Are the Criterion B Binge-Eating Disorder Symptoms Interchangeable in Conveying the Latent Binge-Eating Trait? An Item Response Theory Analysis

Abstract

Binge-eating disorder (BED) is one of the most prevalent eating-disorder diagnoses among nationally representative samples of adults. Yet, few studies have evaluated the psychometrics of the BED diagnostic criteria. The BED Criterion B symptoms represent specific symptoms that may accompany binge-eating episodes. A minimum of three Criterion B symptoms must be endorsed to meet full BED diagnostic criteria. However, few studies have examined the Criterion B symptoms individually to clarify how well each symptom captures underlying BED pathology. The purpose of the present study was to use item response theory (IRT) to identify: 1) how much underlying binge-eating pathology must be present to endorse each symptom (i.e., item difficulty) and 2) how well each symptom differentiates among individuals with different levels of binge-eating pathology (i.e., item discrimination). Participants (N=254) were adults (85.0% female) with BED who completed eligibility screening for a BED group-treatment study. I hypothesized that ‘eating alone’ and ‘eating large amounts when not hungry’ would be the most ‘difficult’ items. Second, I hypothesized that ‘eating alone’ would demonstrate the greatest ability to discriminate those at the high end of the latent trait spectrum from relatively lower trait individuals. Results indicated that ‘uncomfortably full’ and ‘eating large amounts when not hungry’ had the highest difficulty parameters and were most difficult for high trait level individuals to endorse. ‘Eating rapidly’ and ‘eating alone’ were the most discriminating symptoms. Finally, the only symptom that provided unique item information at any trait level was ‘eating large amounts when not hungry.’ Results have implications for refining the Criterion B symptoms given the lack of unique item information relayed by the majority of the symptoms, the symptoms’ dimensionality, and variable monotonicity.