- Top of page
Self-report measures are commonly used to assess eating disorder (ED) symptoms in clinical and research settings.[1-4] Many ED self-report measures have undeniable strengths, including excellent internal consistency and test-retest reliability, criterion-related validity for distinguishing among patient groups,[5, 6] and good-to-excellent convergent validity with other measures of disordered eating.[7-10] Nevertheless, many ED self-report measures were developed on the basis of small samples comprised exclusively of young adult women, and are associated with other psychometric issues that may limit their utility, such as poor factor structure replicability, limited discriminant validity, and reduced psychometric properties in specific populations, such as men and overweight and obese individuals [for additional review of these limitations, see Ref. . In response to the limitations of certain ED self-report measures, our research group developed the Eating Pathology Symptoms Inventory (EPSI), which we describe briefly, below.
The EPSI was developed according to the rigorous scale development procedures recommended by Clark and Watson. Based on previous theoretical and empirical models of EDs, the initial measure included 160 items designed to assess 20 potential dimensions of eating pathology, and included items to assess all of the Diagnostic and Statistical Manual of Mental Disorders, fourth edition [DSM-IV] criteria for EDs. The initial item-pool was administered to large independent samples of community participants and college students. Exploratory, confirmatory, and multiple-group factor analyses (which test for factor invariance between groups) were used to refine the initial measure and test whether the structure was invariant between men (vs.) women and normal weight (vs.) overweight/obese participants. A revised 88-item measure was then administered to large independent samples of ED patients and general psychiatric outpatients. Based on the results of additional multivariate statistical analyses, the measure was revised a second time, and the finalized 45-item measure was administered to an independent sample of college students to confirm the structure of the measure and determine test-retest reliability.
The final 45-item version of the EPSI contained eight (factor analytically derived) scales that were highly replicable across all samples, and reflected dimensions with clear conceptual relevance to ED psychopathology. The first scale, “Body Dissatisfaction” (dissatisfaction with one's weight and/or shape), represented a large general factor that emerged in all samples and explained a substantial amount of variance in the EPSI. Body Dissatisfaction appeared to represent the “core” psychopathology of EDs, as indicated by significant factor loadings of other EPSI scales on this higher-order factor. The other scales were smaller and defined various aspects of eating pathology, including: “Binge Eating” (ingestion of large amounts of food and accompanying cognitive symptoms), “Cognitive Restraint” (cognitive efforts to limit or avoid eating, whether or not such attempts are successful), “Purging” (self-induced vomiting, laxative use, diuretic use, and diet pill use), “Excessive Exercise” (physical exercise that is intense and/or compulsive), “Restricting” (concrete efforts to avoid or reduce food consumption), “Muscle Building” (desire for increased muscularity and muscle building supplement use), and “Negative Attitudes toward Obesity” (negative attitudes toward individuals who are overweight or obese). Our results indicated that the EPSI was able to capture the majority of variance associated with established ED measures, yet possessed additional content that was not represented in any existing multidimensional ED measure (e.g., Muscle Building and Negative Attitudes toward Obesity). EPSI scales were internally consistent, reliable over 2- to 4- weeks, and had excellent discriminant, convergent, and criterion-related validity. For example, the EPSI showed strong evidence of criterion validity for distinguishing among community and college samples (vs.) patients with EDs, and between general psychiatric patients (vs.) patients with EDs. Moreover, results indicated that the EPSI outperformed other multi-dimensional measures of eating pathology, such as the Eating Disorders Inventory-36 (EDI-3) and Eating Disorders Examination-Questionnaire, in terms of distinguishing between patients with anorexia nervosa (vs.) bulimia nervosa. The EPSI, therefore, represents a promising new measure of ED psychopathology and is appropriate for use among a wide range of participant groups (e.g., men, women, normal weight, overweight, obese, community-recruited, general psychiatric patients, and patients with EDs) and in both clinical and research settings.
Here, we report the first large-scale normative data for the EPSI in a sample of young adult men and women. The current study also extends prior research on the construct validity of the EPSI by testing convergent and discriminant validity in a group of participants that is independent from the initial scale development samples. Given that the peak age of onset for many EDs is in young adulthood[14, 15] and the need for brief, cost-efficient, and psychometrically sound assessments of disordered eating in this population, normative and validity data will be particularly useful.
Based on our previous findings, we hypothesized that the EPSI would show strong evidence of construct validity. Specifically, we hypothesized that: (a) Body Dissatisfaction would represent a higher-order factor in both college men and women and would demonstrate the highest convergent correlations with other measures of shape and weight concerns, (b) other EPSI scales would demonstrate small-to-moderate correlations with eating pathology measures, given that the EPSI scales were designed to possess distinct, non-overlapping content (i.e., EPSI scales that were not designed to assess body dissatisfaction should demonstrate lower correlations with shape and weight measures compared to EPSI Body Dissatisfaction, which directly assesses shape and weight concerns), and (c) EPSI scales would demonstrate strong discriminant validity (as indicated by smaller correlations with non-ED measures [vs. ED-related measures]). Finally, consistent with our previous research, we hypothesized that men would score significantly higher than women on Muscle Building, Excessive Exercise, and Negative Attitudes toward Obesity, whereas women would score higher on other EPSI scales, given that women appear to be at greater risk for traditionally defined eating pathology.[16-18]
- Top of page
As predicted, the EPSI's eight-factor model had a good fit to the data in both men (χ2 = 2,157.53 , p < .001, CFI = .947, TLI = .942, RMSEA = .052) and women (χ2 = 2,562.85 , p < .001, CFI = .953, TLI = .949, RMSEA = .054). Given that previous research indicates that EPSI Muscle Building has less relevance to women, we omitted this scale and re-ran the CFA in women, which resulted in an excellent fit (χ2 = 2,132.32 , p <.001, CFI = .958, TLI = .954, RMSEA = .056). The majority of the EPSI scales were internally consistent. Median coefficient alpha values were 0.87 in college men, 0.87 in college women, 0.86 in the combined student sample, and 0.89 in patients with EDs. However, Purging and Muscle Building were less internally consistent in college women (see Table 1).
Table 1. Descriptive statistics for the Eating Pathology Symptoms Inventory (EPSI), Eating Disorder Examination Questionnaire Weight Concern Scale (EDE-Q WC), Weight Concerns Scale (WCS), and SCOFF in college students and eating disorder patients
|Scale||College Students||Eating Disorder Patientsd||Effect Sizes|
|Mena||Womenb||Combined Student Samplec|
| Body dissatisfaction||13.40a||5.73||.90||20.68b||6.89||.91||17.44c||7.35||.92||27.79d||7.04||.89||−1.14||−1.42|
| Binge eating||18.17a||5.34||.85||19.21b||5.45||.87||18.74c||5.43||.86||17.05d||9.01||.93||−0.19||0.28|
| Cognitive restraint||6.70a||2.80||.80||8.39b||2.86||.80||7.64c||2.96||.82||12.60d||3.24||.89||−0.60||−1.66|
| Negative attitudes toward obesity||15.18a||4.69||.88||14.48b||4.65||.90||14.79c||4.68||.89||15.58c||6.85||.95||0.15||−0.16|
| Muscle building||9.75a||4.12||.77||6.98b||2.05||.54||8.21c||3.43||.75||6.95d||2.28||.56||0.88||0.38|
| EDE-Q WCg||.75a||1.12||.85||2.11b||1.62||.89||1.53c||1.58||.90||5.61d||1.39||.80||−0.95||−2.65|
Independent t-tests indicated that college women had significantly higher EPSI scores than college men on Body Dissatisfaction, Binge Eating, Cognitive Restraint, Purging, and Restricting. Consistent with our hypotheses, college men had higher scores than college women on Excessive Exercise, Muscle Building, and Negative Attitudes toward Obesity (Table 1). Additional independent t-tests in the ED patient sample indicated that men (n = 9) and women (n = 140) with EDs did not differ significantly on EPSI scores, except that men with EDs had significantly higher scores on Muscle Building than women with EDs. Cohen's d values for men (vs.) women with EDs were large for EPSI Muscle Building, medium for EPSI Body Dissatisfaction, and small for all other EPSI scale comparisons (data available upon request from first author).
Table 2 shows convergent and discriminant correlations for the EPSI scales with the EDE-Q WC scale, WCS, SCOFF, and TAM. As predicted, Body Dissatisfaction demonstrated strong, positive correlations with the SCOFF, EDE-Q WC, and WCS; Cognitive Restraint had a strong positive correlation with the WCS in college women. The SCOFF, EDE-Q WC, and WCS had small-to-moderate positive correlations with other EPSI scales. The overall pattern of convergent correlations was similar between college men and women; yet, the magnitude of convergent correlations was stronger in college women. Discriminant correlations between EPSI scales and the TAM were lower than correlations between EPSI scales and other measures of eating pathology, demonstrating evidence for discriminant validity (see Table 2).
Table 2. Convergent and discriminant validity for the Eating Pathology Symptoms Inventory (EPSI) scales in men and women
| ||EPSI Scale||Body Dissatisfaction||Binge Eating||Cognitive Restraint||Purging||Restricting||Excessive Exercise||Negative Attitudes Toward Obesity||Muscle Building|
EPSI scale scores generally were significantly different between the ED patient and college samples (see Table 1). Consistent with the composition of the ED sample, which was comprised of a substantial portion of individuals with anorexia nervosa, college students had significantly higher scores on Binge Eating and Muscle Building compared to patients with EDs (additional comparisons indicated Binge Eating was significantly higher in individuals with bulimia nervosa compared to college students; data available from first author). No significant differences were found between college students and patients with EDs on Negative Attitudes toward Obesity. For other EPSI scales, patients with EDs had significantly higher scores than the college students. (Note: Given that the ED sample was significantly older and had a greater proportion of female participants than the college sample, we re-ran between-group comparisons for EPSI scales in an age- and gender-matched sub-sample of college students. The results of these analyses replicated the pattern of findings in the larger sample of college students).
- Top of page
The goal of this study was to provide normative and validity data for the EPSI in a sample of young adults. Consistent with predictions, our results indicated that the EPSI had a robust eight-factor structure in both sexes, which replicated the findings of Forbush et al. in a sample that was independent from those used to develop the measure. As hypothesized, between-group comparisons indicated that college women scored significantly higher than college men on most EPSI scales, except Excessive Exercise, Muscle Building, and Negative Attitudes toward Obesity, which were significantly higher in men. Our findings are consistent with published research indicating that men exhibit a greater desire for larger, more muscular bodies,[34, 35] and may use various behavior strategies (e.g., excessive protein supplement use and excessive exercise to increase lean muscle mass) to achieve these sex-specific aesthetic ideals.[20, 21] Notably, our results indicate that the EPSI may be more sensitive to certain “male-specific” aspects of eating pathology compared to other multidimensional ED measures. With the exception of Negative Attitudes toward Obesity, between-group comparisons showed that EPSI scale scores significantly distinguished college-recruited young adults from individuals with EDs. The results of our analyses contribute to prior research on the validity of the EPSI by demonstrating that EPSI scales possess excellent criterion validity for distinguishing college-recruited young adults from individuals with EDs.
Analyses of convergent and discriminant validity revealed three important properties of the EPSI. First, the Body Dissatisfaction scale correlates most strongly with other measures of weight and shape concerns, showing clear convergent validity and supporting conceptualizations of overvaluation of weight and shape as the core psychopathology of EDs. Second, EPSI scales that were not characterized by body image concerns were significantly (albeit less strongly) correlated with other measures of weight and shape concerns. These results demonstrate that although EPSI scales are significantly related to other eating pathology measures (demonstrating evidence for convergent validity), they are also able to efficiently assess distinct dimensions of eating pathology (demonstrating evidence for discriminant validity). These results are important because many established ED measures differentiate scale content in a way that reduces discriminant validity. For example, the EDI-3 Bulimia scale contains content that assesses purging, inappropriate compensatory behavior, and binge eating. To the extent that the behaviors assessed in the Bulimia scale are distinct, the inclusion of these dimensions within a single scale would diminish discriminant validity. Conversely, certain EDE-Q scales may artificially discriminate between behaviors that potentially should be grouped together (e.g., Shape Concern and Weight Concern). In contrast, the EPSI assesses distinct ED dimensions that have only modest intercorrelations. The strong discriminant validity of the EPSI may be a particularly useful feature for future behavior genetic and clinical trials research, in which it is important to identify phenotypes that are maximally homogeneous to better characterize genetic and environmental risk factors for disordered eating or to track symptom change more precisely over time. Finally, all scales had stronger correlations with other measures of ED symptoms than with negative and positive trait affect, providing excellent evidence for discriminant validity.
The initial item pool contained questions to comprehensively assess theoretically important dimensions of eating pathology, as well as all of the DSM-IV criteria for EDs. Although the DSM-5 recently was published, changes to the ED diagnostic criteria are fairly minimal and focus primarily on the frequency and duration of specific symptoms (e.g., binge eating, purging), which would not affect the content or structure of the EPSI scales. Thus, the results of the current study (as well as findings from Forbush et al.) have important taxonomic implications for the organization of ED symptoms. First, results from confirmatory factor analyses support a higher-order eight-factor structure, in which Body Dissatisfaction exists at a superordinate level to the other seven EPSI factors. These results support theoretical models in which body image concerns represent the ‘core’ psychopathology shared across eating disorders. Second, however, results also indicate that EPSI scales are clearly distinguishable from one another and that there is important additional variance beyond a one-factor model (e.g., a one-factor model in which all EPSI items were indicators of a latent Body Dissatisfaction factor had a very poor fit to the data [in both our previous work and the present study]). Finally, results show that out of the eight EPSI scales, Body Dissatisfaction and Binge Eating demonstrate the strongest correlations with measures of depression, anxiety, and negative affect. These results are informative in providing information about which aspects of eating pathology may contribute to the high rates of comorbidity observed among mood, anxiety, and eating disorders. Together, these results indicate that EDs share a common underlying psychopathology, but that additional dimensions of EDs can be readily identified, and that certain ED symptom dimensions are correlated with other forms of internalizing psychopathology.
Although current results provide encouraging evidence in support of the validity of the EPSI, it will be important to replicate these findings in additional samples. In particular, the current sample had low representation of individuals in ethnic and racial minorities, and no studies have examined norms or validity of the EPSI in young adolescent samples. Future normative studies of the EPSI also would benefit from further characterization of threshold-level eating pathology within normative samples, which was lacking in the current study. Finally, Negative Attitudes toward Obesity was not significantly related to other ED measures in men, which indicates that this scale may have limited convergent validity in men.
In conclusion, the EPSI demonstrated good-to-excellent psychometric properties that are comparable to other well-established multidimensional ED measures. EPSI scales showed clear evidence for factor structure replicability in men and women—a feature that is not characteristic of many other self-report measures of eating pathology—and excellent convergent, discriminant, and criterion validity. These normative data can be used to guide future research and clinical work by providing a context for interpreting young adults' scores on the EPSI. Taken together, results provide the first large-scale normative data for the EPSI in young adults, as well as additional evidence supporting its psychometric properties and construct validity.