The use of the Positive and Negative Syndrome Scale to assess psychiatric symptomatology in a mixed-diagnosis sample

Abstract

The Positive and Negative Syndrome Scale (PANSS) has long been widely used to assess the extent and severity of psychiatric symptomatology in research studies with samples diagnosed with schizophrenia, particularly in pharmacological studies. Recently it was used in a federally funded multisite study with a mixed-diagnosis sample of persons with serious mental illness. Though numerous factor analytic studies of the PANSS (both exploratory and confirmatory) have been undertaken in samples with schizophrenia, its use in mixed-diagnosis samples characteristic of those encountered in research on psychiatric rehabilitation has not been much studied. This paper reports on a series of confirmatory factor analyses undertaken to explore the fit of various models found in the schizophrenia literature to data from the more heterogeneous multisite sample. Two of the models tested attained marginal fit to the data (using the conventional fit criteria of CFI > .90 and RMSEA), however, both failed to represent a broad range of symptoms because they incorporated only a few PANSS items. An alternative model was developed using all 30 PANSS items and tested using parceling to improve item fit. This 5 factor model demonstrated adequate fit to the data. These findings are discussed in light of the need for a model that fits a mixed-diagnosis sample.