An Analysis of the Validity of Retraction Annotation in PubMed and the Web of Science

Research on scientific misconduct relies increasingly on retractions of articles. An interdisciplinary line of research has been established that empirically assesses the phenomenon of scientific misconduct using information on retractions, and thus aims to shed light on aspects of misconduct that previously were hidden. However, comparability and interpretability of studies are to a certain extent impeded by an absence of standards in corpus delineation and by the fact that the validity of this empirical data basis has never been systematically scrutinized. This article assesses the conceptual and empirical delineation of retractions against related publication types through a comparative analysis of the coverage and consistency of retraction annotation in the databases PubMed and the Web of Science (WoS), which are both commonly used for empicial studies on retractions. The searching and linking approaches of the WoS were subsequently evaluated. The results indicate that a considerable number of PubMed retracted publications and retractions are not labeled as such in the WoS or are indistinguishable from corrections, which is highly relevant for corpus and sample strategies in the WoS.