Article type
Abstract
BACKGROUND: The Retraction Watch (RW) database collects retractions from across scholarly publishing. In 2023, it became freely available, opening opportunities for citation software developers, publishers and academic researchers. It is the most comprehensive source of its kind with 43,000 retractions from across all fields and disciplines. Studies are retracted for a range of different reasons, from the purely administrative (e.g. duplicate publications) to the research integrity-relevant (e.g. plagiarism, serious error in data). Retracted studies also have downstream implications for publications that cite the original study, which is most salient for systematic reviews where the retracted study may affect the review results and findings.
OBJECTIVE: To test a novel method for using Retraction Watch data for research integrity research and demonstrate that not all retractions are created alike and there are issues with categorizing the reasons for retractions.
METHODS: We used a test dataset of first and last authors of Cochrane Reviews (CR) drawn from the Cochrane database (N = 21845). We searched the Retraction Watch (RW) database based on first and last names, using the software package “nameparser” from Python. For both datasets (CR and RW), some normalization was applied to the first and last names, and a CR author was then matched to an RW author based on specific rules. We then broadly categorized the diverse array of reasons for retractions into administrative vs research integrity concerns.
RESULTS:We determined that this method is in principle suitable for determining the percentage of authors in a given dataset with integrity-relevant retractions. However, the categorization of retraction reasons remains a constraint, since in some cases information about reasons is absent or unclear.
CONCLUSIONS: Studies can be retracted for a variety of reasons. Retractions with research integrity-related reasons are of most concern to systematic reviewers, but identifying the true reason for a retraction can be difficult. With the Retraction Watch database now freely available, editors and publishers can build resources to ensure retracted studies are not included in systematic reviews, though more reliable and objective methods for categorizing reasons for retractions are needed.
OBJECTIVE: To test a novel method for using Retraction Watch data for research integrity research and demonstrate that not all retractions are created alike and there are issues with categorizing the reasons for retractions.
METHODS: We used a test dataset of first and last authors of Cochrane Reviews (CR) drawn from the Cochrane database (N = 21845). We searched the Retraction Watch (RW) database based on first and last names, using the software package “nameparser” from Python. For both datasets (CR and RW), some normalization was applied to the first and last names, and a CR author was then matched to an RW author based on specific rules. We then broadly categorized the diverse array of reasons for retractions into administrative vs research integrity concerns.
RESULTS:We determined that this method is in principle suitable for determining the percentage of authors in a given dataset with integrity-relevant retractions. However, the categorization of retraction reasons remains a constraint, since in some cases information about reasons is absent or unclear.
CONCLUSIONS: Studies can be retracted for a variety of reasons. Retractions with research integrity-related reasons are of most concern to systematic reviewers, but identifying the true reason for a retraction can be difficult. With the Retraction Watch database now freely available, editors and publishers can build resources to ensure retracted studies are not included in systematic reviews, though more reliable and objective methods for categorizing reasons for retractions are needed.