Measurement invariance of the affective reactivity index across five countries.
Abstract
Objective
Irritability is a globally relevant dimension of youth psychopathology, yet cross-national comparisons require evidence that assessment tools function equivalently across countries and languages. We examined the reliability, factor structure, and measurement invariance of the Affective Reactivity Index (ARI) across samples from Brazil, China, Japan, Spain, and the United States (US).
Method
Community samples (N = 6,195; ages 6-18 years) completed parent- or youth-report ARI versions. Single-factor confirmatory factor analyses evaluated model fit and internal consistency. Multi-group CFA tested measurement invariance. Differences in observed total scores between samples and pooled age and sex associations were examined using linear regression.
Results
Internal consistency was acceptable to good across country samples and informants. A single-factor structure was broadly supported. For parent report, measurement invariance was well supported across China and Japan using the original three-category format and also supported across China, Japan, and the US after dichotomization of all items, due to lack of endorsement of the highest category of Item 4 in the US sample. For youth report, threshold and loading invariance were supported, but scalar invariance was not. Parent-reported irritability was higher in the Chinese and Japanese samples than in the U.S. sample, whereas youth-reported irritability was highest in the Brazilian sample. Parents reported higher irritability in boys and younger youth; youth self-report showed no significant age or sex effects.
Conclusion
The ARI showed broadly comparable measurement properties across the participating samples, with stronger evidence for parent report than youth report. However, parent-report comparisons involving the US were based on a dichotomized response format, and full intercept invariance was not established for youth report. Observed score differences should therefore be interpreted cautiously, and as differences between participating samples rather than population-level cross-national differences.