Critically Revisiting the Psychometric Properties of the EDE-Q and Its Short Forms: Is It Time to Move On?

Abstract

Background: The Eating Disorder Examination Questionnaire (EDE-Q) remains the most widely used self-report measure of eating disorder symptoms, despite concerns regarding its factorial instability, construct redundancy, and local item dependence. Numerous brief versions have been proposed to address these limitations. Yet, whether these revisions represent true psychometric advances remains questionable.

Methods: In the present study, we critically examined the psychometric properties of the full EDE-Q and its short forms in a sample of 772 Australian women. Dimensionality was examined using bootstrap Exploratory Graph Analysis (EGA), followed by Confirmatory Factor Analysis. Factor redundancy was assessed through correlations between factors, HTMT2, and differential associations with an external criterion. Item-level analyses evaluated the functioning of putative doublets, and receiver operating characteristic analyses examined classification accuracy and overlap across scale versions.

Results: EGA yielded more parsimonious structures than the originally proposed models, often collapsing highly overlapping factors. Analyses of factor redundancy indicated limited discriminant validity between several factors. Item-level analyses indicated that putative doublets had highly similar associations with the external criterion and comparable discriminative performance. Finally, global scores were highly correlated across versions and demonstrated similar moderate classification accuracy, although versions based exclusively on shape/weight concern performed worse.

Conclusions: Although brief forms reduced length, they did not meaningfully resolve the core structural weaknesses of the EDE-Q. These findings suggest that these modifications of the EDE-Q may perpetuate, rather than solve, its psychometric limitations. We therefore emphasise the need for the development of new measures and offer recommendations grounded in contemporary psychometric best practices.