Improving assessment of personality disorder traits through social network analysis

ABSTRACT When assessing personality disorder traits, not all judges make equally valid judgments of all targets. The present study uses social network analysis to investigate factors associated with reliability and validity in peer assessment. Participants were groups of military recruits (N5 809) who acted as both targets and judges in a round-robin design. Participants completed self- and informant versions of the Multisource Assessment of Personality Pathology. Social network matrices were con-structed based on reported acquaintance, and cohesive subgroups were identified. Judges who shared a mutual subgroup were more reliable and had higher self-peer agreement than those who did not. Partitioning networks into two subgroups achieved more consistent improvements than multiple subgroups. We discuss implications for multiple informant assessments. In both research and clinical settings, personality disorders (PDs) are most often diagnosed on the basis of self-report, obtained through written inventories or clinical interview. However, information gathered from self-report tends to differ substantially from how others view the individual (Clifton, Turkheimer, & Oltmanns, 2004). A review of 30 published studies of self- and informant reports of This research was supported by a grant from the National Institute of Mental Health (MH51187). We would like to acknowledge the many helpful suggestions contributed by Timothy D. Wilson and Thomas M. Guterbock, and by two anonymous reviewers. This research was originally presented as part of Allan Clifton’s doctoral dissertation. Correspondence concerning this article should be addressed to Allan Clifton, De