Research Communications Librarian Lane Medical Library, Stanford School of Medicine, Stanford University Mountain View, California
Objectives: One essential, yet time-consuming, phase in the systematic review process is identifying duplicate citations as searches are conducted comprehensively across multiple bibliographic databases. Although reference management programs contain algorithms designed to identify and remove duplicate records, literature suggests inconsistency in its success. This study seeks to evaluate the effectiveness of the duplication function in the commercial programs, EndNote and Covidence.
Methods: A sample systematic review search string is designed and translated in the databases Embase (Elsevier) and MEDLINE (Ovid) with a total of 4070 records extracted. Three duplication methods are applied to test the effectiveness of the duplication functions. First, the auto-duplication function in EndNote is used. Second, the systematic EndNote duplication method developed by Bramer et al. (2016) is tested. Lastly, the auto-duplication in Covidence is assessed. Total number of duplicates found, false positives, and false negatives are recorded for each method and calculated for its accuracy.
Results: Preliminary assessments are favorable to Bramer et al. (2016)’s systematic method of duplicate record identification in EndNote. The auto-duplication features in both EndNote and Covidence are limited in its success to correctly identify all duplicate records.
Conclusions: While Covidence has the potential to greatly facilitate the screening and review of citations for systematic reviews, it lacks rigor and advanced de-duplication features. A citation management program, such as EndNote, may better facilitate in identifying duplicates following Bramer et al. (2016)’s method, paired with manual screening. Both commercial programs have strengths and limitations in successfully identifying all duplicates. They can be used in combination to support different stages of the systematic review process.