Study selection

We systematically searched Medline and EMBASE on 18/04/2018 and on 26/02/2020. The search strategy included a combination of free text words and database-specific subject-headings (Additional file 1) using Ovid interface. We used Endnote X7 for the creation of the literature database and the removal of duplicates. Two authors (FW, AB) independently screened titles/ abstracts and full texts for eligibility resolving discrepancies by consulting a third author (DK). Additionally, an expert panel (MR, JM, Rainer Rossi) highlighted missing relevant papers, a screening of reference lists and citing works of included studies was performed. After full-text-screening, we conducted a hand search including forward (citing literature) and backward (cited literature) screening of included studies. Discrepancies during screening, extraction or quality assessment were solved by consulting of another reviewer (JS). For interpretation of reliability, we applied the prevalence-adjusted bias-adjusted kappa (PABAK). The advantage of PABAK in contrast to Kappa value is the consideration of the high class imbalance [25].