Assessing DSM–5 Level of Personality Functioning From Videotaped Clinical Interviews: A Pilot Study With Untrained and Clinically Inexperienced Students

Abstract

Several authors have raised the concern that the DSM–5 Level of Personality Functioning Scale (LPFS) is relatively complex and theory laden, and thus might put high requirements on raters. We addressed this concern by having 22 untrained and clinically inexperienced students assess the personality functioning of 10 female psychotherapy inpatients from videotaped clinical interviews, using a multi-item version of the LPFS. Individual raters’ LPFS total scores showed acceptable interrater reliability, and were significantly associated with 2 distinct expert-rated measures of the severity of personality pathology. These findings suggest that, contrary to the previously mentioned concerns, successfully applying the LPFS to clinical cases might require neither extensive clinical experience nor training.

Acknowledgments

This research was funded by grants from the OPD Task Force and the University of Kassel to Johannes Zimmermann. We thank Katharina Rek for helping with the data collection, and Lena Braun, Sabine Gluth, Sara Homburg, Doreen Jeske, Alan Krause, Annika Linder, Kerstin Schrage, Alina Urban, and all other student raters for their great commitment in conducting the LPFS ratings.

Notes

Morey et al. (2013) also conducted receiver operating characteristic (ROC) curve analyses to explore the associations between global LPFS ratings and DSM–IV PD diagnoses. They reported an area under the curve (AUC) of.83 for predicting the presence of a PD, and an AUC of.70 for predicting cases with multiple PDs. To explore how our results compare to their findings, we computed AUCs for each of the 22 raters using the R package “ROCR.” The median AUC across raters for predicting the presence of a PD was.69, and the median AUC for predicting cases with multiple PDs was.70. Thus, the students’ LPFS ratings performed similar to clinicians’ ratings when identifying severely disturbed patients (i.e., patients with more than one PD), but were less efficient when distinguishing patients with and without PDs.

Note that, across the 10 patients, correlations between the number of DSM–IV Axis I disorders and the three focal predictors ranged from.29 to.40. However, due to the small sample size, these associations were not significant, ps >.25. Moreover, OPD–LSIA ratings were significantly associated with the number of PDs, r =.77, p <.01, but not with the presence of a PD, r =.44, p =.21.

Note that the conception of test–retest reliability adopted by the DSM–5 Task Force assumes that the interviews are conducted within a short interval during which the features of the patients are unlikely to change (Kraemer et al., 2012).