Measurement invariance of the Eating Pathology Symptoms Inventory (EPSI) in adolescents and adults

Highlights

• The Eating Pathology Symptoms Inventory (EPSI) structure replicated across ages.

• The EPSI scales demonstrated reliability in adolescents and adults.

• The EPSI can be used for eating-disorder symptom monitoring across the agespan.

• EPSI scale norms by age-group/sex for Recovery Record app users were provided.

Abstract

Adolescence is a common period for eating disorder (ED) onset. The availability of psychometrically sound measures of ED psychopathology enables clinicians to accurately assess symptoms and monitor treatment outcomes continuously from adolescence and adulthood. The purpose of this study was to assess if the Eating Pathology Symptoms Inventory (EPSI) is invariant across adolescents and adults. Participants (N = 29,821) were adolescent (n = 5250) and adult (n = 24,571) users of the Recovery Record (RR) mobile phone application who provided EPSI responses through the application. Measurement invariance testing was conducted to assess invariance of the EPSI Body Dissatisfaction, Restricting, Excessive Exercise, Purging, Cognitive Restraint, and Binge Eating scales across adolescents (age 13 through 17) and adults (age 18 and older). Findings indicated that all EPSI factors administered in the RR app replicated in both adolescent and adult users. The EPSI factor structure was largely equivalent in adolescents and adults, demonstrating evidence for configural and metric invariance, as well as some evidence for scalar invariance. Our results indicated that EPSI scales measured the same constructs across development. Clinicians and researchers may benefit from utilizing the EPSI to measure ED psychopathology in adolescents and for continued progress monitoring into adulthood.

Introduction

Eating disorders (EDs) are serious, impairing mental illnesses that often begin in adolescence (Nagl et al., 2016; Swanson et al., 2011). Because EDs are often chronic, the ability to accurately measure ED psychopathology in adolescents is critical to monitoring treatment response and assuring appropriately intensive treatment. Aspects of cognitive development may limit adolescents’ understanding of complex cognitive constructs related to EDs (Goldschmidt et al., 2011; Tanofsky-Kraff et al., 2008). Thus, it is important to evaluate how ED symptom measures developed for adults perform in adolescents. Although adolescent-specific measures of ED psychopathology exist, many of these measures are psychometrically problematic and limited to one developmental period. An advantage to a consistent measurement approach across the lifespan is that it may better characterize developmental changes in EDs in both research and clinical settings. Below, we review the psychometric properties of common self-report measures of ED symptoms that were developed for adults, but commonly used with adolescents.

The EDE-Q (Fairburn & Beglin, 1994) is the most widely used measure of ED symptoms in adults. A strength of the EDE-Q is its scales' correspondence with common ED treatment targets, such as overvaluation of weight and shape and dietary restraint (Fairburn et al., 2003). Several studies have evaluated the factor structure of the EDE-Q in adolescents or mixed adolescent-young adult samples (Becker et al., 2010; Penelo et al., 2013; White et al., 2014). Results across these samples suggest that the EDE-Q has an inconsistent factor structure across development. Becker et al. (2010) evaluated a culturally-adapted version of the EDE-Q in a large sample of young adult Fijian females, with results failing to replicate the original structure. Mantilla et al. (2017) conducted exploratory factor analyses on the EDE-Q; results supported a one-factor solution in adolescent girls and a three-factor solution in adolescent boys. Other studies in clinical-adolescent and mixed clinical-community adolescent samples yielded factor structures consisting of two (Penelo et al., 2013; Sepúlveda et al., 2019) or three (White et al., 2014) factors. Though this factoral inconsistency could be attributed to cultural differences (e.g., Becker et al.'s Fijian sample, Sepúlveda et al.'s Spanish sample) or gender differences between groups (Mantilla et al., 2017), these studies nonetheless raise concern about how the EDE-Q performs in adolescents. Further, the original EDE-Q structure generally fails to replicate in independent samples of adults; thus, psychometric concerns are not limited to adolescents (Berg et al., 2012).

The EDI-3 (Garner, 2004) measures scales related to ED psychopathology (i.e., Bulimia, Drive for Thinness) as well as more general constructs often associated with EDs, such as Emotional Dysregulation and Perfectionism. The EDI-3 demonstrated good internal consistency in adolescent patients (Cumella, 2006). However, a review by Gleaves et al. (2014) noted that certain EDI subscales yielded lower internal consistency in adolescents, with the Bulimia scale demonstrating the largest discrepancy (α = 0.72 in adolescents compared to 0.81 in adults).

A limitation of the EDI-3 may be factor-structure inconsistency in adolescents compared to adults. For example, the EDI-3 factor structure did not provide adequate model fit in a large nonclinical adolescent Spanish-speaking sample (García-Grau et al., 2010). Another study that examined the EDI-2, which shares many EDI-3 items, found that the factor structure did not replicate in adolescent girls (McCarthy et al., 2002).

The EPSI is a more recently developed measure that has demonstrated strong psychometric properties in adults (Forbush et al., 2013). Scales measure both behavioral (i.e., purging, binge eating, excessive exercising, restricting, muscle building) and cognitive features (i.e., cognitive restraint, body dissatisfaction, negative attitudes toward obesity) of EDs. The EPSI was developed and validated in several samples, including a mixed adolescent-adult ED sample, a college sample, and a general psychiatric outpatient sample (Forbush et al., 2013). EPSI scales demonstrated evidence for strong convergent validity, discriminant validity, criterion-related validity, and good test-retest reliability (Forbush et al., 2013; Forbush et al., 2014). The factor structure was also recently supported in a mixed adolescent-adult sample of individuals presenting for outpatient ED treatment (Coniglio & Becker et al., 2018). Although the EPSI has been tested in samples that included adolescents, no studies have tested the EPSI's factor structure or internal consistency in an exclusively adolescent group. Further, no studies have tested if the EPSI factor structure is invariant between adolescents and adults.

Several existing measures of ED psychopathology that were developed for adults and used to assess adolescents show evidence for inconsistent factor structures, which suggests that certain measures may not assess the same latent constructs across development. For example, inconsistent factor structures between groups make it difficult to meaningfully compare subscale scores across development. Though the EPSI has shown promising structural replicability, it is important not only that scale constructs replicate in different groups, but also that similar magnitudes of factor loadings and intercepts emerge. Though the EPSI has a reading level appropriate for teens, EPSI items are tailored primarily to adult ED presentations. For example, certain body dissatisfaction items focus on body-parts that may change throughout adolescence, so it is possible that such items perform differently in adolescents. Although a strength of the EPSI is its scales for various ED behaviors such as purging, certain ED behaviors may have a later age of onset (Stice et al., 1998), which could translate to structural differences and/or mean-level differences (which are not synonymous with structural differences). It is important to confirm that EPSI score differences reflect developmental differences in ED behaviors rather than how scales perform in adolescents. Thus, the primary aim of this study was to test the measurement invariance of the EPSI in a sample of adolescents and adults seeking treatment for an ED. Our secondary aim was to provide EPSI norms for both adolescent and adult patients with an ED. Based on past research on the EPSI factor structure in mixed adult-adolescent samples (Coniglio & Becker et al., 2018; Forbush et al., 2013; Forbush et al., 2014), we hypothesized that the EPSI's factor structure would demonstrate a good fit to the data in both adolescents and adults. In addition, we hypothesized that the EPSI would be invariant between adolescent and adult patients with an ED.