Investigating ethnic variations in reporting of psychotic symptoms: a multiple-group confirmatory factor analysis of the Psychosis Screening Questionnaire.

Abstract

Epidemiological evidence suggests risk for psychosis varies with ethnicity in Western countries. However, there is little evidence to date on the cross-cultural validity of screening instruments used for such comparisons. Combining two existing UK population-based cohorts, we examined risk for reporting psychotic symptoms across White British (n = 3467), White Irish (n = 851), Caribbean (n = 1899), Indian (n = 2590), Pakistani (n = 1956) and Bangladeshi groups (n = 1248). We assessed the psychometric properties of the Psychosis Screening Questionnaire (PSQ) with a multiple-group confirmatory factor analysis, assessing the equivalence of factor loadings, response thresholds and residual variances in an analysis of measurement non-invariance. Compared with prevalence among British Whites (5.4%), the prevalence of self-reported psychotic symptoms was greater in the Caribbean group (12.7%, adjusted OR = 2.38 [95% CI 1.84-3.07]). Prevalence was also increased among Pakistani individuals (8.3%, adjusted OR = 1.36 [1.01-1.84]) although this difference was driven by a greater likelihood of reporting paranoid symptoms. PSQ items for thought interference, strange experience and hallucination were measured in equivalent ways across ethnic groups. However, our measurement models suggested that paranoid symptoms were measured less reliably among ethnic minorities than among British Whites and appeared to exaggerate latent differences between Pakistani and White British groups when measurement non-invariance was not accounted for. Notwithstanding evidence for measurement non-invariance, the greater risk for reporting psychotic symptoms among Caribbean individuals is unlikely to be an artefact of measurement. Greater residual variance in the recording of paranoid symptoms among ethnic minority respondents warrants caution in using this item to investigate ethnic variation in psychosis risk.