Differential item functioning for items in Berger’s HIV Stigma Scale: an analysis of cohorts from the Indian, Swedish, and US contexts

PURPOSE: To examine whether items in Berger’s HIV Stigma Scale function differently with persons of different age, gender, and cultural backgrounds. METHODS: Secondary data from cohorts, collected in South India (n = 250), Sweden (n = 193), and the US (n = 603) were reanalyzed to evaluate DIF within, between, and across these cohorts. All participants had answered the revised version of the HIV stigma scale consisting of 32 items forming the subscales Personalized stigma, Disclosure concerns, Concerns about public attitudes, and Negative self-image. Differential Item Functioning (DIF) for these items was assessed using hybrid ordinal regression-IRT technique. When DIF was detected, the cumulative impact of DIF on individual subscale scores was evaluated. RESULTS: DIF was detected for 9 items within, between, or across cohorts, but the DIF was negligible in general. Detected DIF between the Swedish and Indian cohorts had a cumulative salient impact on individual scores for the subscale Disclosure Concerns; Disclosure concerns were overestimated in the Swedish cohort and both over- and underestimated in the Indian cohort. CONCLUSIONS: The items in the 32-item version of the HIV stigma scale did not seem to be particularly prone to present DIF. The DIF between the Indian and Swedish cohort for items in the subscale Disclosure Concerns could, however, result in both type I and type II errors if scores should be compared between the Indian and Swedish cohort. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (10.1007/s11136-018-1841-4) contains supplementary material, which is available to authorized users.