Presentation (CTI)

OBJECTIVES: To evaluate the Covidence relevance ranking tool for accelerating the identification of eligible publications in literature reviews.
METHODS: For 3 previously conducted systematic reviews, records were re-screened at title/abstract stage using previous screening decisions with records set to order by relevance. The number of eligible publications identified at multiple stages during title/abstract screening was recorded, and the proportion of records screened required to identify 90% and 100% of included publications was determined. A sensitivity assessment was conducted excluding eligible review records without associated abstracts.
RESULTS: Assessed systematic reviews were of effects and safety (N=2) and prognosis (N=1). Searches had retrieved 2,300, 4,098 and 3,468 records for screening, with 115, 30 and 136 included publications, respectively. To identity 90% of publications, reviewers needed to screen 22%, 22% and 27% of the total records. To identify 100% of publications, 58%, 32% and 53% of records needed to be screened. Excluding eligible records without abstracts (excluding 29, 0 and 2 records, respectively), sensitivity improved, with 17%, 22% and 26% of records screened to identify 90% of publications, and 28%, 32% and 35% of records screened to identify 100% of publications.
CONCLUSIONS: The Covidence ranking tool is useful for rapid identification of relevant publications in systematic reviews. In pragmatic reviews, there is potential for the tool to be used to reduce the number of records screened. Records without abstracts could be prioritised for manual screening to optimise workload savings.