Crossref and PubMed Correction and Linking have the same basic goals (see What do PubMed and Crossref Correction and Linking do?): to compare the references you submit against metadata in a database of publications, return a link if possible, and check for discrepancies between what you submitted and what the database has on file for that source. In other words, these tools try to link your reference to a metadata database and then make or suggest corrections based on what information is found in that database.
From the user perspective, these two processes look almost identical. However, there are critical differences in how Crossref and PubMed Correction and Linking work with your submitted references--more specifically, in how corrections are handled. To understand how Crossref and PubMed correction in Edifix are different, it’s also important to understand why.
Both Crossref and PubMed Correction and Linking use patented fuzzy matching heuristics to match each of your submitted references to a corresponding database record. So far, so identical!
But once a match has been made, the next steps differ. Here’s why.
PubMed, a free resource developed and maintained by the National Center for Biotechnology Information at the U.S. National Library of Medicine (NLM), includes metadata for more than 22 million citations in the fields of biomedicine and health, covering portions of the life sciences, behavioral sciences, chemical sciences, and bioengineering.
Traditionally, metadata deposited to PubMed have been curated (i.e., checked against the version of record to ensure accuracy) by NLM staff; errors in PubMed records exist, but are rare. However, because PubMed indexes primarily journal articles in specific disciplines, Edifix PubMed processing will return very few PMID links for content outside those disciplines.
Crossref is a not-for-profit membership organization for scholarly publishing. Content publishers create a unique digital object identifier (DOI) for each item they wish to deposit, attach metadata for the item and the URL where it resides, and submits the record to Crossref, which then registers each article DOI and URL in a central DOI directory.
Because Crossref is not discipline or format specific, Edifix Crossref processing returns DOI links for content in a wide variety of fields (STM, social sciences, humanities…) and a variety of formats (journal articles, book chapters, books…) However, metadata deposited to Crossref are not curated, and Crossref records are therefore more likely to contain errors (often minor, sometimes major) picked up from the publisher’s incorrect metadata.
Corrections and Error Flags
Edifix checks both PubMed and Crossref records for discrepancies in
- Author names
- Article titles
- Journal titles
- Volume, issue, and page numbers
- Publication date
Edifix also checks Crossref records for discrepancies in
- Chapter titles
- Book titles
- Publisher location
- Author-supplied DOI, if any
When a discrepancy is found, what happens next depends on
- which database the discrepancy was found in (PubMed vs Crossref);
- which element of the reference is affected (author name, title, volume/issue/page number, date…); and
- whether the discrepancy is major or minor.
Because PubMed metadata records are curated, Edifix’s PubMed Correction and Linking is more likely to make a correction based on PubMed metadata. For example,
In this entry, Edifix has abbreviated the journal title (as the user-selected editorial style requires) and has also, based on the PubMed record, added the month of publication and the issue number.
In this entry, based on the PubMed record, Edifix has corrected the journal title (from Perspective to Perspectives and added the missing issue number, as well as correcting the format of author names to match the user-selected editorial style.
If the author list is different from the PubMed record, you’ll likely see a comment like this:
Note that Edifix, based on the PubMed record, has also corrected 2 words in the article title and corrected the journal title.
Because Crossref metadata records are created by publishers and are not curated, Edifix’s Crossref Correction and Linking is more likely to add a comment when the submitted reference and the Crossref record don’t match. For example,
The following example demonstrates why correcting a discrepancy from Crossref metadata instead of flagging it can be a bad idea:
What’s going on here? Well, it appears the publisher deposited the article title separately from the article subtitle--which means that when Crossref goes looking for the title associated with this DOI, it finds only the first half of it!
When Crossref and PubMed disagree
Sometimes, Crossref’s and PubMed’s records for a publication that appears in both databases don’t agree. Because errors are more likely in publisher-deposited Crossref metadata, Edifix will prioritize PubMed metadata in these cases.
This is particularly important to understand when you are using an editorial style that requires abbreviated journal names, because while Crossref publishers are asked to supply both full and abbreviated journal titles, Crossref’s process has no way of verifying that they have (a) provided both titles and (b) provided them correctly.
For more information, see Is data from PubMed and Crossref always reliable?