Application of structural equation mixture modeling to characterize the latent structure of eating pathology.

Abstract

Several theoretical models describe the structure of eating disorders (EDs), and a burgeoning empirical literature has sought to identify whether eating pathology is conceptualized best as categorical (presence or absence of disorder), dimensional (continuous), or a hybrid of categories and dimensions. This study used structural equation mixture modeling (SEMM) to identify the latent structure of EDs. Items from the Eating Pathology Symptoms Inventory (EPSI) were administered to individuals with EDs (N = 344). Select EPSI scales and body mass index were indicators in subsequent SEMM analyses. The Inventory of Depression and Anxiety Symptoms (IDAS), ED diagnoses, and select demographic variables were used as validators using chi-square or MANOVA. Categorical models fit the data better than latent dimensional or hybrid models. Latent profile 1 (LP1) was non-fat-phobic restricting anorexia nervosa; LP2, an obese, binge-eating class; LP3, non-purging bulimia nervosa; LP4, fat-phobic restricting anorexia nervosa; and LP5, multiple purging bulimia nervosa. External validation analyses indicated that LP4 and LP5 had the highest non-ED-related psychopathology. These findings indicate that there is substantial variability in the phenomenology of traditional DSM-based ED categories across latent profiles, and highlight the salience of certain ED phenotypes that have been debated in the literature. © 2016 Wiley Periodicals, Inc.(Int J Eat Disord 2017; 50:542-550).