F119. MULTILEVEL ANALYSIS IMPROVES THE MODEL FIT OF THE DIMENSIONAL STRUCTURE OF THE PANSS IN PATIENTS WITH SCHIZOPHRENIA

摘要

BackgroundPrincipal component analyses (PCA) studies show that schizophrenia symptoms are usually grouped into five domains. However, to infer a latent dimensional structure, confirmatory factor analysis (CFA) is more appropriate than PCA. Most CFA studies addressing the five-factor model yielded poor fit indices. One single study achieved a good fit using a multilevel CFA structure with the interviewers as level. Other possible reasons for sample heterogeneity and subsequent poor model adjustments, such as differences in patients’ clinical profiles across clinical units and clinical staging, were not measured in this study. We aimed to replicate the effect of the CFA multilevel analyses and evaluate the possible influence of other heterogeneity sources as levels, i.e., clinical staging, on the Positive and Negative Syndrome Scale (PANSS) five-factor structure.