Continuing my thoughts about great sessions from the 2015 EAHIL workshop in Edinburgh (have to get thoughts down quickly because the sugar rush from the Tunnocks Teacakes will only last so long….).
The other session that I found really interesting too was this one – even after a tremendous evening ceilidh dancing with the best librarians Europe (and the world) has to offer in a fabulous location of the National Museum of Scotland.
sessions that blew my mind 1.2
Improving efficiency and confidence in systematic literature searching by Wichor Bramer & Gerdien de Jonge.
Now I have to admit that I’m probably a bit stuck in my ways regarding putting together a literature search strategy, but when it comes to re-running it in other databases, it can be dull and is certainly pretty repetitive. What Wichor and Gerdien say is that their technique can reduce the time it takes to run systematic literature search strategies from average of 16 hours, to 2 or 3 hours tops. Who wouldn’t be interested to hear more about this type of time saving.
The technique is essentially to move away from line by line searching but to make a single line strategy, optimised as much as possible. Here’s the science part: you can then use microsoft word macros to instantly adapt the strategy for the syntax that matches the interface of the next database that you want to use. Wichor’s macros are available here.
So that, for example:
(exp “osteoarthritis, hip”/ or (coxarth* or ((hip*) ADJ3 (osteoarthrit*))).ab,ti.)
which is in Medline via OvidSP syntax will be instantly changed into
(MH “osteoarthritis, hip+” OR (TI (coxarthr* OR ((hip*) N3 (osteoarthrit*))) OR AB (Coxarthrit* or ((hip* N3 (osteoarthrit*)))))
which is the syntax for medlinc via EbscoHost.
Wichor and co have created macros so that you once you’ve created an optimum strategy in one database (they start in Embase for preference), you can easily and quickly translate the strategy for use in Cochrane, Web of Science, Medline etc etc whichever interface that these databases come in. [see slide 29 of his slidedeck for more details]
You just copy the appropriately coded strategy in, as a single line, and start exporting.
Clearly the trick is to a) get an optimal strategy to begin with, and b) to ensure that you are sufficiently comfortable with the code/syntax of your starting database to create a single line strategy to begin with.
Now this is a seriously new trick for an old dog. So much of the time I like the multi-line strategy because I can see where the mistakes are (ahem! not that I *ever* make any….), but I absolutely agree that if you want to adapt/improve it, you need to re-type the entire horrific thing if you want to make it look pretty for publication or to make it vaguely comprehensible for anyone else who is going to review it or make comments. I suppose the important aspect using of the single-line approach is that you need to be really confident and comfortable with the syntax of the databases. Not all librarians are (I suspect), and certainly not all researchers, though of course we can try to explain it.
Wichor made some really good points about trying to encourage ourselves/researchers that they dont’ need all the words of their research question in the strategy.
Some phrases are irrelevant: “pre-eclampsia in pregnant women” – the only important term is pre-eclampsia, since it only occurs during pregnancy, and only women get pregnant.
Sometimes your search terms result in unconscious bias: “eating broccoli to prevent cancer” – the term prevent should not be part of your search, because what if broccoli causes cancer. And your search should not be restricted to broccoli because what if other brassicas have an impact on cancer.
But hopefully these issues would be ironed out during a solid reference interview.
Another top tip was to run a preliminary search, arrange the results by relevance ranking, and see what new terms are included in these most relevant papers. I do a variation on this technique by using key papers already in the possession of the researcher to check for any additional keywords, to test whether my strategy is robust enough (if these key papers are not retrieved, then something is wrong).
There was a lot to take in from this session, and as I freely admit, it’ll be hard to change my old habits, but I’m really looking forward to hearing the results of a piece of work that Wichor says he has planned – doing a direct comparison of his technique against a more traditional (?) approach to the same question, and directly compare number of results, relevance etc. Now if the results of that piece of work are convincing, this old dog might have to learn a new trick.