3rd Call for Submissions for “Medical Information Matters”: Tools for Searching the Biomedical Literature

8 05 2011

It takes some doing to breathe life into Medical Information Matters” (blog carnival about medical  information).
A month ago I wrote a 2nd call for submissions post for this blog carnival. Unfortunately the next host, Martin Fenner, didn’t have time to finish a blog post and has come up with a new (interesting) variation on the theme “A Wish list for better medical information”.

Martin asks you to philosophize, blog and/or comment about “Tools for Searching the Biomedical Literature.

You can base your contribution on a recent (editable) survey of 28 different PubMed derivative tools by Zhiyong Lu (NCBI) [1].

Thus, write your thoughts on the various PubMed derivative tools mentioned here or write about your own favorite 3rd party PubMed tool (included or not).

For details, see Martin’s blog post announcing this upcoming edition. The Blog Carnival FAQs are here.

And if you don’t have time to write about this topic, you may still find the survey useful, as well as the views of others on this topic. So check out Martin’s blog Gobbledygook once in a while to see if the blog edition has been posted.

Note [1]: If you have already submitted a post to the carnival, or would like to write about another theme, we will take care that your post (if relevant)  will be included in this or the next edition. You can always submit here.

Note [2]: Would you like to host “Medical Information Matters” at your blog? Please comment here or write to: laika dot spoetnik at gmail dot com. We need hosts for June, July, August and September (submission deadline first Saturday of every month, posting on the next Tuesday)

  1. Lu Z. PubMed and beyond: a survey of web tools for searching biomedical literature. Database. 2011 Jan;2011. doi: http://dx.doi.org/10.1093/database/baq036
Advertisements




Advanced Neuritis in PubMed

8 03 2009

pubmed-logoAlmost a year ago (June 2008) I discussed PubMed’s Advanced Search Beta in a series entitled PubMed: Past, Present and Future. At that time I was not particularly impressed by disliked Advanced Search Beta and I still do.

November last year some of its features have improved: like the addition of a Clear Button, Focused Queries, providing links to the Clinical Queries and Special Queries pages, and the author/journal search has been extended with optional fields so that it looks more like the valuable Single Citation Mapper in the blue side bar of the Basic PubMed page. And there is a link to the MeSH-database (see NLM Technical Bulletin November 2008).
Although these are real improvements, the links to the Queries and to the MeSH database are inconspicuous, at the end of the page below all kind of limits. My major objections to the Advanced Search is that people are more inclined to narrow their search by using as many limits as possible (because these are so prominently present) and that the MeSH cannot be easily looked up and/or are wrongly translated. Previously I gave some examples, where lung cancer[mesh] was searched, whereas the MeSH is lung neoplasms, or where recurrent pregnancy loss[MeSH] returns no result, because the term is habitual abortion (see previous post).

I avoid Advanced Search as long as I can, but the problem is, the library-users don’t. They like to experiment, especially when they consider themselves as advanced searchers.

Last month a Neurologist asked me if I could check his search for a Diagnostic Systematic Review. A search for a Systematic Review should be comprehensive and thus contain both MeSH-terms (Controlled terms of MEDLINE) and free text words (tw).

He was a resident in Neurology for 5 years and knew how to search PubMed.

Below is the first part of his search.

((((((((motor neuropathy[MeSH Terms] OR motor neuron[tw] OR motor neuropathy[tw]) OR multifocal motor neuropathy[tw]) OR demyelinating neuropathy[tw]) OR multifocal demyelinating motor neuropathy[tw]) OR neuropathy[tw]) OR neuropathies[tw]) AND (((((((((((((((((((((((((…..

Grosso modo it looked all right and well structured. The awful number of brackets is often seen when people combine directly in PubMed (although I was already glad there were no brackets around every single word and he didn’t copy the entire translation from the Details-Tab). And some terms were superfluous: you don’t have to search for multiword terms with neuropathy (i.e. motor neuropathy) because these are already found by searching neuropathy.

So we made the search simpler, like this:

(motor neuropathy[MeSH Terms] OR motor neuron[tw] OR neuropathy[tw]) OR neuropathies[tw]) AND (………

Just to be sure I asked him: “Do you mind if we check the MeSH? Motor Neuropathy looks just fine, but you never know.”

To my surprise, typing motor neuropathy in the MeSH search bar yielded 4 suggestions, none of which was motor neuropathy.

pubmed-motor-neuropathy-mesh-1

The most suitable term appeared Neuritis. When bringing this MeSH-term to PubMed we got exactly the same number of hits as with Motor Neuropathy. Mere coincidence? No, the hits weren’t any different (#1 NOT #4 giving zero results).

pubmed-motor-neuropathy-search-1

Looking Up the Query Translation under the Details Tab confirmed my suspicion: motor neuropathy[mesh] was translated as “neuritis”[MeSH]. This is disturbing. Not only doesn’t there exist any MeSH specific for motor neuropathy, people are put on the wrong track since it looks like motor  neuropathy[mesh] is recognized as such.

pubmed-motor-neuropathy-search-1b-details

Then it came to my mind that I had seen a similar odd “translation” when using PubMed Advanced Search (see above). And I asked him: “Did you by any chance use the Advanced Search”, which he did.

To check this I searched in Advanced Search for the MeSH: motor neuropathy. And, yes indeed, the motor neuropathy[MeSH] was searched so it seemed. (in reality we now know: Neuritis was searched). The difference with searching the MeSH database is that here I know that I search for neuritis (I choose to), whereas the Advanced search misleads me by suggesting I’m searching for motor neuropathy.

pubmed-motor-neuropathy-2

pubmed-motor-neuropathy-2a

Why do I bother? Why don’t I just use motor neuropathy[mesh]. First because I don’t get what I want: I get neuritis[mesh] not neuropathy! Second, and most important, because it is not the most appropriate MeSH-term.

To find more appropriate MeSH I use a trick. I look for MeSH-terms assigned to articles, having motor neuropathy in their title, assuming that motor neuropathy is an important aspect of those papers.

Although you can look up MeSH assigned to each individual citation in PubMed in the citation display format, it takes a lot of time to go through the papers one at a time. Therefore I rather use GoPubMed or even better PubReminer for this purpose, because these give you a frequency list of the MeSH assigned.

Of the 379 hits found in GoPubMed, 219 were categorized as Motor Neuron Disease, 153 as Demyelinating Diseases and 145 as Polyneuropathies. These categories are MeSH term you can use for your search.

gopubmed-neuropathy

Similarly of the 380 references found in PubReminer, many papers were indexed with Motor Neuron Disease, Demyelinating Diseases, Polyneuropathies, peripheral nervous system diseases and motor neuron.

(Below are the number of papers, indexed with the indicated MESH in PubReminer; PubReminer shows the subheading coupled to the MeSH)

  • 65 Motor Neuron Disease/diagnosis
  • 32 Motor Neurons/physiology
  • 26 Demyelinating Diseases/diagnosis
  • 16 Peripheral nervous system diseases/diagnosis
  • 8 Polyneuropathies

Using this approach we were able to set up a more complete search in PubMed. Remember it was the neurologist’s purpose to to an exhaustive search, for a less exhaustive search we would have only used motor neuropath* and perhaps motor neuron disease[mesh].

How different is it when you use the OVID interface for searching MEDLINE.

When you type Motor Neuropathy, several MeSH are suggested, many of which are useful:

ovid-motor-neuropathy-1

When you click on Motor Neuron Disease, you see the hierarchal context and can choose which terms you would like to add. We choose not to explode Motor Neuron Disease, but only include one narrow term in our search: amyotrophic lateral sclerosis.

ovid-motor-neuropathy-2

Finally the first part of the search in MEDLINE (OVID) looked like this. It is rather broad but the second part of the search (not shown) puts it into context.

1. motor neuron disease/ or amyotrophic lateral sclerosis/
2. exp Motor Neurons/
3. Demyelinating Diseases/
4. neuromuscular diseases/ or peripheral nervous system diseases/ or neuritis/ or polyneuropathies/
5. (neuropathy or neuropathies).tw.
6. motor neuron*.tw.
7. or/1-6

OVID MEDLINE was easier to use, you get what you see (and want) and the search is easier to save and edit. Furthermore the entire MEDLINE search can be easily transformed to a search in EMBASE: just replace MESH by EMBASE keywords.

I’m not happy with the Advanced Search for reasons explained above. I don’t find the altered mapping and citation sensor a success either. I don’t like that they removed the blue side bar in some display formats. And I’m really getting depressed by NLM’s announcement (November 2008):

PubMed Advanced Search will soon no longer be a beta site. It is now the place to go to use features such as field searching and limits. In the near future the tabs for Limits, Preview/Index, History, Clipboard, and Details will be removed from the basic PubMed pages. History, Limits, Index of Fields, and a link to Details are available from the Advanced Search screen. A link for the Clipboard appears to the right of the search box on the PubMed screen when the Clipboard has content.

If I understand it correctly this means that Pubmed Advanced Search is taking over the basic search.

It looks that my original idea was right: PubMed is going for the mass, it is going for the Google-like quick searches by people that don’t know much about MEDLINE and don’t want to learn it. But you have to know some basic principles to get the most out of subject searching. It is such a pity, that PubMed tries to copy its clones, whereas it holds all the trumps. No other 3rd party tools offer the same possibilities that PubMed offers, although they are more suitable for certain purposes (see examples of GoPubMed and PubReMiner above).

At least make two interfaces, one for the beginner (the present Advanced Search) and one for librarians and other people doing subject searches.

But I don’t have the illusion that the people of PubMed/NLM will listen to me and I’m not going to contact them for a 3rd time. PubMed’s route is determined, I guess.





Finding assigned MeSH terms and more: PubReMiner

24 09 2008

Generally when searching PubMed I use both MeSH and textwords. If you already have some nice articles, either by performing a quick and dirty search or looking at the Related Articles or your colleague gave you one or two, then you can find the MeSH assigned to these papers by looking in citation format (see Fig). However going through a set of articles looking at all indexed terms takes quite some time and one doesn’t easily get an overview of the overall frequency of MeSH in a set of records.

Therefore, Rachel Walden, asked first at Twitter and then at David Rothman’ site (see here):

“What I’d like to do is to be able to enter the PMIDs of several citations and have the tool search MEDLINE via PubMed for the assigned MeSH terms, and return a single list of the terms used by any of the entered citations with a measurement of frequency. For example, if I input PMIDs 16234728, 15674923, and 17443536, the tool would return results telling me that 100% or 3 of 3 use the term “Catheters, Indwelling”, 2 of 3 use “Time Factors,” 1 of the 3 uses “Urination Disorders,” and so on. Although this example uses 3 PMIDs, I’d like to be able to input at least 10, just based on personal experience.”
(PMID is the unique PubMed-identifier, by searching for the PMID you get one specific record.)

The suggestions made by several people were summarized by David in another post (see here). The following 3rd-party PubMed/MEDLINE tools seemed most promising:

Both give useful results. The layout- is very user-friendly.

I tried my own sets of 20 PMIDs (from a systematic review search on “predictive models for in vitro fertilization”) and http://www.docmobi gave this result:


This was the result when I selected: “primary terms only“. If I selected “include secondary terms” I would find female and pregnancy as well, but at rates of 175%. First I didn’t understand, but than I realized that pregnancy occurs as “Pregnancy” and as “Pregnancy Rate”, whereas female occurs as “Female” and as “Infertility, Female”. Therefore Pregnancy and Female can occur more than once as single words in one record (thus accounting for >100%). It is odd however, that important MeSH like Pregnancy and Female (which are really check tags, MeSH that should be assigned to each article that is about pregnancy or females) are not included in the primary list.

Although useful and nicely presented I still not find this ideal:

  • Check tags are not in the list (with “primary terms only”)
  • Publication types and substance names are in neither list.
  • Subheadings are not included. For words not captured by a single MeSH, but by a combination of MeSH and subheadings this is especially important.
    For instance, there is no MeSH for EGFR-inhibitors, you have to use:
    Receptor, Epidermal Growth Factor/antagonists & inhibitors
  • Textwords are not included (not on Rachel’s wish list but on mine)
  • Personally I find the list so simple that I would have find the terms immediately myself.

Coincidentally I found another 3rd party tool which does the job much better (I think): (PubMed) PubReminer, produced by Jan Koster at our hospital (AMC, Amsterdam). I looked at it, because my colleagues and I are going to discuss it today.

This is the procedure I would recommend to find assigned MeSH and more, starting from PubMed (which is not absolutely required, because you can also search directly in PubReMiner)

  • Collect the PMID’s of the papers you want to analyze.
  • If you have selected the papers in PubMed (on the Clipboard, in My collections or in your search set) then Set the Display Tab on: UI List (Unique identifiers), and export these PMID to a textfile by choosing “to Text” from the Send to button. You get a simple list of PMID’s that you can copy/paste to PubReMiner. (not required, but handy)

  • Go to PubReminer, paste the PMID’s in the search box as one string.
  • Click: “Start PubReminer” (if you enter a search you may wish to apply limits)

  • You see the results, in the following columns: Year, Author, Journal, (Text)Word, MeSH, Substance, Country.

  • Afterwards you can decide
    – Whether the words are searched in title only, in Title and Abstract or also in MeSH and RN.
    – which columns are displayed
    – whether similar words are merged or not.
  • You can use selected terms to build up your search in PubReMiner and/or
  • You can export the terms as a text-file (!)
  • And, by the way, you can download a plugin for IE or Firefox (Fig. right)

This tool is not very intuitive, but the result is quite ideal.

You get both a list of MeSH plus their subheadings AND a list of textwords (and substance names).

There are very useful terms in here, for instance logistic models[mesh]. As textwords I should use ‘logistic regression’ or ‘multivariate analysis’. It is only a pity that terms are given individually and spread: the context is lost (‘analysis’ may for instance be too broad, it is only useful in combination with ‘multivariate’ or ‘regression’).
With respect to the MeSH only individual terms are given. ‘Pregnancy’ is mentioned 19 out of 20 times in the MeSH-list. Does this mean I’m missing one paper by searching “Pregnancy”[MesH]? No, because that paper is indexed with the narrower term “Pregnancy Rate”. (and you will find it by exploding Pregnancy).
So the context and the hierarchy are lost here as well.

But it is about as good as one can get.

I’m gonna use this tool (as an adjunct at least). Seems that it has some other potentials as well: look up genes, find the research interest of an author, determine the journal to submit your work to. Well I probably learn that in a few hours. Perhaps I will come back to it later.