Widiger & Clark, 2000).

⁴As a result of this integration, there appears to be the rebirth of an important idea from the earliest understanding of PDs. Westen, Gabbard, and Blagov (2006) noted that using personality as a basis for conceptualizing psychopathology represents a return to ideas initially set forth by Freud and other psychoanalytic theorists in understanding the basic interplay between personality structure and (dys)function. Weston et al., (2006) added that an understanding of psychological processes (e.g., object relations, defenses) are quite informative in understanding personality beyond surface traits.