As anyone who has worked on a systematic review (SR) project can attest – the record screening process can be a frustratingly tedious and time-consuming one. If available, most reviewers would likely welcome some kind of automation that streamlines and potentially reduces the manual record screening portion of their SR workload.
What is an RCT classifier algorithm?
An RCT classifier algorithm is “a tool to help you sort out the non-RCTs so that you can focus your effort on studies more likely to be included in your review”. In other words, researchers working on SRs that specify in their protocol that only studies reporting on RCTs will be included can now take advantage of tools that help them predict – using an automated algorithm derived from machine learning – whether a study is using a possible RCT or a not an RCT study design.
The research team behind the leading RCT classifier algorithm tool (which includes members of the EPPI-Centre and Cochrane) published a paper in May 2022 describing the development and evaluation of their tool:
Thomas J, McDonald S, Noel-Storr A, Shemilt I, Elliott J, Mavergames C, Marshall IJ. Machine learning reduced workload with minimal risk of missing studies: development and evaluation of a randomized controlled trial classifier for Cochrane Reviews. J Clin Epidemiol. 2021 May;133:140-151. doi: 10.1016/j.jclinepi.2020.11.003. Epub 2020 Nov 7. PMID: 33171275; PMCID: PMC8168828.
The good news for our MSK community is that this RCT classifier functionality has already been incorporated into Covidence, the systematic review project management system that the MSK Library subscribes to and provides access to. To turn this function on in a review that they are working on, a team member will need to have first selected the “Medical and health sciences” option under the “Area of Research” drop-down menu. After choosing to create this kind of review, the option to “Automatically tag studies reporting on RCTs using the Cochrane RCT Classifier” will become visible for a user to decide to enable of not. If enabled (only works with titles that have >15 characters and abstracts that have >400 characters), their SR records will be tagged as “Possible RCT” or “Not RCT” and can be filtered accordingly.
Learn more about “How to tag studies not reporting on RCTs” by checking out this article from the Covidence knowledgebase.
Questions? Ask Us at the MSK Library.