Considering Measurement Equivalence in the Assessment of Quality of Life in Immigrants and the Autochthonous Population

Thursday, 19 July 2018: 09:30
Oral Presentation
Patrick BRZOSKA, Chemnitz University of Technology, Germany
Introduction: Health-related quality of life (HRQOL) is an important outcome in health research and a frequently studied social indicator. The SF-36 inventory is one of the most often used instruments for assessing this construct. When HRQOL dimensions are compared between immigrants and non-immigrants, often composite scores are used, calculated as the sum of the underlying item responses. This, however, neglects that groups may differ in their item responses despite having the same position on the latent dimension these items are supposed to measure. This study examines this differential item functioning (DIF) for two frequently used subscales—vitality (VT) and mental health (MH)—of the SF-36 in immigrants and non-immigrants in Germany.

Methods: Data from a representative population-based survey (n=22,050) is used. The measurement baseline model tested by means of confirmatory factor analysis comprised the VT and the MH SF-36 factor. DIF related to immigrant background was analyzed by means of multiple indicators multiple causes models. Additionally, it was examined whether DIF related to immigrant background is moderated by sex, age and socioeconomic status (SES).

Results: 11.5% of the respondents were immigrants. DIF related to immigrant background was observed in 6 out of 9 items. It was both confounded and moderated by sex, age and SES in 4 items. Effect sizes were moderate in size. A comparison of composite scores and latent means revealed significant differences.

Discussion: The comparison of item composite scores between groups may be biased as a result of DIF, leading to under- or overestimation of true group differences as well as of the role of relevant determinants. To obtain valid estimates, appropriate approaches such as latent variable modeling need to be applied. A mixed-method approach can help to identify causes of DIF and can guide appropriate measures aiming to address non-invariance in different phases of the research process.