Acta Psychologica (Nov 2023)

The Perceived Stress Scale is essentially unidimensional: Complementary evidence from ancillary bifactor indices and Mokken analysis

  • Tyrone B. Pretorius

Journal volume & issue
Vol. 241
p. 104058

Abstract

Read online

Stress is a significant contributor to physical and mental health. However, there is considerable variability in response to stress, underscoring the role of stress perception in health outcomes. Perceptions of stress are influenced by socio-demographic and cultural factors. The Perceived Stress Scale (PSS) has been used extensively to assess stress among different populations and for cross-cultural comparisons. Given its wide-ranging application, the rigorous assessment of its psychometric properties is necessary to solidify its standing as a dependable measure. The scale was originally conceptualized as unidimensional but there has been controversy regarding its factor structure, leading to various factor models. Conclusions about the structure of the PSS were largely based on fit indices in confirmatory factor analysis and often did not go beyond fit indices. In addition, classical test theory indices are sample dependent and it is not surprising that instruments can exhibit different properties in different samples. Over reliance on fit indices can lead to overly simplistic interpretations of underlying dimensionality. Hence, the current study extends this research by using item response theory and examining ancillary bifactor indices to determine the amount of variance explained by the total scale and subscales. Participants were students (n = 322) at a South African institution who completed the PSS and a demographic questionnaire. Fit indices in confirmatory factor analysis indicated that a one-factor model, a bifactor model and a correlated-two factor model fit the data to an acceptable degree (GFI > 0.95, CFI > 0.90, RMSEA <0.08). Ancillary bifactor indices supported the essential unidimensionality of the scale as the two subscales accounted for 18.9 % and 17.2 %, of the variance, respectively, while the total scale accounted for 63.9 % of the variance. This was confirmed by Mokken analysis, which indicated that all the items loaded on one scale. The unidimensionality of the PSS suggests that it can be an effective clinical screening tool and that it is easily adaptable across different populations and contexts, enabling standardized comparisons. However, it may lose sensitivity to cultural differences in how stress is perceived and experienced. Hence, integrating the PSS with tailored assessments would facilitate a more robust and inclusive approach to stress assessment in various settings and populations. The study did not take specific sources of stress into account and was conducted among a distinct population group. Future studies undertaken among diverse populations and linking global stress with the impact of stressful life events are needed to further confirm our results and understand the interconnection between these factors.

Keywords