RESUMO
BACKGROUND: The aim of this study was to investigate the performance of the Toronto Alexithymia Scale (TAS-20) in patients with eating disorders. As a secondary objective, we analyzed a subset of items in order to determine whether the total score is a good tool for classifying these patients and whether the items are ordered according to the difficulty of choosing a particular answer depending on the severity of the patient's condition. SUBJECTS AND METHODS: We administered the TAS-20 to 103 consecutive patients who met DSM-IV-R criteria for anorexia nervosa, bulimia nervosa, or eating disorders not otherwise specified. Mokken scale analysis, nonparametric item response theory (NIRT), and confirmatory factor analysis (CFA) were used to test the psychometric properties of the scale. To improve our understanding of the structure underlying the TAS-20, we used an automated item selection procedure based on the Mokken scale. RESULTS: We identified a subset of 13 independent and clinically interpretable items that are potentially sufficient to rate patients with alexithymia. This subset was considered a unidimensional reduction of the TAS-20. However, the scale composed of this subset needs to be validated. Thirteen of the 20 items were maximally related to alexithymia because they have a one-dimensional structure. The NIRT and CFA enabled us to identify a maximal subset of items that conform to the requirements for good measurement of alexithymia in patients with eating disorders.