MeSH Terms and the New PubMed
Yesterday I attended an online webinar focused on the changes in the new PubMed. Holly Ann Burt from the GMR was the instructor and she was great. (I will be linking to the recording when it is available.) However at one point in time during the session one librarian became rather frustrated with how the designers made it that much more difficult to search PubMed using MeSH. Her complaint was that doctors were not going to know where to click and how to search using MeSH within the new version of PubMed. My first thought was that most doctors weren’t using MeSH prior to the redesign anyway. Sad, but I think pretty much true.
I must confess, I am not a huge PubMed user. My MoC (Medline of Choice) is through Ovid. I really love Ovid’s mapping in Advance Search, I think that is one of its strongest features. The mapping allows me to quickly type in a keyword in the search box and be presented with the correct MeSH term. The mapping forces you to at least see a MeSH term, which is one of the strongest features of MEDLINE. Does it work all the time? No sometimes you can get some wonky results when you have an odd term (or when you unknowingly type it incorrectly). But even in that case I like it because it forces me to re-examine my terms and strategies early in the search process. I am not blindly searching via keyword and either missing vital information or retrieving everything under the sun.
I have always felt PubMed hides the MeSH in the closet. It really never had very good front end mapping. If you really wanted to be sure you were searching using MeSH terms you had to go to the MeSH database, search it and then ADD it to your PubMed search. Clunky. What doctor wants to do that? Many don’t. Heck I don’t want to do that and I am a librarian, but I do it when I search PubMed. Of course over the years PubMed has made various improvements on the keyword searching. In this latest version, PubMed uses Automatic Term Mapping (all behind the scenes) to search for the keyword as a MeSH term, subheading, Publications type, Pharmacologic action term, and All Fields. Personally I think this produces a ton of results with a lot of extra junk citations. I think it is just one step better than Google.
There is one way to stop the citation deluge, and that is to have sound search strategies using MeSH. But through out history, PubMed has done such an outstanding job of hiding the MeSH that it is no wonder that the average user doesn’t know or care about MeSH. Must I remind you of Ana Kushnir’s blog entry, I Am Not Yelling. Not Out Loud, about why she hates PubMed. She had no idea of really how to use PubMed and of MeSH terms. She is not alone. There are tons of doctors, nurses, medical students, etc. who do a Google search on PubMed and they think they are getting the right stuff.
I believe that PubMed’s inability or refusal to actively map people’s keyword searches to MeSH terms out front glaring at you on the page (like Ovid), is reason why people don’t search MeSH in PubMed. In Ovid I am forced to pick a MeSH term, in PubMed I am blindly given results. Doctor’s, nurses, and medical students didn’t go through school speaking or using MeSH, all the more reason to bring it out of the forefront of the search and force them to see it and choose it when searching a database that indexes using it! Yet PubMed designers chose to hide it behind a curtain. While we librarains were doing are darndest Toto impersonation to reveal the truth behind curtain, there was no way our message could be heard by everybody. Getting doctors and researchers to search via MeSH was a losing battle.
Now with the latest redesign the battle is lost. The ability to search MeSH is even more clunky, the MeSH terms are minimized (hidden) within the abstract results and the details box in Advanced Search will be disappearing. PubMed has made their product the Google database of medical literature. Congratulations. Searchers like Google, it is the number one search engine. I am just not sure I want my medical database to end up to be that. Before you know it NCBI may wonder why we are even bothering with this whole MeSH thing anyway, why index when everybody is doing keyword searching anyway? Everybody is Googling. Now let’s be clear, I never heard anybody say that nor do I have any super secret handshake knowledge of that happening. Quite frankly I bet those in charge of PubMed right now would say we would never think of ending the MeSH indexing of articles. But, that is right now. What happens 10, 20, 30 years from now? PubMed’s hiding the MeSH in the background was the first, but crucial step in making MeSH irrelevant to the searcher.
Interesting take! I agree that the way Ovid forces you to use MeSH is a good way to introduce the concept of controlled vocabulary, and that it’s easier to teach people to use Medline in that interface.
This only works if you default to Advanced search, though, and I’ve run into quite a few people who would switch to the Basic search and get pretty much the same keyword-based results you do with PubMed’s front page search bar.
When I teach PubMed, I just say straight out “you can use this Google-style search box and get Google-style superfluous results, OR you can click here and use MeSH to focus your search.”
Does anyone actually remember the classes and use PubMed this way? We can’t guarantee it, but we can hope.
Again, I do agree that Ovid’s design makes it easier to teach some of the advanced search concepts, and PubMed is a little more obscure, but it’s certainly possible to get poor results with Ovid as well.
I also find that Ovid spreads information all over the place (scope note here, MeSH tree here, subheadings on another page), where on PubMed it’s all on one page, which I like. But then, I’ve always kind of preferred PubMed. Different strokes for different folks, as they say.
I wonder if we can draw any conclusion about cat people/dog people and Ovid people/PubMed people? Cats for me, if we’re going to start counting.
I wish they would go all the way to making to making the search google-like. Why can’t a search return what I am looking for? Why should I have to look through thousands of irrelevant results when doing a search?
I totally agree — I miss my Ovid MEDLINE. A lot of people like PubMed because they perceive it as faster, but that says nothing about the results they’re getting. Ovid makes you think about your search. I’ve spent a lot of time teaching people to click the Details tab, so they can at least make sure they’re searching what they want to search (I always point out that PubMed thinks CT means contraindications, not computed tomography, but you need to look at Details to see that), but now Details is even harder to find.
I completely agree that no system is perfect. Both Ovid and PubMed could do some things better. I just think that PubMed’s way of making searching easier by hiding the MeSH from normal people is unintentionally contributing to the misguided perception that MeSH isn’t important. Don’t get wrong, I’m all for easier searching. I just don’t agree with PubMed designers methods.
PubMed and Ovid are like driving a stick and automatic. You may know how to drive both but you will always prefer one over the other.
I have never been a big PubMedder, I use it when I have to. I am much more efficient in Ovid. Perhaps it is because I was trained on it for my grad assistantship job in library school, or perhaps it is because I am a dog person.
I use Ovid Medline for work (and by preference) and I Google. But let me be clear-I google for *known items* and when I want *one or two* useful resources. I don’t undertake comprehensive searches in Google. This is what the PubMed people need to understand-if they make their product mirror Google, they will alienate the people like me who shy away from using Google (and keyword-based search engines) for comprehensive searches.
I’ve never been too pleased with PubMed as a resource. Even when I need to get a PubMed URL, I will put the title and sometimes journal of the article into Google as this gives me a nice clean PubMed URL. Not the long, ugly within-PubMed URL that one is given. Why can I not get a nice http://www.ncbi.nlm.nih.gov/pubmed/PMID here] URL when I click on an article from a PubMed search results page?
I must couch this in ‘I have barely touched new PubMed’ but am fairly familiar with the old one.
Great post, Krafty. I feel you pretty much speak for me on this.
Eliminating MeSH may not be such an unlikely idea – I do recall of some discussion in the past year or so about moving to automatic indexing, because indexing takes time and money.
I, too, prefer my Ovid MEDLINE. I occasionally use PubMed to explore a new concept, or occasionally Basic Ovid for the same purpose. But I was raised on MeSH – my husband says I dream in MeSH (after nearly 40 years of seaching, he’s probably right). For a quick and dirty search, PubMed may be adequate – but I spend most of my time running complex, multidatabase systematic reviews.
I’m not so sure that I agree with the premise that MeSH is more deeply buried in the new PubMed interface than in the old — or that it is harder to search.
The PubMed search box has a drop down arrow adjacent to the word PubMed — a quick way to search MeSH (and other NCBI databases) is to simply use that to change the database you’re searching. Now, I realize that some people might argue it’s not intuitive. But then, I’ve heard from plenty of clients of our library that the Ovid MEDLINE way isn’t entirely intuitive, either.
The new interface with its single abstract option makes finding the MeSH terms much easier. It certainly wasn’t intuitive that the “Citation” format contained the indexing terms. Now, it’s simply a click to see the MeSH. And, if you are a user that wants to see the terms all the time, you can set it to be open all the time via your My NCBI account.
And, from what I understand, we’ll be able to leave the Details box open on the search results page — which I think will be fabulous.
Finally, I just have to comment that if all the citations in PubMed were indexed with MeSH (as is the case in Ovid MEDLINE), then I would agree that it would be preferable to do your searching with MeSH. But, that’s not the case. There are a lot of non-indexed citations in PubMed — including the newest citations in the database. That’s why the Automatic Term Mapping uses both the MeSH with the [All Fields] search. And, we’re coming up to that time of year when the non-indexed citations start piling up (the end of the year processing time).
I am a PubMed user. Have been for years. I’m also an Ovid MEDLINE user. I like them both. My own personal suspicion is that whatever people learn to use first or have used the longest – that’s what continues to be their favorite. And that’s fine.
But, PubMed is really quite phenomenal, in my opinion. And, since it is freely available on the web, it seems to me that it’s pretty important.
When I wrote an article in August – MeSH is a Buried Treasure – I had hopes that the new version of PubMed would address problems with MeSH that were discussed at the annual Canadian Health Libraries Association conference, as reported by Mark Rabnett. But alas, it looks like things went the opposite direction, for un-burying MeSH.
The idea of doing away completely with MeSH is unthinkable, for many reasons — Start with Explosions, one of the primary pillars upon which MeSH is built. Even though the importance of MeSH is hidden in using PubMed, explosions are there in the background, making it work.
I agree that Ovid is much a much better tool for learning MeSH. Does anyone have a handle on which medical/health sciences libraries continue to offer Ovid as well as PubMed?
I understand your opinions on the MeSH within PubMed. However, the Details box is going away for regular users. According to the technical bulletin, “When we initially transition to the redesigned version, Search Details will display for all users with search results (see Figure 1). The current plan is that at a later date this will change and users must use My NCBI to select the option to display Search Details with search results.” Right now the Details box displays to any user in Advanced Search. In the future only users searching via My NCBI will have the option to turn on the Details box. I honestly don’t think many users use MyNCBI. I could be wrong but that is my feeling. Your power user doctor or researcher, yes they have MyNCBI accounts. Your average doctor or researcher most likely doesn’t. They hop on to the system and search just like they search using Google. I think the Details box should stay as is and should not be removed. It is out of the way, but is helpful. Why the developers are playing “now you see it now you don’t” when it is a valuable tool that is not in the way of the overal search screen.
The instances you describe are perfect reasons to go “off MeSH” and use keywords. But my experiences and observations lead me to believe that many more users are going “off MeSH” and searching keywords when they shouldn’t. The people who know that not everything is indexed in MeSH are usually librarians (or possibly super power users) and they already know when and why somebody should search keyword and when to search using MeSH. The average user doesn’t know this and more often than not does a crummy keyword Google search and gets a ton of junk results where they have sift through and separate the wheat from the chaff. And the typical user won’t go through that many to do that, they will pick the first one or two “decent” ones. That is Google searching and that is what most people do on the computer and in PubMed. PubMed just fostered it by leaving MeSH in the background and unknow to the average user.
Ovid was my MoC for over 15 years and it is what I tried to teach. Two things pushed me to swith to PubMed.
1. PubMed’s related articles search consitenty brought up relevant citations that me Ovid Medline search missed. I was first trained by NLM in searching Medline in 1979 so I have literally done thousnds of searches but I still missed good citations that the related articles search found.
2. The longer I have been doing this the more I have been moving towards the KISS (keep it simle, stupid) approach. In trying to teach Ovid I have known if I can just get my students to have a basic understanding of Mesh and Ovid’s mapping of it that they could perform a much more focused search in Oivd. I found that I could not achieve this in the majority of the students I was teaching. I found that with there familiarity with Google that I had much more luch teaching them to do their Google-like search in PubMed and then using related articles on the few good articles they found.
By the way I also was very used to the old Pubmed and am adjusting to the new interface but expect to be used to it within the next month or two. We can teach those who know and love Mesh how to use it in the interface but I think that will be an increasingly smaller subset of suers.
[...] librarian has responded to the piece with enthusiasm, suggesting she had become fed up with PubMed. The Krafty Librarian has sounded off about PubMed recently, to say while PubMed had always hidden the controlled vocabulary, the MeSH, it had become [...]
I’ll join the chorus. I really don’t like PubMed’s de-emphasizing of MeSH. I have too many docs who send me PubMed citations for a bunch of articles from the 80s that they need on ILL when there are newer, better articles out there. I notice often that they all have the same phrase in the title but a MeSH search brings up much better articles under newer, more focused terminology. In fact I often will also send along a newer article along with the older stuff. If you don’t know much about MeSH and wind up using keywords you get inexact results or get buried in citations. That’s why I really, really like Ovid’s emphasis on mapping to MeSH. There are so many cases when it has turned up such better results. I know Ovid is working very hard on Googlizing their searches–you know, one easy box, put in some junk and get some junk and a few nuggets of helpful stuff back–that I know it is where everyone is going. Fast isn’t really fast when you have to page through tons of stuff for a few good articles. You save thinking time at the beginning but waste more time at the end by going through a bunches of off topic things.