Measurement invariance of a brief disordered eating scale in a large sample of secondary school students in Canada.

Background: For use in population-level surveys, there is a need for brief measures of disordered eating (DE) that demonstrate validity among diverse adolescents. This study aimed to (1) assess the measurement invariance of a short DE scale among high school students in Canada, and (2) estimate differences in scale scores across student sociodemographic subgroups.

Methods: We used student-level survey data collected during the 2021/2022 school year in the COMPASS study. The sample included 24,639 students in grades 9-12 from 69 secondary schools in Alberta, British Columbia, and Ontario, Canada. Multi-group confirmatory factor analysis was applied to assess measurement invariance (i.e., configural, metric, scalar, and strict invariance) of the 6-item DE scale, and one-way ANOVA and mixed linear regression were implemented to estimate scale score differences across population subgroups (by grade, gender, race and ethnicity, family affluence, and weight status).

Results: Results demonstrated full measurement invariance of the DE scale across student sociodemographic subgroups. Higher sum scale scores indicated greater engagement in DE thoughts or behaviours. Students in higher school grade (e.g., grade 12), transgender and gender diverse youth, and cisgender girls reported higher DE scores compared to their grade 9 and cisgender boy peers. Elevated DE scores were also observed among adolescents of Middle Eastern and Another or Multiethnic identity, students with less affluent households, and those with higher body weights, relative to their White, more affluent, and relatively lower weight peers, respectively.

Conclusions: Results indicate the DE scale measures the same construct across various subgroups of adolescents, providing confidence that differences in scale scores found by grade, gender, race and ethnicity, affluence, and weight reflect actual differences in DE rather than artifactual differences in scale interpretation. This scale warrants additional psychometric testing as a promising brief DE measure suitable for large population-level youth surveys. Our results provide important new evidence from a large contemporary sample of adolescents demonstrating elevated risk of DE among specific subgroups and emphasizing the need for targeted interventions.