Objectives: Because databases differ in subject matter, indexing, and holdings, more than one is used when conducting literature searches. As a result, the same reference can be captured multiple times, creating duplicate records in the retrieval. The duplicates must be removed before reference screening and data extraction. This study reviews the deduping features in the reference management software EndNote and the systematic review software tool DistillerSR. The hypothesis is that EndNote allows for more robust duplicate detection. However, the advantage to deduping in DistillerSR is that the duplicates are automatically accounted for in the PRISMA diagram that it can generate.
Methods: EndNote and DistillerSR each have default settings for duplicate detection which include the author and title fields. In each product it is possible to adjust what fields are used when detecting duplicates. Bramer et al’s 2016 JMLA article provides a method using various field combinations. Three separate searches were conducted in PubMed and Embase. The resulting references were exported to EndNote and then uploaded into DistillerSR. Deduping was done using the default settings and Bramer et al’s method in EndNote as well as in DistillerSR. While Bramer et al’s method was designed for EndNote, it is possible to use the same concept of manipulating the fields in DistillerSR.
Results: When EndNote was used the number of duplicates found using Bramer et al’s method was greater than the default. When DistillerSR was used the number of duplicates found was the same with either method. Both tools allow the user to compare the references identified as duplicates side by side. The user can also accept wholesale the duplicates identified by both programs. One difference is that DistillerSR displays a per cent confidence for each duplicate pair. The user can select the confidence threshold above which the duplicates are removed without review by the user.
Conclusions: If the researcher has access to DistillerSR it is recommended to use it for the deduplication process. There is no difference in which duplicates are found and doing it here allows the researcher to use the PRISMA diagram function. If the deduping is done in EndNote before uploading the references to DistillerSR the PRISMA diagram would not account for the duplicates. If the researcher does not have access to DistillerSR it is recommended to use Bramer et al’s method since more duplicates are found using it than with the default settings.