Using multidimensional modeling to combine self-report symptoms with clinical judgment of schizotypy

Abstract

This study investigated psychometric properties of two widely used instruments to measure subclinical levels of psychosis, the Community Assessment of Psychic Experiences (CAPE) and the Structured Interview for Schizotypy-Revised (SIS-R), and aimed to enhance measurements through the use of multidimensional measurement models. Data were collected in 747 siblings of schizophrenia patients and 341 healthy controls. Multidimensional Item-Response Theory, Mokken Scale and ordinal factor analyses were performed. Both instruments showed good psychometric properties and were measurement invariant across siblings and controls. The latent traits measured by the instruments show a correlation of 0.62 in siblings and 0.47 in controls. Multidimensional modeling resulted in smaller standard errors for SIS-R scores. By exploiting correlations among related traits through multidimensional models, scores from one diagnostic instrument can be estimated more reliably by making use of information from instruments that measure related traits.