Background: The factor structure of four competing models of positive symptoms and their clinical validity was studied in a sample of 253 schizophrenia inpatients.
Methods: The following models were tested using confirmatory factor analysis: a one-dimension severity model, a two-dimension model comprising a psychosis factor and a disorganization factor, a four-dimension model based on the Scale for the Assessment of Positive Symptoms (SAPS) structure in subscales, and a five-dimension model derived from the previous one by further differentiating Schneiderian delusions from non-Schneiderian ones.
Results: More complex multifactorial models fit the data better than simpler models. The five-dimension model was the best adjusted (goodness of fit index = .844, nonnormed fit index = .812, normed fit index = .728). Whereas the one-dimension model did not display significant association with the clinical variables, multidimensional models were related to age at onset and illness severity. The two-dimension model captured well the clinical correlates of the more complex models.
Conclusion: None of the tested models showed good fit to the data. The one-dimension model displayed both poor factor validity and poor external validity; therefore, research relying on the SAPS total score may reach misleading conclusions.