On the Dimensionality, Suitability of Sum/Mean Scores, and Cross-Country Measurement Invariance of the Perceived Stress Scale 10 (PSS-10)-Evidence From 41 Countries

Applied researchers typically use mean/sum scores as simple scoring method for the 10-item Perceived Stress Scale (PSS-10). However, inferences from such scores can be biased when measures are not essentially unidimensional. A bifactor modeling approach is especially useful to evaluate whether and to what degree ignoring (potential) multidimensionality of the PSS-10 might affect person score estimates. However, so far only few studies on the PSS-10 have used this approach and have not yet fully exploited its potential. In addition, studies that investigate measurement invariance across countries are scarce. Thus, the present study aimed to (a) evaluate the dimensionality of the PSS-10 by means of an exploratory graph analysis and an item response theory bifactor modeling comparison approach, (b) assess conditional and marginal reliability, (c) investigate cross-country measurement invariance, and (d) examine whether the relationships between the PSS-10 and different validity criteria depend on the scoring method of the PSS-10. Therefore, community samples from 41 countries (N = 109, 264) of the COVIDiSTRESS Global Survey data were used. In most countries (i.e., n = 33) ignoring multidimensionality did not substantially affect person score estimates and only slightly overestimated reliability. Measurement invariance and differential item/test functioning analysis of the PSS-10 on these 33 countries revealed only minor differential test functioning across countries. Overall, the PSS-10 showed a concurring nomological network with the external criteria, although with some substantial differences in effect sizes between countries.