Can we reduce the workload of systematic reviews without compromising quality? A new platform to accelerate search and study identification
2Cochrane Madrid, Universidad Francisco de Vitoria (UFV-Madrid), Clinical Biostatistics Unit, Hospital Universitario Ramón y Cajal (IRYCIS), CIBERESP, Institute for Complementary and Integrative Medicine, University Hospital Zurich and University of Zurich, Spain
3Iberoamerican Cochrane Centre, Biomedical Research Institute San Pau (IIB Sant Pau-CIBERESP), Barcelona, Spain
4Cochrane Madrid, Clinical Biostatistics Unit, Hospital Universitario Ramón y Cajal (IRYCIS), Spain
5Cochrane Madrid, Universidad Francisco de Vitoria (UFV-Madrid), Clinical Biostatistics Unit, Hospital Universitario Ramón y Cajal (IRYCIS), Spain
6Cochrane Madrid, Clinical Biostatistics Unit, Hospital Universitario Ramón y Cajal (IRYCIS), CIBERESP, Spain
7Iberoamerican Cochrane Centre, Biomedical Research Institute San Pau (IIB Sant Pau), Barcelona., Spain
Background: The development of systematic reviews (SRs) is resource intensive. The Sustainable Knowledge Platform (SK platform) is a new system that streamlines all the steps of SRs, applying technology to make the process efficient. It includes a comprehensive taxonomy, and off-the-shelf boolean strategies and predeveloped artificial-intelligence algorithms for each term. This platform decreases the workload, but it is not clear if this is not at the expense of compromising quality.
Objectives: To compare the performance of the SK platform against the traditional approach, to search and select studies in the context of a SR of trials of aspirin for the prevention of colorectal cancer in average-risk adults.
Methods: For the traditional approach, we designed a standard search strategy for MEDLINE, EMBASE and CENTRAL up to January/2023. For the alternative approach, we matched the components of the review question to the relevant terms of the taxonomy, ran the predeveloped boolean strategies, and applied the automated classifiers. Title/abstracts will be assessed by a single reviewer for each approach, and the full-text of the potentially eligible studies will be assessed in duplicate (independent reviewers for each approach). As a measure of workload, we will estimate the total number of records. We will calculate the sensitivity and specificity of the alternative approach using the traditional approach as the reference standard. We will report the number of studies detected by the alternative approach only, and will attempt to explore the potential reasons (e.g. different repositories, different strategies).
Results: The traditional approach identified 2,766 potentially eligible records. The alternative approach identified 916 records based on the predeveloped search strategies. After applying the automated classifiers, the number to screen decreased to 189. The complete results will be reported during the Colloquium.
Conclusions: New approaches may save time and resources. However, it is important to know if this efficiency gain does not come at a cost in quality. If our results are positive, the SK platform would offer an alternative worth exploring in future research.
Patient, public and/or healthcare consumer involvement: Using technology in SRs process may help to produce rigorous, relevant, and rapid evidence synthesis.