From: Vonville, Helena M <[log in to unmask]>
Subject: Re: cost of searching for evidence....
To: [log in to unmask]
Received: Tuesday, April 10, 2012, 11:14 PM

I'm curious why you would go through two steps-- titles first, then abstract.  Why not do title/abstract (when available) as one step?  I find that many authors do not know how to write a title.  Given enough words, they eventually are able to describe their study (more or less) in an abstract. The PRISMA flowchart moves from citation/abstract to FT so that's what we follow. 

According to the National Library of Medicine:

"Generally, there are no abstracts for records created before 1975."

If there's no abstract, then we review the title closely and look at the MeSH terms to try to get a clue (including publication type).  However, titles are often cryptic enough that the FT is required.


As for breaking up a team-- what about quality control?  At some point, the team should have been checked for agreement to make certain eligibility criteria are clear to all.  If they were in agreement, then broken up, the new team would have to be checked again.

I've not seen truncated abstracts but I guess I haven't screened enough citations.  However, either the study author(s) have described their study by that point (in which case you decide on your eligibility criteria) or they haven't.  If they haven't and you truly have questions as to whether it should be included, get the FT. 

Also according to the National Library of Medicine:

"The maximum length of abstracts for records created after 2000 is 10,000 characters. Original policy on inclusion of abstracts set a limit of 250 words for acceptance. Effective with the January 1984 data (i.e., NLM's ELHILL legacy system 8401 Entry Month) two changes were made in this policy: 1) the word limit was expanded to 400 words for abstracts from articles ten pages or more in length or from articles in the cofre journals identified by the National Cancer Institute and 2) abstracts exceeding the 250- or 400-word limit are to be included in truncated form at the end of the sentence closest to the word limit. The percentage of records with abstracts has increased over the years as more publishers gave permission for NLM to include these data. "

So, older articles won't have an abstract and, since there really weren't standards for reporting, the titles are often less than informative.  Since 2000, abstracts can be up to 10,000. 

Those citations between 1975 and 2000 are the ones that could be potentially truncated (with some caveats).  But-- there should be a protocol in place if you encounter a truncated abstract so that everyone follows the same progression of steps.

Helena

Helena M. VonVille, MLS, MPH
Library Director
University of Texas School of Public Health Library
Houston, TX  77030
[log in to unmask]">[log in to unmask]
713.500.9131 (office)
713.500.9125 (fax)
Need help? Ask a Librarian!
Or text us: 713-965-6968

-----Original Message-----
From: Evidence based health (EBH) [mailto:[log in to unmask]">[log in to unmask]] On Behalf Of Patricia Anderson
Sent: Tuesday, April 10, 2012 11:35 AM
To: [log in to unmask]">[log in to unmask]
Subject: Re: cost of searching for evidence....

Hi, folks,

I've been advising a couple systematic review teams here, and a few methodology questions came up. I gave them my two cents, but wanted to check with the wider and wiser community at large.

1) Title review has been completed, about to start abstract level review. They want to switch the people doing the reviews midstream.
The only review I've worked with that encountered a need to do this was when one of the reviewers took seriously ill, and they ended up starting the review over. In this case, they mostly have some people busier than others and want to offer a variety of people the experience of working on a review. I recommended against switching reviewers midstream for the sake of consistency throughout the process. Thoughts?

2) Medline often truncates long abstracts. At the abstract level review, what do folks do when they encounter one of these truncated abstracts? What I'm familiar with and recommended is to review the part made available from MEDLINE as supplied by the librarian in search results files. For one team, the PI started downloading the full article for every truncated abstract, and the team was having trouble reaching consensus because the two reviewers were working from different sets of information. I was also concerned about the risk of introducing bias. This has come up before, and I am usually told (from different points of view):
- "I'm a researcher. Just trust me to not look at the rest of the article."
- "I don't mean to, but I can't help it. It's so interesting, and my eyes just go there. I promise, I won't include that information in my decision."

3) Citations with no abstract. Are they automatically shunted to the full article review level, or automatically excluded? For one topic, we ran a test and found RCTs with no abstract, which has us reluctantly leaning toward automatically including all cites with no abstract in the full article review level. Most of them were comments or letters, but they weren't necessarily coded as such in MEDLINE.
This topic is an emerging one, so many of the citations lack full records. Makes for some novel challenges in trying to honor the intent of the methodology.

Many thanks!

- Patricia

Patricia Anderson, [log in to unmask]">[log in to unmask]
Emerging Technologies Librarian
University of Michigan
http://www.lib.umich.edu/users/pfa