Construct validity and factor structure of the Kessler-10 in South Africa

Hoffman, Jacob; Cossie, Qhama; Ametaj, Amantia A; Kim, Hannah H; James, Roxanne; Stroud, Rocky E; Stevenson, Anne; Zingela, Zukiswa; Stein, Dan J; Gelaye, Bizu

Construct validity and factor structure of the Kessler-10 in South Africa

Journal Article

2022-07-18

Journal Title

BMC Psychology

Link to Journal

https://bmcpsychology.biomedcentral.com/

Department

Department of Psychiatry and Mental Health

Faculty

Faculty of Health Sciences

License

http://creativecommons.org/licenses/by/4.0/

Abstract

Background The Kessler Psychological Distress Scale (K-10) is a short screening tool developed to identify, with good sensitivity, non-specific psychological distress in the general population. Sensitivity and specificity of the K-10 have been examined in various clinical populations in South Africa; however, other psychometric properties, such as construct validity and factor structure, have not been evaluated. We present evidence of the prevalence and severity of psychological distress in an outpatient setting in South Africa and evaluate the internal reliability, construct validity, and factor structure of the K-10 in this population. Methods We explored prevalence estimates of psychological distress using previously established cutoffs and assessed the reliability (consistency) of the K-10 by calculating Cronbach’s alpha, item-total correlations and omega total and hierarchical coefficients. Construct validity and factor structure of the K-10 were examined through split-sample exploratory factor analysis (EFA) followed by confirmatory factor analysis (CFA), comparing several theoretical models and the EFA. Results Overall, there was low prevalence of psychological distress in our sample of 2591 adults, the majority of whom were between the ages of 18–44 (77.7%). The K-10 showed good construct validity and reliability, with a Cronbach’s alpha of 0.84 and omega total of 0.88. EFA yielded a four-factor solution with likely measurement artifacts. CFA showed that the four-factor model from EFA displayed the best comparative fit indices, but was likely overfitted. The unidimensional model with correlated errors was deemed the best fitting model based on fit indices, prior theory, and previous studies. Conclusion The K-10 displays adequate psychometric properties, good internal reliability, and good fit with a unidimensional-factor structure with correlated errors. Further work is required to determine appropriate cutoff values in different populations and clinical subgroups within South Africa to aid in determining the K-10’s clinical utility.