Programme no. 163-P
Quality Improvement
Construct validity of the Perceived Stress Scale among adult Danes in the Central Denmark Region
Marie Mortensen1, Mogens Vestergaard2, Per Bech3, Eva Ørnbøl4, Finn Breinholt Larsen5, Mathias Lasgaard6, Kaj Sparle Christensen*7
1Research Unit for General Practice,Institute of Public Health, Aarhus University,Aarhus,Denmark, 2Research Unit for General Practice and Section for General Medical Practice,Institute of Public Health, Aarhus University,Aarhus,Denmark, 3Clinimetrics Centre for Mental Health (CCMH),Mental Health Centre North Zealand,Hillerød,Denmark, 4Research Clinic for Functional Disorders and Psychosomatics,Aarhus University Hospital,Aarhus,Denmark, 5Public Health and Quality Improvement,Central Denmark Region,Aarhus,Denmark, 6Public Health and Quality Improvement,Central Denmark Region,Aarhus,Denmark, 7Research Unit for General Practice,Institute of Public Health, Aarhus University,Aarhus,Denmark
* = Presenting author
Objectives: In the present study, we aimed to explore the construct validity of the PSS using methods from modern test theory.
Background: Psychological distress has considerably impacts on the quality of life worldwide, and stress is a known risk factor for mental and physical disorders. Originally the Perceived Stress Scale (PSS) was argued to be a single construct with six negatively –and four positively worded items (item 4, 5, 7 and 8) which are reversely scored. Subsequent studies indicated that a two-factor structure was more dominant. The PSS has previously been studied by methods from classical test theory.
Results: The analysis showed floor effect in the responses. The data did not fit the Rasch model for all ten items and the five response categories. Item 4 had the largest misfit in most performed analyses. Response categories were collapsed for two items, 4 and 7 but it did not improve the fit of data. The assumption of unidimensionality was not fulfilled and data did not fit the model for the suggested two dimensions. The collapsed response categories for two of the ten items did not improve the model fit. Data did not fit the model for subgroups of age and gender.
Material/Methods: The study population consists of 30,000 citizens in the Central Denmark Region, who completed the PSS in the Danish National Health Survey in 2010. We examined if data fitted the Rasch model for a unidimensional model and for the suggested two dimensions. We investigated fit to the model with ordering of response categories and collapsed response categories. Additionally we tested fit of data to the model in subgroups based on age and gender and by subsequently deleting items with the largest misfits.
Conclusion: The results suggest that we face scalability challenges with the current version of the PSS scale.
Points for discussion: For future research using PSS we suggest revision of the scale to improve fit of data to the Rasch model. We need to assess the content validity of the scale, especially regarding the positive formulated items.