Exploring differential item functioning on eating disorder measures by food security status.

Abstract

Food insecurity is associated with elevated eating disorder (ED) pathology, yet commonly used ED measures may not fully capture ED pathology in the context of food insecurity. The present study used differential item functioning (DIF) analyses to explore whether item endorsement on two commonly used ED questionnaires differed by food security status. Participants were 634 cisgender women recruited through Amazon's Mechanical Turk. DIF was explored for items on the Short Eating Disorder Examination Questionnaire (S-EDE-Q) and the Eating Disorder Diagnostic Scale for DSM-5 (EDDS-5). DIF analyses used a hybrid ordinal logistic regression/item response theory approach, with the presence of both statistical (p < .01) and clinical significance (pseudo ΔR2 ≥ .035) indicating DIF. There was no evidence of clinically significant DIF within the S-EDE-Q. Two items on the EDDS-5 exhibited statistically and clinically significant DIF, with moderate effect sizes. Specifically, compared to food-secure participants, food-insecure participants were more likely to report (1) eating large amounts of food when not physically hungry and (2) feeling disgusted, depressed, or guilty about overeating at lower levels of overall ED pathology but less likely to report these experiences at higher levels of overall ED pathology. Findings highlight a potential need to adapt ED measures to fully capture ED pathology in food-insecure populations. Level III, well-designed cohort study.