Applying a Little Common Sense to Your Search Results
Recently I was asked to perform a literature search looking for studies on off-hours care for acute myocardial infarction and stroke. Off-hours care means care during non-normal working hours like nights, weekends, and holidays. When I formulated my search strategy and ran the search I retrieved the following set of results.
I immediately suspected something was wrong. The number of hits on the Cochrane Database of Systematic Reviews is too large. It doesn’t make sense that the number of systematic reviews on a topic greatly outnumbers the number of studies that are used to produce those systematic reviews. With the number of potential studies being a little over 300 (162+173), and not all of those hits being relative to the topic, I would expect Cochrane to give me only a handful of results. Maybe 20 or less.
So I had to dig through my search strategy and the Cochrane database guide to find out what was happening. It turns out that Cochrane indexes certain words, called Stopwords, differently than Medline and Embase. Here’s the definition of Stopword from the Cochrane database guide.
Stopwords are words of little intrinsic meaning that occur too frequently to be useful in searching text. You cannot search for the following stopwords by themselves, but you can include them within phrases by placing the entire phrase within quotation marks.
My search terms included “after hours”, “after hour”, “out of hours”, among others, combined with care, admission, admitted, etc. “After”,” out”, and “of” are in Cochrane’s list of stopwords. I thought I could search them since I was including them in quotation marks, but Cochrane adds another “feature” to their database that prevents this.
The default search in Cochrane searches several fields including the full text and abstracts. Here is the description of the abstract field from the Cochrane database guide.
The Abstract (AB) field summarizes the content of the document and generally describes the background, methods, results, and conclusions. Only documents which appeared in print with an abstract will include an abstract in this database.
Stopwords such as “of” or “the” display in documents but are not searchable and do not appear in the Abstract field.
The last sentence about stopwords is the most important. Stopwords work the same way in the full text field.
So what was happening was that searching for “after hours” was really only searching for “hours” since “after” is not included in the index for the abstract or full text fields. This results in retrieving a large number of irrelevant references. The way I fixed this was to limit my search in Cochrane to the title, abstract, and keyword fields. It still doesn’t properly fix the problem, but it greatly reduced the hits in Cochrane to around 25. It is also a reasonable strategy, since if the systematic review really is about the topic of after hours care you would expect “after hours” and its synonyms to show up in either the title, abstract, or keywords.
It should be noted that Embase and Medline in Ovid are indexed differently and stopwords are searchable using quotation marks, e.g. “after hours”, “out of hours”.
So, apply a little common sense to your search results. Does what you’re finding look correct? If not, it might take some detective work to find out why not. If you need help, you can always call the sleuths at your local Mayo Clinic library.
Reference Librarian, Plummer Library
Entry filed under: What's Your Reference Question?.