Issue 15 of our digest introduces you to the classifier in CRS Web. We're also looking at the differences between MEDLINE and PubMed, and presenting a useful resource - the SR Toolbox. If you have anything you'd like us to look at for the next edition, let us know: cis-support@cochrane.org.
Introducing the Classifier in CRS Web
New feature in CRS Web
One of the new features in CRS Web that will be introduced to early adopters in the coming months is the classifier. The classifier came about as part of the Pipeline component of Project Transform, with the aim of identifying previously untapped RCTs within the CRS that are likely to be of interest to Cochrane review groups. It works by analysing each group’s scope, based on the reviews it has published, its CENTRAL records and specialised register. It then uses machine-learning to generate “models” to apply to new references to determine how likely they are to be within a group’s scope.
CRS Web sends records to the classifier and retrieves a classifier score automatically, which is presented to Cochrane information specialists in a number of ways. A dashboard panel shows how many records are very likely to be of interest to a group, giving a quick overview. For more control, CISs can set the classifier score above a certain threshold. For those interested in classifier detail, each record has a graphical representation of the classifier scores for each Cochrane group indicating how likely a new record is to be of interest to their group as well as how it relates to other groups.
Initially, the classifier may make suggestions that seem difficult to fathom, this is because the models work on probabilities and so are not perfect. For example, there might be a record with just a title which gives little information for the classifier to use. It might also be that in some subject areas there are few or no unique concepts to distinguish records of interest from irrelevant records. The intention is to continually train the models with new records, so over time it will become more accurate. Long term, there is potential for the classifier to become an accurate and time-efficient method of populating our registers and CENTRAL, rather than through traditional scope searches.
CISs attending the Seoul colloquium will get the opportunity to see the classifier and learn more about how it works at the CRS Web training event.
Systematic Review Toolbox
Resource for systematic reviewers
Julie Glanville (Co-Convenor of the Cochrane Information Retrieval Methods Group) emailed the IRMG mailing list recently to alert us to a resource called SR Toolbox.
This has been compiled by Chris Marshall, a research consultant at the York Health Economics Consortium. It is a web-based catalogue of tools to support systematic reviewers. It includes resources which support searching. You can find it here. If you have any suggestions for new resources to add, you can get in touch with Chris Marshall via the contact page of the SR Toolbox here.
ePub ahead of print news
ePub ahead of print now available on Ovid
There was some discussion on the CIS mailing list recently about ePub ahead of print now being available via Ovid. But if you search MEDLINE Ovid, plus ePub ahead of Print and In Process is this the same as searching PubMed? The short answer is no!
Searching PubMed also gives access to:
Citations to articles that are out-of-scope (e.g., covering plate tectonics or astrophysics) from certain MEDLINE journals, primarily general science and general chemistry journals, for which only the life sciences articles are indexed with MeSH.
Citations that precede the date that a journal was selected for MEDLINE indexing (when supplied electronically by the publisher).
Pre-1966 citations that have not yet been updated with current MeSH and converted to MEDLINE status.
Citations to some additional life sciences journals that submit full text to PMC® (PubMed Central®) and receive a qualitative review by NLM.
Citations to author manuscripts of articles published by NIH-funded researchers.
Citations for the majority of books available on the NCBI Bookshelf (a citation for the book and in some cases each chapter of the book).
If you want to access these records on PubMed, you can do a search for publisher supplied records and OLDMEDLINE by using the syntax (publisher[sb] OR oldmedline[sb]).
See this useful blog post by Jacqueline Laikas on Things to Keep in Mind when searching Ovid MEDLINE instead of PubMed. There's more information on the differences between PubMed and MEDLINE in this NLM Factsheet.
CRS Web training opportunity
CRS Web Training in Seoul
If you're attending the Cochrane Colloquium in Seoul, you can now register to attend the CRS Web training event on Saturday 22nd October. This event is presented by Metaxis and the CIS Support Team, and participants will get the opportunity to see CRS Web and do some hands-on training.
If you would like to attend, email Anne if you haven't already done so.
Questions from the helpdesk
Do I need to search all sources again for a "top up" search, even if the grey literature searches haven't yielded any studies?
This was a question that came into the CIS Support mailbox recently. This was for a top-up search, prior to first publication of a review, not for a review update. The authors of the review were asking whether a top up search of the grey literature sources was needed, as they are not a mandatory MECIR standard requirement. Grey literature searching is "highly desirable" according to MECIR.Due to a delay in publication, the searches were 2 years old by this point.
There is no official policy on this. We advised that as this is a top-up search, authors should be encouraged to re-run all of the searches so that the review is compliant with the following MECIR standards:
R37 and R38: Rerun or update searches for all relevant databases within 12 months before publication of the review or review update, and screen the results for potentially eligible studies.
MECIR conduct standard 38: Incorporate fully any studies identified in the rerun or update of the search within 12 months before publication of the review or review update
But next time the review is updated, the authors may want to leave out the searches of the grey literature databases, and use the following wording:
"We changed the databases that we searched in the previously published version of the review because no unique relevant records were identified on the following databases in the original search: [list databases]"
How would you have handled this situation? Let us know on the CIS discussion forum.
Contact us...
Don't forget, previous editions of the Cochrane Information Specialist Support Team digest are available on the CIS Portal (Archie login required)
Contact the Support Team for help with any Cochrane Information Specialist related issue (including CRS technical support): cis-support@cochrane.org
The Cochrane Information Specialist Support Team:
Liz Doney I Sam Faulkner I Ruth Foxlee
Anne Littlewood I Doug Salzwedel
