Differential item functioning of the CESDR-R and GAD-7 in African and white working adults

African and white working adults. SA Journal of Industrial SA Tydskrif vir Bedryfsielkunde Differential item functioning of the CESDR-R and GAD-7 in African and white working adults Authors: Carolina Henn 0 Brandon Morgan 0 Corresponding author: Brandon Morgan Introduction Orientation 0 Department of Industrial Psychology and People Management, College of Business and Economics, Faculty of Management, University of Johannesburg , Auckland Park , South Africa Orientation: Depression and anxiety can have undesirable consequences for employees and their employers. It is therefore important that employers pay attention to the existence and extent of depression and anxiety. However, measuring these constructs requires unbiased, reliable and valid instruments. Research purpose: To facilitate unbiased measurement of depression and anxiety, we investigated differential item functioning of the Centre for Epidemiologic Studies Depression Scale-Revised (CESD-R) and Generalised Anxiety Disorder Scale 7 (GAD-7) in a sample of non-clinical African and white working adults. Motivation for the study: Biased measurement instruments can lead to serious problems when comparing scores between groups, using raw score cut-offs, or creating norm scores. Practitioners are legally and ethically required to ensure that any instrument used is unbiased. Research approach/design and method: A cross-sectional survey design was used. The CESD-R and GAD-7 were administered to working adults. A final sample of 551 CESD-R responses and 529 GAD-7 responses were included in the analyses. Ordinal logistic regression was performed to investigate differential item functioning. Main findings: Both CESD-R and GAD-7 showed some evidence of differential item functioning although it was mostly small in magnitude. Item bias had some minor nonnegligible impact on aggregated observed scores within specific ranges of the underlying traits. Practical/managerial implications: Both CESD-R and GAD-7 show promise as instruments that can be utilised to explore the experience of anxiety and depression in African and white employees. Contribution/value-add: This study is a promising first step towards the measurement fairness of the CESD-R and GAD-7 in the South African context. Depression; Anxiety; GAD-7; CESD-R; Differential Item Functioning - Scan this QR code with your smart phone or mobile device to read online. employee well-being and organisational effectiveness (Bender & Farvolden, 2008; Evans-Lacko et al., 2016) . A workplace is ‘an ideal setting for depression and anxiety interventions’ (Mykletun & Harvey, 2012, p. 868) because adults spend much of their time in the workplace (Tan et al., 2014) . Indeed, research supports the effectiveness of workplace interventions for depression and anxiety (Joyce, Modini, Christensen, & Mykletun, 2016; Tan et al., 2014) . Despite these findings, anxiety and depression in the workplace appear to have received little research attention, particularly in South Africa, with limited research available on their workplace impact. Therefore, it is important that more research is conducted on these topics. However, to be able to do this, it is necessary to measure depression and anxiety in the workplace with appropriate, psychometrically sound and unbiased measuring instruments. The Centre for Epidemiologic Studies Depression Scale (CESD) (Radloff, 1977) and its revised version (CESD-R) (Eaton, Smith, Ybarra, Muntaner, & Tien, 2004) and the Generalised Anxiety Disorder Scale 7 (GAD-7) (Spitzer, Kroenke, Williams, & Löwe, 2006) are often used to measure depression and anxiety in clinical and non-clinical samples. These two instruments are particularly useful as they measure the symptoms most commonly associated with depression and anxiety. Research purpose and objectives Unfortunately there is limited evidence available on the reliability and validity of these two instruments when used in South African workplaces. It is particularly important that measurement bias, also referred to as Differential Item Functioning (DIF), in these two instruments as one form of validity is investigated. Differential item functioning means that respondents from different groups1 who have the same relative standing on some latent trait have different response probabilities on one or more items of an instrument that measures the latent traits (Chalmers, Counsell, & Flora, 2016; Zumbo, 1999) . These differences (after respondents are matched on the latent traits) usually reflect construct irrelevant factors that can confound observed scores (De Sa-Junior et al., 2019). This is problematic because decisions are made on observed scores with the assumption that these scores are uncontaminated by unwanted sources of variance (e.g. Gamerman, Gonçalves, & Soares, 2018; Steyn & De Bruin, 2018) . Differential item functioning often leads to item bias (Sireci, 2011) and failure to account for this bias can lead to biased or even incorrect decisions when item scores are translated into aggregated scale scores. Investigating DIF is therefore both a legal and an ethical imperative (Jodoin & Gierl, 2001) . In South Africa, the Health Professions Act of 1974 (Republic of South Africa, 1974) and the Employment Equity Act (EEA) of 1998 (Republic of South Africa, 1998) are clear on the legal and ethical imperatives of using unbiased instruments, with 1.Group is defined broadly and could include, for example, ‘groups differing in terms of gender, cultural background, education, ethnic origin, or age’ (International Test Commission, 2013, p. 17) . Zumbo (1999, p. 13) indicates that ‘standard comparisons are based on gender, race, sub-culture, or language’. the EEA indicating that no instruments (or scores obtained from these instruments) should be used in the workplace if they are unreliable, have limited validity evidence, and/or are biased against any person. It is the responsibility of practitioners to ‘contribute to specific empirical studies related to the psychometric properties of the tests they use’ (Health Professions Council of South Africa Form 208, 2006, p. 1) and to ensure that there is ‘[e]vidence relating to … DIF … [w]hen tests are to be used with individuals from different groups’ (italics in original, International Test Commission, 2013, p. 17) . Applied to the CESD-R and GAD-7 item bias (i.e. DIF) means that observed scores between groups cannot be directly compared because these scores consist of construct relevant and construct irrelevant sources of variance (De Sá Junior et al., 2019; Van De Vijver & Leung, 2011; Zumbo, 1999) . It also holds implications for cut-off scores used to indicate depression and anxiety (Carleton et al., 2013; Spitzer et al., 2006) and norm scores created for these instruments (Carleton et al., 2013). These two instruments must therefore be free of DIF if they are used to measure depression and anxiety in South African workplaces. Against (...truncated)