Abstract
Background All research has room for improvement, but authors do not always clearly acknowledge the limitations of their work. In this brief report, we sought to identify the prevalence of limitations statements in the medRxiv COVID-19 SARS-CoV-2 dataset.
Methods We combined automated methods with manual review to analyse manuscripts for the presence, or absence, either of a defined limitations section in the text, or as part of the general discussion.
Results We identified a structured limitations statement in 28% of the manuscripts, and overall 52% contained at least one mention of a study limitation. Over one-third of manuscripts contained none of the terms that might typically be associated with reporting of limitations. Overall our method performed with precision of 0.97 and recall of 0.91.
Conclusion The presence or absence of limitations statements can be identified with reasonable confidence using automated tools. We suggest that it might be beneficial to require a defined, structured statement about study limitations, either as part of the submission process, or clearly delineated within the manuscript.
Competing Interest Statement
The authors are the founders and shareholders of Scholarcy Limited. Scholarcy technology was used to extract the information from the manuscripts used in this study.
Footnotes
Conflicts of Interest: The authors are the founders and shareholders of Scholarcy Limited. Scholarcy technology was used to extract the information from the manuscripts used in this study.
Results section updated to clarify the denominators used in the calculations. Discussion section updated to clarify results in comparison to earlier work using automated methods. Added Acknowledgement
https://ai2-semanticscholar-cord-19.s3-us-west-2.amazonaws.com/latest/biorxiv_medrxiv.tar.gz