The search engine will produce about 40 citations. Note that during this workshop I will often refer to the number of citations returned by a particular search. Because the literature is continually growing these numbers will also grow. They were current at the beginning of Nov. 2001

Most of these seem to be on the subject to some degree. We need to make sure the search is reasonably comprehensive before trying to whittle it down.

Page down to one of the citations which seems specially helpful. I chose the one with the highest numbers.

Click on the link to the citations (ie the authors' names). Confirm that the paper is indeed on the subject.

In the language of MeSH or subject of intererst would be described as Clavicle and Fractures, Ununited. The terms /surgery are subheadings which can be attached to any MeSH term. To begin with we don't want to restrict the search only to surgery. So we won't use subheadings.

Copy and paste the following search string into the PubMed search box
Clavicle[MeSH] AND Fractures, Ununited[MeSH]
Then Press GO

The capitalization doesn't matter except for the word AND. Since this is a Boolean logic operator it must be captialized (PubMed's rules not mine!)

WOW! This search produces 227 citations - and you thought 38 were quite enough! What's more, even if you look at the 227th way back in 1965 the papers (by title) are on the subject.

Note, however, that most of the early papers do not have abstracts online.

Although there are some papers in this list which you might not consider on topic it is clear that this is a substantial list and it is as comprehensive as I can make it. So now the task it to reduce it to manageable proportions without losing quality. Since this is an exercise we will be arbitrary and exclude papers without abstracts, papers after 1980 and papers in languages other than English and French

On the PubMed page make sure that the correct search string is entered

Click on the Limits button   This will open the Limits dialog page. It should look (in part) like this.

The only drop-down menu boxes we are interested in are the languages, human or animal, the date box and the tickbox for "only items with abstracts". Change the boxes until they look like this

While still on the Limits page press the Go button and you will get the collection indicated below. Note the yellow highlighted line describing the limits (arrowed)

Now we need the equivalent collection in French. Return to the Limits page and swap the languages French for English. Press Go again. 4 more papers are added.To combine these two lists you must use the History function of PubMed.

On the PubMed page find the History link under the entry box and click on it. What you will see depends on whether you have done the searches in the previous parts of this workshop. For this to work you need to have done the last two (English and French) searches. The other alarming thing that may happen is a message that "this function needs cookies" and your browser won't accept cookies. Because the history expires after 1 hour your clock also has to be (approximately) right for this function to work!. If you know how to enable cookies and/or correct your clock go ahead. Otherwise just skip this section and do the rest of the workshop on the English version of the list.

All being well this is what you will see

To combine or otherwise manipulate the searches you can now use the numbers. In this example #7 OR #6 would combine searches 6 & 7. Because of Boole #7 AND #6 would result in nothing because there is no overlap between the two. (Remember OR to expand AND for the overlap and NOT to exclude - Boole in one line!). In fact I made a different error when I prepared this, not turning off the limits. Since the limits still require French language the combination didn't work!.

  You can Preview the results by clicking on the Preview button rather than Go

As expected there are 72 citations in the combined collection. To see the result of the search from the History function click on the number.

We have seen before that a number of these citations are about congenital pseudarthrosis. To exclude these run a search on #9 NOT congenital and you will get about 51 citations. After some experimentation this was search # 15 for me. Likely it will be another number for you. Using the history number is a matter of convenience. You could use the full search string.

51 citations is still too many for a "short reading list" which is what we started out looking for. There are a number of choices for shortening it

• going through it by hand
• looking for review articles
• using a subheading

Going through by hand

This may be your last stage. To help with the final printout/list of links to abstracts you can get PubMed to print out your selected articles with or without the abstracts.

Set the number of citations per page to 100 so they will all be on one page.

Click on Display

Click on the checkbox of selected articles.

When you have completed your selection click on display again and the selected citations will be assembled. My selection of the articles that showed the variety of surgical treatments used in this situations may be seen by clicking here. Note that the links to the abstracts still work. This type of collection can be printed or copied and pasted into a document.

Looking for Review Articles

This can be done by using the Limits function

Make sure that #9 NOT congenital is still your search string

Open the limits section by clicking on Limits Open the Publication Types Drop Down menu. Select Review.

Click Go and you will see that only one citation out of 51 was indexed by NLM as a review. Unfortunately it is one of the few articles which are mostly off the subject of clavicle non-unions - so this part of the demonstration is unsuccessful. Or rather it demonstrates the limits that can be achieved with this system.  In the end we are dependent on the librarians for much of this sorting.

Using a Subheading

One last attempt to produce a useful short list may be to use the MeSH subheadings. If the papers have been correctly indexed papers about surgical treatment of non-union will be subclassified as Fractures, Ununited/surgery. We can also require that this aspect is the major subject of the paper by using the subheading [majr] (for major). So our final search string for this section would be

#9 AND Fracture, Ununited/surgery[majr] NOT congenital

When I pasted that into the search box I obtained 32 citations. If you don't care about the filters for language etc enter (exactly)

clavicle[mesh] AND Fracture, Ununited/surgery[majr] NOT congenital

I obtained 72 citations of which 36 had abstracts and were published 1980 or later (all languages)

If you wanted to specify complications, risk factors or etiology there are subheadings which will select out those topics

Summary

• You couldn't do this at all without a computer system, That puts the annoyance of having to struggle with the system in perspective. Getting the medical librarian to do it will be great, if he/she can read your mind! Learning to do it and practicing the skill isn't so tough.
• To make a good short list you need to make a good long list and whittle it down
• A comprehensive list is best obtained by defining your subject using MeSH terms
• Narrowing down the list can be done using the limits function, the history function and the MeSH subheadings
• In the end you often have to hand select the most relevant papers
• Don't forget you still have to read them! Abstracts are not enough!

Myles Clough mylesclough@shaw.ca
Clinical Instructor, Department of Orthopaedics, University of British Columbia
B.C. Canada
Feedback, Comments and Questions welcome.