Examining Dimensionality and Item-Quality of the Eating Disorder Examination Questionnaire in Individuals With Eating Disorders Using Item Response Theory Analysis.

Abstract

Objective: The Eating Disorder Examination Questionnaire (EDE-Q) is a widely-used measure of eating-disorder symptoms. However, inconsistent replication of the subscale structure raises concern about validity. To provide a rigorous test of the EDE-Q's dimensionality and item-quality, we applied modern and classical test theory approaches to data obtained from a large, transdiagnostic sample of people with clinical eating disorders.

Method: We analyzed data from 1197 individuals (M_age = 27.9 years, SD = 10.08, 95% female) with various eating disorders, who had been assessed for treatment at a specialized program. Exploratory analyses (including Parallel Analyses), Confirmatory Factor Analyses (CFA) and graded-response Item Response Theory (IRT) analyses, were conducted with Mplus.

Results: Factor analyses showed inappropriate fit to the original EDE-Q subscales, as well as for alternative 1,2,3, and 4-factor solutions. Parallel analyses suggested a one-dimensional structure as best fit. IRT analyses showed substantial variability in EDE-Q-item quality and indicated that five items (fear of weight gain, feeling fat, desire to lose weight, importance of weight, importance of shape) were most pertinent to determining severity. The construct validity of the five EDE-Q items was confirmed by a CFA, showing excellent fit.

Discussion: Our results suggest that EDE-Q scores are best interpreted as spanning a one-factor continuum. IRT results suggest that some items are more pertinent than others for determining eating-disorder severity. Results could be useful for establishing short EDE-Q versions, such as a five-item version, which, in turn, would be helpful for measurement-based clinical practice and for data-collection in epidemiological and experimental studies.