Generalized network psychometrics of eating‐disorder psychopathology. Issue 11 (29th August 2022)

• Record Type: Journal Article
• Authors: Forbush, Kelsie T.; Swanson, Trevor J.; Chen, Yiyang; Siew, Cynthia S. Q.; Hagan, Kelsey E.; Chapa, Danielle A. N.; Tregarthen, Jenna; Wildes, Jennifer E.; Christensen, Kara A.
• Is Part Of: International journal of eating disorders. Volume 55:Issue 11(2022)
• Page Start: 1603
• Page End: 1613
• Publication Date: 2022-08-29
• Subjects: classification -- diagnosis -- eating disorders -- Eating Pathology Symptoms Inventory -- generalized network psychometrics -- latent variable modeling -- network analysis -- psychonetrics; Appetite disorders -- Periodicals; Ingestion disorders -- Periodicals; Eating disorders -- Periodicals; 616.8526
• DOI: 10.1002/eat.23801
• Languages: English
• ISSNs: 0276-3478

Abstract

Abstract: Objective: As network models of eating disorder (ED) psychopathology become increasingly popular in modeling symptom interconnectedness and identifying potential treatment targets, it is necessary to contextualize their performance against other methods of modeling ED psychopathology and to evaluate potential ways to optimize and capitalize on their use. To accomplish these goals, we used generalized network psychometrics to estimate and compare latent variable models and network models, as well as hybrid models. Method: We tested the structure of the Eating Pathology Symptoms Inventory (EPSI) and Eating Disorder Examination‐Questionnaire (EDE‐Q) in Recovery Record, Inc. mobile phone application users ( N = 6856). Results: Although all models fit well, results favored a hybrid latent variable and network framework, which showed that ED symptoms fit best when modeled as higher‐order constructs, rather than direct symptom‐to‐symptom connections, and when the relationships between those constructs are described as a network. Hybrid models in which latent factors were modeled as nodes within a network showed that EPSI Purging, Binge Eating, Cognitive Restraint, Body Dissatisfaction, and Excessive Exercise had high importance in the network. EDE‐Q Eating Concern and Shape Concern were also important nodes. Results showed that the EPSI network was highly stable and replicable, whereas the EDE‐Q network was not. Discussion: Integrating latent variable and network model frameworks enables tests of centrality to identify important latent variables, such as purging, that may promote the spread of ED psychopathology throughout a network, allowing for the identification of future treatment targets.