The update process
When updating a previous search, the medical library uses its own method. An article on this method is currently being prepared as part of a larger series of articles on EndNote for Systematic Reviews. It works like this:
All results as they currently come from the databases are downloaded in EndNote and deduplicated. The numbers of results from this action are included in the Word document that is sent to account for the method.
The previous EndNote file is added in its entirety to that new EndNote file. After that, deduplication takes place again, but both duplicates are removed. The EndNote file therefore only contains the articles that were not in the previous EndNote file.
The numbers are different
However, the numbers of the results in the two EndNote files added together are usually not equal to the total number documented in the Word document. That’s because articles that were found last time are now no longer found. This can be for several reasons:
- An article no longer falls in the relevance ranking that has been used. For example, the first 200 articles from Google Scholar are used regularly. When other relevant articles are found, or Google changes the sorting algorithm, previously found articles can now be placed in a place above 200.
- An article is no longer in the database. For example, PubMed as supplied by publisher is searched quite widely (recently this database is no longer used for new publications, because Medline Ovid now also contains Epub ahead of print). An article may now have been included in Medline, and no longer fall within the somewhat stricter search.
- The thesaurus terms of an article have been changed. Particularly in embase, keywords are often automatically assigned initially. These will be revised later by specialized employees. Perhaps the automatically added thesaurus terms were irrelevant.
For similar reasons, it may also be the case that the update finds articles that have a publication date before the original search date.
How do you get your flow chart right?
A flow chart of the inclusion/exclusion process is often published in a review article. The numbers thereof can be entered correctly as follows.
# Records screened for relevance is known: this equals the number mentioned in the Word document
# Full articles accessed for eligibility is known: that is the number of articles that were read full text in the previous screening, plus the number of articles that were read full text in this round.
The difference between the to numbers equals # records excluded.
N.B. In the first round of screening for title and abstract, the explicit reasons do not have to be given with the numbers of why articles were excluded. This is only necessary for full text screening.