Abstract
Aims: Reliabilities of the work environment questionnaire Copenhagen Psychosocial Questionnaire (COPSOQ) have previously been estimated by Cronbach’s alpha, but since the internal consistency assumption may not apply to all COPSOQ scales, Cronbach’s alpha may underestimate true reliability. This study aims to evaluate reliability in a test—retest design. Methods: We analyzed postal questionnaire data from 349 persons (of whom 283 were employees) who completed two forms with a median interval of 22 (range 6—65) days between baseline and follow-up. Test—retest reliabilities were estimated by the intraclass correlation (ICC). For scales where the internal consistency assumption was theoretically plausible, reliabilities were also estimated by Cronbach’s alpha and by Green’s test—retest alpha. Results: With one exception, the ICC estimated reliabilities of the COPSOQ scales were adequate or good (range 0.70—0.89). A scale concerning mutual trust between employees had a low reliability of 0.64. Among the scales where the internal consistency assumption was plausible, Cronbach’s alpha was adequate or good (0.75—0.85) for seven out of eight scales. Green’s retest alpha was adequate or good for six out of eight scales (0.72—0.81). Conclusions: Standard criteria for acceptable intraclass correlation reliability were achieved for all COPSOQ scales but one. The test—retest design and intraclass correlation appears to be more appropriate than Cronbach’s alpha for assessing the reliability of psychosocial work environment scales.