Assessing Discriminant Validity through Structural Equation Modeling: The Case of Eating Compulsivity.

Abstract

Food addiction (FA) and disordered eating behaviors related to obesity are gaining attention in clinical and research fields. The modified Yale Food Addiction Scale 2.0 (mYFAS2.0) is the gold standard questionnaire to measure FA, while another tool is the Measure of Eating Compulsivity 10 (MEC10). Discriminant validity is present when two measures of similar but distinct constructs show a correlation that is low enough for the factors to be regarded as distinct. However, the discriminant validity of these measures has never been tested. Through a cross-sectional study design, 717 inpatients (females: 56.20%, age: 53.681 ± 12.74) with severe obesity completed the MEC10, Binge Eating Scale (BES), and mYFAS2.0. A structural equation model (SEM) was fitted, freely estimating latent correlations with 95% confidence intervals (95% CI). The results confirmed the scales' excellent psychometric properties. Importantly, latent factor correlations between MEC10 and mYFAS2.0 (est = 0.783, 95% CI [0.76, 0.80]) supported their discriminant validity. In contrast, the latent correlation of MEC10 and BES (est = 0.86, 95% CI [0.84, 0.87]) exceeded the recommended thresholds, indicating the absence of discriminant validity and suggesting a potential overlap, consistent with previous evidence. In conclusion, MEC10 demonstrates excellent psychometric properties but is more a measure of BED and not FA.