27 The effects of exit from work on health across different socioeconomic groups 4165) were screened by one reviewer (RS) on title and abstract. Screening of 4765 articles on title and abstract resulted in 108 articles that were screened on full text. Screening of full-text articles was performed by two reviewers (RS and AdW) independently. Discrepancies were discussed until consensus was reached and a third reviewer (CB) was consulted in case consensus could not be reached. Finally, references of the included articles were checked for other possibly relevant articles. Data extraction and quality assessment One reviewer (RS) performed the data extraction by using a predefined dataabstraction form, extracting the following data per study: author, publication year and country, population (i.e. dataset, cohort or register, n, sex and age), design (i.e. type and follow-up period) statistical analyses (i.e. stratification and/ or interaction term), assessment of exit route (i.e. early/statutory retirement, unemployment or disability pension), health and SES, and the results of the effects of exit from work on health across socioeconomic groups. In case of uncertainty about the extracted data a second reviewer (AdW) was consulted. The quality assessment was performed by two reviewers (RS and AdW) independently and based on a set of nine predefined criteria (Table 1). The criteria were predominantly based on one review that focused solely on the relation between exit from work and health and on already existing criteria lists in the field of public health (6, 24-26). Each quality criterion was rated positive (+), negative (−) or not applicable (n.a.). Criteria 3, 4 and 5, were rated not applicable in studies with register data, because they could not provide information on participation rates. Differences in scores between reviewers (RS and AdW) were discussed and were resolved in consensus meetings. Studies with a minimum of 5 points (> 50%) were regarded as of high methodological quality (6, 24, 25). Studies in which criteria 3, 4 and 5 were rated not applicable and with a minimum of 3 points (> 50%) were regarded as of high methodological quality. The data extraction and quality assessment were performed per study to avoid multiplication. This means that some articles resulting from the same dataset, register or cohort were merged. Nevertheless, many articles resulting from the same dataset, register or cohort were not merged as they differed with regard to the health outcome. Consequently, different (smaller) datasets were retrieved from one large dataset, resulting in different studies. 2
RkJQdWJsaXNoZXIy MTk4NDMw