PubMed Workshop Page 3

Defining your subject using MeSH

One gets the feeling that somehow the definition of the subject of interest has not been accurate enough to pick up all the papers that Medline has. This is why NLM introduced the concept of standardised terms to define medical subjects - the Medical Subject Heading (MeSH) system.

On the PubMed Page find the link to MeSH Browser (MeSH Browser) Enter the term "distal radius fracture" in the box marked "Search for"

You will find it returns "No exact match for your term was found. Please select from the following possibilities: " and provides a large number of possible terms. The closest match amongst these is "Colles' Fracture". Highlight this and click on the box which says "Browse this term" This will show you the path to the term :-

MeSH Tree Location

All MeSH Categories
      Diseases Category
            Injuries, Poisonings, and Occupational Diseases
                  Wounds and Injuries
                        Fractures
                              Radius Fractures
                                    Colles' Fracture
What this means is that Colles' Fracture is a subset of the subject "Radius Fractures" which in turn is part of "Fractures" and so on. More importantly if you use the term Colles' Fracture as your search string AND define it as a MeSH term your search will be maximally sensitive and specific.

Return to the PubMed page and enter (exactly) "Colles' fracture [MESH]" as the search string. The number of citations returned Sept 8th 2001 was 522 all of which were about wrist fractures or had the term Colles Fracture as part of the full citation. If you use the search term "Colles Fracture" without the qualifier "[MESH]" I obtained 607 citations. A brief look suggested that all the citations of the "MESH" search were included in the second search. You can frequently define the search as a MeSH term from the MESH browser.

*Key Point* Using the MeSH term and defining it as a MESH term makes the search very specific. Using the same term but NOT defining it as a MESH term makes it more sensitive.

[MeSH] and [MAJR]: two important subheadings

If you use the subheading [MESH] (or [mesh] or [MeSH] !) with a search string that is an exact MeSH term the result will be maximally sensitive. If, however, you use the subheading [MAJR] which stands for major topic, the collection will include only those papers in which the main topic was Colles Fractures. In the following table comparison is made between the results of various different search strings. For the purposes of discussing specificity I have arbitrarily defined the subject as distal radious fractures, including epidemiology, treatment and complications. "Off subject" papers thus are mostly on osteoporosis as a main topic with distal radius fractures being secondary to the study.

Table 1 Results of PubMed searches using different search strings

Using a MeSH term may improve the quality of your search. However, for a subject like Orthopaedics the MeSH classification is very coarse and will result in large numbers of papers to select from. Furthermore it is clear from the table above that using a MeSH term may not give you the most sensitive collection. The largest and best collection arose out of the combination of "Colles fracture/surgery" which is a MeSH term and "Distal radius fracture/surgery" which is not. 

This table displays the overlap between the collections taking the first 20 fully indexed citations in each collection and comparing them to each of the other full collections.

Table 2 Overlap between the collections of citations using diferent search strings

There were a total of 7 citations in the 39 citations examined which were represented in every collection. There is a remarkable lack of overlap between citations found by entering "distal radius fracture" and "Colles' Fracture" 19/20 of the former collection and 15/20 of the latter did not overlap. This can only mean that the librarians entering the information into the Medline database do not recognize that Colles Fracture and Distal Radius Fracture are synonymous (in this context)

If you enter "Colles' fracture OR distal radius fracture" you get 1551 citations. [cf table 1 1079 citations for distal radius fracture and 607 for Colles' fracture] If you add Smith Barton Fracture to that you get 1556 citations. So one is always left with a nagging doubt that you have found every citation on the subject in the Medline database. It is likely that a recent review about the subject would yield references which are not in the list - and this is of course a very necessary step to take after you have formulated your basic list. Nevertheless we have probably come as far as we can in creating a sensitive search strategy. Unless you are planning to write the definitive review on distal radius fractures 1556 citations to read is quite enough! We should turn to strategies for refining the search and producing specificity - always remembering that this may be at the cost of sensitivity.

However, the PubMed site can be quite helpful in filtering the papers.

Return to the MeSH Browser and re-enter Colles' Fracture. You will see the display

Colles' Fracture[Detailed display]

Fracture of the lower end of the radius in which the lower fragment is displaced posteriorly.

Click on the link [Detailed display] and you will obtain the display :-

Colles' Fracture[Brief display]

Fracture of the lower end of the radius in which the lower fragment is displaced posteriorly.

Year introduced: 1991

This filter is self-explanatory. For instance if you click on the box "Restrict Search to Major Topic headings only" the only citations which will come up will be those in which Colles Fracture is the major topic. This is the same as adding the [MAJR] subheading as we did earlier. If you do this you will get about 432 citations most of which are on the subject (See #5 in Table 1 above).

Even such oddities as

Bloch H.           [See Related Articles]

The Asclepiads of Dublin: a moment in Ireland's medical history.
South Med J. 1990 Jun;83(6):664-8. No abstract available.

turn out to be a historical account including Colles' original description!

Return to the [Detailed Display]  page and check the box   surgery

When you click on the "ADD" box you will see that the search is now for

Current Query:

This yielded papers with high specificity to the topic (but to increase sensitivity you need to add a similar term for distal radius fractures). There is a tutorial on Search Field Descriptions on the PubMed site. This is well worth going through if you have an important and complex query and don't have a medical librarian to help you.

In these days of Evidence Based Medicine a premium is placed on evidence supported by randomized controlled trials (RCTs). PubMed reflects this trend by making it easy to find papers on your subject which are RCTs.

From the PubMed page click on the link to Clinical Queries You will see a page like this

Clinical Queries using Research Methodology Filters

This specialized search is intended for clinicians and has built-in search "filters" based largely upon Haynes RB et al.. Four study categories--therapy, diagnosis, etiology, prognosis--are provided, and you may indicate whether you wish your search to be more sensitive (i.e., include most relevant articles but probably including some less relevant ones) or more specific (i.e. including mostly relevant articles but probably omit a few). See this table for details regarding filtering.

NOTE: If you want to retrieve everything on a subject area, you should not use this page. The objective of filtering is to reduce the retrieval to articles that report research conducted with specific methodologies, and retrieval will be greatly reduced.

Enter the subject "Colles' Fracture MAJR" and then click on the radio buttons "therapy" and "specificity". (Because these belong to different groups they do not cancel each other). In theory you have now asked for high quality papers on the treatment of Colles' Fractures. In fact, as you will see when you click the search button you get a collection of 14 papers on RCTs of medical treatment associated with Colles' fractures - even though there are RCTs of surgical treatment. This is because the search string generated by this section assumes a search for medical rather than surgical treatment.

Note that the search string generated by these moves is "(colles' fracture[All Fields]) AND ((double [WORD] AND blind* [WORD]) OR placebo [WORD])". Since it is highly unlikely that any RCT in this surgical field will be double blind, or involve placebo surgery, it is unsurprising (even though annoying) that the surgical RCTs are missed out.

Now return to the PubMed Clinical Queries page and again enter "Colles' fracture/surgery OR distal radius fracture/surgery" as the search string, click on treatment but this time click on sensitivity rather than specificity. The search string generated is -(colles fracture/surgery OR distal radius fracture/surgery) AND (randomized controlled trial [PTYP] OR drug therapy [SH] OR therapeutic use [SH:NOEXP] OR random* [WORD])

This leads to a collection of 49 papers, mostly about clinical trials of surgical treatment of Colles' fracture. The one that caught my eye was

 

Kapoor H, Agarwal A, Dhaon BK.

Related Articles

Displaced intra-articular fractures of distal radius: a comparative evaluation of results following closed reduction, external fixation and open reduction with internal fixation.
Injury. 2000 Mar;31(2):75-9.

Clicking on Related Articles for this paper produced 500 citations, most of the first 50 are relevant to the issue.

*Key Point* Carrying out this kind of search takes persistence and practice and refinement of your search. Computerised systems are inherently dumb and no good at reading your mind. Your medical librarian is trained to do this for you and may be the best resource if the quality of your search is an important issue. However, doing a rapid and competent literature search should be part of every orthopaedic surgeon's skill set. So you should practice.

Continue with PubMed Workshop - using Boolean terms to customize your search

Myles Clough MD, FRCSC, Clinical Instructor, Dept. of Orthopaedics,
University of British Columbia
Nov.2001
Email comments most welcome mylesclough@shaw.ca