We’re going to give you a taster so that you understand how the searching element of the review works – how we go about developing a search at Cochrane. Anyone who goes on to do a Cochrane systematic review will get full support from the Editorial Base – developing the search strategy, running the electronic searches, advising on content for the search methods section
The presentation today will cover the following.
Should be searching more than one database It should be a sensitive search to identify all the available studies – review could be biased if an RCT is missed A well reported search strategy means that others can replicate your study – Cochrane advice is to make sure they appear AS RUN in the appendix of your review
MEDLINE via PubMed – one of the biggest biomedical databases, based in US. Contains over 21 million citations. EMBASE via OVID at NYU – European database with pharmacological focus, also now a good source for conference proceedings Access to the Cochrane Library is available at NYU – via the library. The Cochrane Library is where all the Cochrane systematic reviews are published, also contains a database of RCTs/CCTs. This includes records from MEDLINE and EMBASE, but also the many records that have been found by handsearching journals (more about this later)
CINAHL via EBSCO – examples – Oral hygiene for nursing home residents PsycINFO is via Proquest – example – Interventions for reducing dental anxiety in children and adolescents LILACs – large number of dental schools in South America, so there are a number of dental journals on here, dates back to early 1980s Each cochrane group keeps a specialized register of RCTs/CCTs in their particular field, OHG’s register is substantial with 26,000 citations. It can only be searched for authors of Cochrane reviews.
OpenGrey has been relaunched and is now a resource that’s free to use and features grey literature (mainly European) I wouldn’t search web of knowledge as a general rule as it’s not the best source for RCTs, but it can be useful for discovering conference proceedings. There are other trials register platforms also – for example, World Health Organization have one, although it’s future is a little uncertain at the moment. There are some national trials registers which are freely available and also can provide you with access to unpublished studies.
As we’ve already discussed in the workshop, a well constructed question is the place to start. The better the question, the clearer the concepts, the easier the search is to construct.
So we’ve already seen this… Some more examples…
Searching is a trade off between sensitivity (picking up all studies that might be out there) and precision (picking up only studies of interest to your question). Cochrane systematic reviews aim for maximum sensitivity in order to pick up every trial on the topic. It’s not uncommon to get 1000s of references to screen (the most I’ve seen was 16,000 for the review on Oral Mucositis). This is an extreme example: search results are likely to be in the 100s!
Pt 1: So if you search for a particular subject heading, for example, Toothbrushing, you’ll retrieve all articles with the subject heading of toothbrushing. Pt 4: MEDLINE uses MeSH, as does CINAHL. EMBASE has a controlled vocabulary called EMTREE. We’ll be concentrating on MeSH and PubMed today.
So this is an example of the hierarchy – within tooth diseases you have tooth demineralization, and within that dental caries. Tooth diseases itself lies within stomatognathic diseases. It’s a bit like a Russian doll, inside each heading are more headings, each retrieving a smaller subset of results.
Demo – follow the link. Look up dental caries – explain scope notes, entry terms – how entry terms can be used to build up free text search.
Cochrane searches are systematic, and use both free text and controlled vocabulary.
You have to be careful not to exclude things from your search. Databases are not intuitive.
PubMed (which we’re using today) does not.
Now we’re going to discuss how to put all of this together to create the search – example I’m going to use is a systematic review on chlorhexidine for the prevention and treatment of dental caries in children. Our participants are children with caries. Back to task 1 – here are the synonyms I found for caries…
And the synonyms for the intervention: chlorhexidine. Note: some of these are trade names. That is allowed in a systematic search.
We should take a decision on whether to add these terms to the search or not. Will it be too restrictive? Need to test to find out, and talk to the authors about which groups of children to target.
Specific groups: children, elderley, even things like smokers? Caution against using limits generally, because indexers are human and make mistakes! But could be used in the testing process.
Different ways of saying caries!
Remember databases are not especially intuitive!
So you would start to enter your search in the box, and click “preview” to keep track of how many hits as you go along. If you want to check the results at any point, click through on the number to the results to see what is being retrieved.
This is especially true of older citations.
Searching for Studies Anne Littlewood Sue Furness Trials Search Co-ordinator Editor/Methodological Support Cochrane Collaboration Oral Health Group University of Manchester, UK
Use a research question of your own, or use an example research question below to develop a search strategy. List the participants and intervention and find as many synonyms as you can for each term. Use the internet to help you, and try and find an article on the topic that might give you additional keywords.
Example Research Questions:
Alendronate for preventing tooth loss in postmenopausal women
Hypnosis for anxious children undergoing dental treatment
Occlusal adjustment for treating and preventing temporomandibular joint disorders in adults
For example, if the opening sentence of the abstract is:
“ This article looks at cancer but exclusively cancer in adults and not cancer in children”
The search “cancer NOT children” will not pick up this potentially relevant paper, as you have asked the database to search for cancer but not children, and both of these terms are in the opening sentence of the abstract
Searching for participant information can be difficult e.g.
if you are looking for adults rather than children, putting the term “adults” in will not necessarily help as “adult” might not be in all the articles you want to retrieve (could be “patients” or “participants”)
Search terms for participants should therefore be used with caution!
Most helpful when there are specific groups
PubMed has search limits that can be added and tested as part of the search
Go to the National Library of Medicine MeSH browser: http://www.nlm.nih.gov/mesh/MBrowser.html
Enter the search terms and synonyms and make a note of the MeSH terms for your topic. Do any need to be exploded?
2. Look at your identified synonyms and think about how to build a free-text search:
Can any of the terms be truncated? Are there any alternative spellings? (for example, think about US English v British English spelling: -ize versus -ise or -ization versus -isation.) Are there any terms that can be represented by initials alone? How should these be represented in the search?
3. Now test your search in PubMed (use Advanced Search)