#TwitJC week 3 – some thoughts on literature searching

On Sunday night, I participated in the 3rd installment of Twitter Journal Club (#TwitJC), organized by @twitjournalclub (see also their blog: twitjc.wordpress.com).

Last week and the week before, #TwitJC drew a lot of attention and the discussions were pretty hectic. This week, however, there seemed to be slightly fewer participants and the conversation was thus easier to follow. And it was a very good experience – witness these tweets:

A summary of this week’s discussion will be posted on twitjc.wordpress.com.

For now, I want to take the paper that was discussed and run with it – in a direction that’s a bit off-track. I will use it as a case study in literature searching, highlighting some of the concepts we regularly cover in teaching and conducting searches.

The paper reports the results of CRASH-2, a randomized placebo-controlled trial on the effects of the antifibrinolytic agent tranexamin on trauma patients with significant haemorrhage (full-text with institutional login).

This paper recently won the award for Research Paper of the Year at the 2011 BMJ Awards. As also became obvious from Sunday night’s discussion, this is an important paper that should not be overlooked when searching the literature for this subject.

So… will this paper show up in a relevant literature search?

Three-part question
A literature search usually starts with a clinical question, preferably consisting of 3 components: domain, determinant and outcome (alternatively, the PICO-format can be used). In this case, a relevant question could be:

‘Does an antifibrinolytic agent reduce mortality in trauma patients with significant haemorrhage?’

Domain: trauma patients with significant haemorrhage
Determinant: antifibrinolytic agent
Outcome: mortality

Of course, this is only one way of many to formulate a question for which this paper would be expected to turn up in the search results. I purposely broadened the determinant a bit (not using tranexaminin specifically), as it seems possible people might search for the effects of a type of drug, rather than this drug specifically. Other outcomes could also be considered relevant – more on that later.

Textwords or thesaurus?
The components of the three-part question can be searched as textwords (often confined to title/abstract words only) or as thesaurus terms (like MeSH in PubMed or EMTREE in Embase). Both have their advantages and we feel they really should be used in combination. Let’s have a look at how that would work out for our search question/the CRASH-2-paper.

In the title and abstract of this paper, the following words are used to describe the three components of our search question:

Domain: trauma, haemorrhage, bleeding

• trauma seems pretty straightforward. Note that searching for hospital setting instead (e.g. emergency or ER) would have resulted in missing this paper.
• Note that searching for spelling and word variants as hemorrhage or haemorrhaging would not have yielded the paper.

Determinant: tranexamic acid

• Note that searching for ‘antifibrinolytic‘ as title/abstract word would not have yielded this paper.

Outcome: death, mortality

• I almost missed the occurrence of mortality in the abstract, but it is mentioned in the results section.
  

The above example illustrates the importance of using synonyms, including spelling and word variants, in constructing a literature search using title/abstract words.

What about searching thesaurus terms? Read more!