Is All Fat Created Equal?: Assessing Self-Reported Obesity Risk between Racial Groups with Structural Equation Modeling of the Eating Inventory

Background: Obesity prevalence rates differ widely between gender and racial groups in the United States. African American women have the highest prevalence rates, while Caucasian women have the lowest rates. Better assessment methods are needed to discern the varied and complex biopsychosocial risk factors for this disease for each individual. Psychological measures must be tested for cultural validity in the populations in which they will be used. One commonly used measure of eating behavior, the Eating Inventory (Stunkard & Messick, 1988), has not been assessed for cultural bias between the major ethnic groups in the Unites States in terms of its validity for assessment of obesity risk. Methods: Structural equation modeling techniques are used to compare invariance of factor structure of the Eating Inventory between adult Caucasian (n=110) and African American women (n=100). Alternate proposed factor structures are also tested for factor invariance between groups. Body Mass Index (BMI) is used as an optional covariate in the models. Additionally, socio-economic status indicators are examined for differential influence in the models, due to previously demonstrated, highly positive correlation with rates of obesity. Experimenter/participant race dyads are examined for systematic influence on response patterns. Results: The factor invariance of the original, three factor structure (Cognitive Restraint, Disinhibition and Hunger) of the Eating Inventory is supported with this sample. The invariance of alternate factor structures could not be supported with this sample, with one exception (Cognitive Restraint). Relative covariance of factors with BMI appears to be consistent between groups for all models tested. No participant/experimenter race interaction with factor scores was observed. Conclusions: The original three factor structure of the Eating Inventory appears to have invariance of the factor structure between the two groups measured, implying cultural validity. The relative covariance of factor scores with BMI is equivalent between groups. However, alternate factor structures could not be fully assessed due to sample size limitations. Further research is needed to replicate and expand these findings.