Cross-group and longitudinal measurement invariance of the Child Eating Behavior Questionnaire-24G in middle childhood

BACKGROUND: Childhood eating behaviors are most commonly assessed using parent-report questionnaires. To validly compare means across different groups or over time, equivalent item functioning is an important prerequisite. This study investigates the measurement invariance of a short German version of the Child Eating Behavior Questionnaire, the CEBQ-24G, across child sex, age, weight status, family socio-economic status, and parental weight status, as well as across three measurement occasions spanning three years in middle childhood. METHODS: Analyses were conducted using a large community sample (N = 1,657) recruited from primary schools in Brandenburg, Germany. Measurement invariance was tested and latent group means were compared using Multigroup Confirmatory Factor Analysis (CFA). RESULTS: The CEBQ-24G is measurement invariant on the residual level across all groups and measurement occasions, i.e., assessed differences in eating behaviors between these groups and measurement occasions are derived from actual differences in latent constructs and not a result of measurement bias-thus allowing valid comparisons of latent means across child sex, age, weight status, parental weight status, and family socio-economic status. We found significant differences in all group comparisons (except between age groups), although the eating behavior scales in which the differences in mean scores occurred varied depending on which groups were compared. CONCLUSIONS: Our findings suggest that the CEBQ-24G has good psychometric properties in all groups and is suitable for research into eating behaviors in middle childhood in diverse populations.