RESUMEN
The 30-item Positive and Negative Syndrome Scale (PANSS) is used worldwide in the assessment of symptom severity in schizophrenia. The present study uses confirmatory factor analysis (CFA) to compare three different factorial models and to evaluate the best-fitting representation of schizophrenia symptom structure on the PANSS across four samples of patients diagnosed with schizophrenia from the US (the CATIE schizophrenia trial), São Paulo, Brazil, and from Beijing and Changsha, China. We examine the goodness of fit of several previously proposed models. The traditional trifactorial model for the PANSS and two five-factor models were evaluated using absolute and incremental indices. Single group CFA found that the five-factor model proposed by NIMH researchers based on an extensive literature review demonstrates the best fit in each of the four samples. This model used 20 of the 30 PANSS items grouped into five factors: positive, negative, disorganized, excited, and depressed symptoms. Subgroups defined by age, gender, nationality, hospitalization status, and severity of illness also did not differ in overall symptom structure as assessed by several standard indices. Our findings suggest that the five factor NIMH model showed the best representation among all four samples from different countries and potentially contrasting cultures.