(PS7-63) Is the Clinical Impairment Assessment Invariant Across Weight Categories? Implications for Weight-related Impairment

The Clinical Impairment Assessment (CIA) assesses psychosocial functioning and psychiatric impairment secondary to an eating disorder (ED). EDs exist across various weight categories, and prior research has found that a higher body mass index (BMI) is associated with higher CIA impairment. Indeed, some CIA items may be interpreted differently based on weight status. For example, item one of the CIA asks, “To what extent have your eating habits, exercising, or feelings about your eating, shape, or weight made you feel ashamed of yourself?” It is possible that in addition to ED-related impairment, this item also captures internalized weight bias in higher-weight individuals. Thus, it is unclear whether we can interpret CIA scores the same way across weight categories.

Invariance testing (configural, metric, scalar, and residual) was conducted on the CIA’s three-factor structure between two weight categories (e.g., normal weight and higher weight [overweight, obesity type I, II, and III]). Participants (N= 4,225) for this study were students at a Midwestern University who were invited to complete an online self-report survey to better understand the rates of disordered eating behaviors in college students. Participants’ ages ranged from 18-75, with most respondents identifying as White, non-Hispanic women. We hypothesized that we would find at least partial invariance between the two weight groups.

The three-factor model had an acceptable fit (CFI=0.95, TLI=0.94, RMSEA=0.08). The configural invariance model fit was acceptable across groups (CFI=0.95, TLI=0.94, RMSEA=0.08). Full metric invariance between the groups was not supported; however, we found partial metric invariance by freeing six-item loadings. Most variant items pertained to social eating (e.g., finding it “difficult to eat out with others”). Full scalar invariance was also not supported, but we found partial scalar invariance by freeing six intercepts. Intercepts for the higher-weight group were higher than the lower-weight group. Full residual invariance was not supported, though we found partial residual invariance by freeing four residuals. Overall, only partial invariance of the CIA between the normal and higher-weight groups was found.

Our results indicated that CIA items, especially those assessing impairment in social eating situations, are variant across BMI categories. This may be due to societal messages that individuals with a higher BMI “should” eat more healthily than their lower-weight peers. Metric invariance indicated that some items within the CIA function differently between weight groups. Partial scalar invariance may indicate that not all items on the CIA can be compared without bias across weight groups (e.g., utilizing a subscale score when scoring the CIA). Because intercepts for the higher-weight group were higher, it is reasonable to assume that some CIA items may also account for additional, probable constructs, such as weight discrimination or bias. Based on our findings, future research should consider removing variant items from the CIA and reevaluating its factor structure. Lastly, clinicians scoring the CIA should be aware of elevated scores amongst the variant items when utilizing and interpreting the CIA.