The dimensionality of the Perceived Stress Scale: The presence of opposing items is a source of measurement error

Abstract

Introduction: The Perceived Stress Scale (PSS) is a widely recognized patient-reported outcome measure designed to assess an individual's level of perceived stress. The PSS consists of ten items, with six items phrased negatively and four phrased positively. Most studies have found that the PSS consists of two dimensions, with negative and positive items forming separate dimensions. However, some studies suggest a one-dimensional construct.

Objective: We aimed to investigate the dimensionality of the PSS and the impact of wording effects on the measurement properties of the scale.

Methods: We tested a modified version of the PSS (mPSS), with negative wording of all items. In a Danish sample, respondents were randomized to answer the PSS or the mPSS. We used confirmative factor analysis and Rasch analysis to assess the validity and reliability of the two versions. The proportion of person misfit was also evaluated.

Results: A total of 326 respondents completed the PSS, whereas 306 completed the mPSS. For the PSS, a two-factor model fitted the data best, and the first positively formulated item showed under-discrimination (Rasch model item fit rejected). The mPSS had higher measurement precision than the PSS, but evidence of local dependence was seen for both versions. Fewer respondents gave improbable responses to the mPSS compared to the PSS.

Limitations: Data comes from three different respondent samples.

Conclusion: The PSS was deemed essentially unidimensional, with the extra variance being attributed to the reversed items. To reduce response bias, clinicians and researchers may apply the mPSS.

Keywords: Dimensionality; Item wording; Measurement error; Perceived Stress Scale; Stress.