Considerations on the calculation of fractions of cardiovascular disease attributable to psychosocial work factors

Motivated by the recent publication of Niedhammer et al. (2013) we would like to communicate some in our view noteworthy considerations concerning the measurement of psychosocial stress in epidemiological studies and the calculation of the population attributable fraction based on these studies with regard to research aimed at the prevention of disease. Changes in the workplace and in the working population lead to a continuous steep increase in the literature on the association of psychosocial stress experienced at the workplace and disease in particular cardiovascular diseases (CVD) (reviewed by Kivimäki et al. 2006, 2012; Backé et al. 2012; Eller et al. 2009; Belkic et al. 2004). Also in the recent publication of Niedhammer et al. (2013), population attributable fractions (PAF) for psychosocial work factors were calculated in relation to CVD and mental diseases. The choice of the concept of the PAF is reasonable in order to translate epidemiological evidence into policy and practice in the field of cardiovascular health in the workplace. The proportion of cases (morbidity and mortality) in a population attributable to a given exposure should provide information on most urgent factors that need to be addressed in prevention strategies. Most of the studies on CVD investigate the association between job strain measured by the Job Content Questionnaire (JCQ) (Karasek et al. 1998). Fewer studies use the effort–reward imbalance (ERI) model (Siegrist et al. 2004) or the organisational injustice model (Elovainio et al. 2006) or other instruments. There are different ways to derive PAFs for a population (e.g., country or region), either directly from a population-based study or indirectly. With the indirect approach, risk estimates from one or more analytical studies are retrieved and combined with information on the fraction of exposed persons in the general population from other sources (mainly surveys). Risk estimates may be derived from studies selected based on specific quality criteria (e.g., a certain design and/or statistical model including the relevant confounders) or from meta-analyses, respectively. When using this method, survey questions to estimate the prevalence of exposure need to be comparable to the instruments used for the exposure in the observational studies, which are the basis for the calculation of risk estimates. Validity of the PAF depends heavily on the estimation of the prevalence as well as risk estimates, given that they are correctly estimated (Olsen 1995). Niedhammer et al. (2013) used proxies for the job strain and effort–reward imbalance from the fourth European Working Condition Survey (EWCS) and combined the prevalences with risk estimates from published meta-analyses. With this indirect method, the authors describe PAFs between 2.51 and 5.77 % for job strain and 9.78–27.89 % for the effort–reward ratio [1 in the European countries. Reviewing the literature on fractions of CVD attributable to psychosocial work factors, we also saw that the estimated PAFs differ severely between countries (Backé et al. 2013; Backé and Latza 2013). With the indirect approach, PAFs for cardiovascular outcomes attributed to occupational stress have been derived for the United States (Steenland et al. 2003), Finland (Nurminen and Karjalainen 2001), Korea (Ha et al. 2011), and France (Sultan-Taı̈eb et al. 2011). For Sweden, PAFs in relation to several diseases were calculated by Järvholm et al. (2013). Here, with respect to job strain and myocardial infarction, calculations E. Backé H. Burr U. Latza (&) Work and Health, Federal Institute for Occupational Safety and Health, Noeldnerstraße 40-42, 10317 Berlin, Germany e-mail: [email protected]