Ordinal scaling properties of the WHO-5 among young adolescents in Moldova and North Macedonia: a Mokken Scale Analysis

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Abstract Background The World Health Organization Five Well-Being Index (WHO-5) is commonly applied to assess subjective well-being. However, evidence on its ordinal measurement properties among young adolescents, especially in low- and middle-income countries, is scant. In particular, there is a lack of studies that have comprehensively evaluated key Mokken model assumptions, including scalability, monotonicity and invariant item ordering. Evaluating these properties is important as they support the ordinal interpretation of WHO-5 scores and clarify whether the items form an invariant hierarchy. Methods Baseline data from the FLOURISH project, a multi-country study evaluating a family-focused prevention program in Moldova and North Macedonia, were used. The sample comprised 653 adolescents aged 10–14 years. The ordinal scaling properties of the WHO-5 were examined using Mokken Scale Analysis, a non-parametric item response theory approach. Specifically, overall and item-level scalability were evaluated using Loevinger’s H coefficients. Single monotonicity was evaluated by examining local decreases in item step response functions, while invariant item ordering was assessed by examining pairwise intersections between manifest item response functions. The accuracy of the estimated ordering was evaluated using the H T coefficient. Analyses were also conducted separately by gender and country. Results The WHO-5 demonstrated strong overall scalability ( h = 0.53, 95% CI: 0.49–0.57), with item-level scalability coefficients ranging from 0.50 to 0.56. Analyses revealed only minor and infrequent deviations from single monotonicity, with all diagnostic values below established thresholds. In addition, no meaningful violations of the nonintersection assumption were identified in the overall analysis. However, the H T coefficient was 0.12, indicating that the estimated item ordering was too inaccurate to be useful. Conclusions The findings corroborate evidence for the scalability and monotonicity of the WHO-5 and support the use of its summed score to distinguish adolescents with relatively lower or higher well-being at a given assessment, while cautioning against interpreting the group-level item ordering as a hierarchy that is accurately reproduced in individual adolescents’ response patterns.