A sum score ranging from 5 to 30 was calculated—a high score indicates a high level of exhaustion; standard deviation 5.9. Psychometric properties which were shown to be excellent were published for the Swedish version in Hanson et al. (2008). Depressive symptoms were measured with a brief subscale from the Hopkins Symptom Checklist (SCL-90). The scale measures one-week prevalence and includes six items covering the depressive core symptoms ‘feeling

blue’; ‘feeling no interest in things’; ‘feeling lethargy or low in energy’; ‘worrying too MM-102 chemical structure much about things’; ‘blaming yourself for things’; ‘feeling everything is an effort’. Response options cover five steps from ‘not at all’ to ‘a great deal’. A sum score of depressive symptoms ranging from 0 to 24 was calculated with a high score indicating a high probability of clinical depression; standard deviation 5.1. Scale characteristics for this short Swedish version which were shown to be excellent have been published by Magnusson Hanson et al. (2009). Statistical methods In the first step product moment correlations

were calculated between all explanatory study variables using accumulated scores for both cultural activity and the work-related variables. Since the study variables were normally or close to normally distributed, multiple linear regressions were performed in the next step using cultural activity as explanatory and health variables as outcome. Epothilone B (EPO906, Patupilone) Each wave was analysed separately. The regressions were made in three successive versions: 1. with adjustment for age, gender and nlog (income) from the ATR inhibitor actual study year and education from Etomoxir manufacturer self-reported data in 2006.   2. In addition to 1. also with adjustment

for listening/non-listening manager.   3. In addition to 2. also with adjustment for psychological demands and decision latitude at work.   Since income has had a more important role as a confounding factor in the association between cultural activities and health in previous research (Bygren et al. 1996) and since decision latitude and education are strongly correlated, we decided to use income and decision latitude but not education as a proxy for socioeconomic status in the multiple regressions. A final step was two series of prospective analyses, one from 2006 to 2008 and one from 2008 to 2010, in the form of multiple linear regressions. In these analyses age, gender and logarithmically transformed income as well as the work-related variables listening/non-listening manager, psychological demands and decision latitude and finally cultural activities at work and psychological state (emotional exhaustion and depressive symptoms, respectively) at start of follow-up (2006 and 2008, respectively) were used as predictors and subsequent psychological states (emotional exhaustion and depressive feeling 2008 and 2010, respectively) as outcomes.