Lost in Translation? Cross-Cultural Differences in Symptom Report Across Instruments and Language of Administration in a Spanish-English Bilingual Cuban Sample

Abstract In an increasingly diverse world, assessors are likely to encounter examinees who are not proficient in the language of standard service delivery. For such occasions, some tests provide versions translated into multiple languages. Establishing the psychometric and cultural equivalence of scores on the original versus translated versions requires further empirical investigation. We report the results of a preliminary study on the effect of language-of-administration on three rapid-assessment instruments designed to measure emotional functioning in a small bilingual student sample. The PHQ-9, GAD-7 and the V-8 (visual analog scale) were administered both in English and Spanish to 47 Cuban university students, in counterbalanced order. The classification accuracy of the V-8 was calculated against the two legacy measures as criteria. There was no difference as a function of the language of administration. The Depression and the Anxiety scales of the V-8 were significant predictors of the PHQ-9 and GAD-7 during the English administration. On the Spanish version, the V-8 Depression was a stronger predictor of the PHQ-9; the V-8 Anxiety was a non-significant predictor of the GAD-7. Twice as many participants scored ≥ 10 on the PHQ-9 (31.9%) than the GAD-7 (17.0%) during the English administration; the proportion of the sample scoring ≥ 10 was similar on the Spanish version (29.8–23.4%). Although there was no difference across English versus Spanish versions, the classification accuracy of the Depression and Anxiety scales of the V-8 diverged. Results suggest that despite superficial equivalence, the clinical predictive power of a test may be different as a function of the language of administration. Future research on the psychometric properties of translated versions of tests is needed to determine if this is a widespread phenomenon or an isolated finding.