Assessing the psychosocial work environment in the health care setting: translation and psychometric testing of the French and Italian Copenhagen Psychosocial Questionnaires (COPSOQ) in a large sample of health professionals in Switzerland

Background
Measuring work-related stress in a reliable way is important in the development of appropriate prevention and intervention strategies. Especially in multilingual studies the use of comparable and reliable instruments is crucial. Therefore, the aim of this study was to translate selected scales and single items from the German version of the Copenhagen Psychosocial Questionnaire (COPSOQ) into French and Italian and psychometrically test them in a sample of health professionals.
Methods
This study used cross-sectional data from health professionals at 163 randomised selected health organisations in Switzerland. Selected COPSOQ items/scales were backwards- and forwards- translated and cross-culturally adapted from German to French and Italian. Reliability was assessed with Cronbach alpha and intraclass correlation coefficients, construct validity with confirmatory factor analysis (CFA) and structural equation modelling as well as comparative fit index.
Results
Responses from 12,754 health professionals were included in the analysis. Of the overall 24 scales, 20 in the German version, 19 in the French version and 17 in the Italian version attained sufficient internal consistency with a threshold of 0.7 for Cronbach’s alpha. Predominantly high factor loadings on scale level are reported (> 0.35), as well as good and satisfactory fit values with RMSEA below 0.1, SRMR below 0.08 and CFI above 0.95. For 10 out of 15 scales, the test for factor invariance revealed a significant difference regarding the psychological constructs of the scales across the language versions.
Conclusions
The psychometric properties verify the underlying theoretical model of the COPSOQ questionnaire, which is to some extent comparable across the three language versions. Of the 10 scales with significant factor variance, four showed large differences, implying that revision is needed for better comparability. Potential cultural issues as well as regional differences may have led to the factor variance and the different reliability scores per scale across language versions. One known influencing factor for regional differences is culture, which should be considered in scale development. Moreover, emerging topics such as digitization should be considered in further development of the questionnaire.