Equivalence and Bias in the South African Substance Use Contextual Risk Instrument

This article forms part of a larger study that sought to develop and validate a scale to measure individual and contextual factors associated with adolescent substance use in low-socio-economic status South African communities. The scale was developed to inform the process of designing preventative interventions in these communities. This study assessed the construct equivalence and item bias across different language versions of the scale. Exploratory factor analysis, equality of reliabilities, and the Tucker’s phi coefficient of congruence were employed to assess whether the two language versions were equivalent at a scale level. Differential item functioning analysis was conducted using ordinal logistic regression and the Mantel-Haenszel method at an item level. The findings revealed that there are significant differences between the two groups at a scale level. Items were flagged as presenting with moderate to large differential item functioning. The biased items have to be closely examined in order to decide how to address the bias.