Cross-language invariance and item-level differences in the SWEMWBS: a CFA and IRT study in English-, French-, and Spanish-speaking adolescents

Brief, psychometrically robust measures are essential for cross-language assessment and monitoring of adolescent mental wellbeing. Yet most large-scale evidence comes from WEIRD (Western, Educated, Industrialized, Rich, Developed) contexts, underscoring the need for validation in more diverse socio-cultural and linguistic settings. This also calls for brief, inclusive instruments evaluated with rigorous methods beyond classical test theory, such as multi-group CFA (MGCFA) and item response theory (IRT). We evaluated the Short Warwick-Edinburgh Mental Well-Being Scale (SWEMWBS) across English-, French-, and Spanish-language survey versions in 6,445 adolescents (English n = 5,369; French n = 680; Spanish n = 396). We examined internal consistency and convergent validity, tested a one-factor structure and multi-group measurement invariance using confirmatory factor analysis (CFA/MGCFA), and assessed item functioning using graded response item response theory models and differential item functioning (DIF) analyses. SWEMWBS showed good internal consistency (α/ω ≈ 0.80), expected convergent correlations with related wellbeing indicators, and support for a one-factor structure. The pooled one-factor CFA showed good fit (CFI = 0.981, TLI = 0.971, RMSEA = 0.049, SRMR = 0.019). MGCFA supported configural, metric, and scalar invariance across language groups. However, IRT-DIF analyses indicated item-level differences for six of seven items. These item-level results are interpreted as exploratory. In this international school-based sample, SWEMWBS appears to be a useful brief indicator of adolescent mental wellbeing across these language-version groups. However, the item-level non-invariance identified by IRT suggests that raw summed-score comparisons across language versions should be interpreted cautiously. Model-based approaches, such as latent-variable or IRT-informed scoring, may be preferable where precise cross-language comparison is required.