Investigating item bias in some of the scales of the South African substance use contextual risk instrument (sasucri) for use with adolescents in low socio-economic status communities in the Western Cape.

Abstract

Substance use is a continuously growing problem worldwide. South Africa in particular has a high prevalence of substance use disorders, most of which are affecting adolescents and young adults. Over the past few years the Western Cape region has emerged with the highest rates of substance use. Literature pointed toward the need for an instrument which would be able to validly measure the contextual factors associated with adolescent substance use in low socio-economic status communities in the Western Cape. The South African Substance Use Contextual Risk Instrument (SASUCRI) was therefore developed to capture these contextual realities from the subjective point of view of adolescents. The instrument serves as a means to identifying at-risk youth and communities in which risk factors are present. Exploratory research on the reliability and validity of the instrument was previously conducted. This initial validation study concluded that further revision was required before the instrument could validly measure what it intended to measure. It was recommended that a DIF analysis be done on six of scales on the measure which proved to not be equivalent across the two language versions. The current study is based on this recommendation and serves as a continuation of the validation process for this instrument. This study contributes to the validation process for the instrument by investigating item bias on six inequivalent scales. This study aimed to: 1) establish whether items function differently across the two language versions, 2) identify problematic items on the six scales and 3) inform the process of rewriting these items. Item bias was investigated using two well supported statistical procedures; Ordinal logistic regression and the Mantel-Haenszel method. The analysis was conducted on secondary data which was collected by researchers of the larger study. Permission was requested from the researchers of the larger study for use of this data. The main findings of the study indicated that there were discrepancies between the results obtained across the two methods.