Psychometric evaluation of the Kessler 10 measure of psychosocial distress using the 2022 South African population-based household survey.

Abstract

The Kessler Psychological Distress Scale (K10) is widely used in epidemiological research but requires validation within South Africa's multilingual setting. This study evaluated the psychometric properties of the K10 using national population-based survey data and Item Response Theory (IRT). Graded Response Models (GRM) were used to assess factor loadings, item discrimination, and difficulty parameters, while Test Information Curves evaluated measurement precision across the distress continuum. Differential Item Functioning (DIF) analysis examined language effects, supporting cross-group reliability. All items exhibited high factor loadings (0.86-0.96) and substantial communalities (> 0.74) in the unidimensional GRM. Although the multidimensional GRM showed marginal improvement, relatively low factor loadings limited its suitability. All items had high discrimination parameters, ranging from 2.89 (K1) to 5.78 (K6). Difficulty parameters (β1-β4) followed the expected increasing pattern across response categories. Items K6 and K9 were the most informative given by the highest discrimination values. Internal consistency was excellent (Cronbach's alpha and McDonald's Omega total = 0.95), while marginal reliability under the unidimensional IRT model was moderate (0.65). Items 4, 5, 7, 8, and 10 were flagged for possible DIF by English-speaking status; effect sizes were negligible (R² < 0.002). This study provides robust evidence supporting the psychometric soundness of the K10 in a South African context. The scale functions as a unidimensional, internally consistent, and discriminating measure of psychological distress. Marginal reliability and information curves indicate it is most precise for detecting moderate-to-high distress levels, with potential for refinement in screening for milder symptoms.