Now that you’ve articulated your clinical question, the next step to locate the best relevant research information to answer your clinical question, which is step 2 in the practice of EBP.
So, how do you locate relevant information? There are three broad steps within this step:
First, you’ll want to select a resource. Depending on your time frame and how obscure your question is, you may want to go for a prefiltered source like UpToDate or Micromedex, where an expert has already summarized the information for you and you can get a quick answer or review of the available evidence. Or, you may want to go for an unfiltered resource like PubMed or Embase, or another secondary source, where you can search for the primary literature and summarize it for yourself.
Next, you’ll want to execute your search strategy, which will depend a lot on which source you’ve selected. For the prefiltered sources, you can get away with a simpler search. For the unfiltered sources, you may want to use a more complex search. Especially when you’re executing your more complex search, you’ll want to make sure you think about searching with natural language or free-text terms, plus adding in any controlled vocabulary terms like MeSH or Emtree if you’re searching in a database that uses those (MeSH and Emtree are covered in the previous sections). Then you’ll want to think about how you want to combine your terms. You’ll also want to consider the logic of your search, where to put and, or and not, and how to use parenthesis.
Finally, once you’ve executed your search, you’ll want to look at your results to see if they’re on target. Not everything will be relevant, but if you scroll through the first page of results, you’ll want to see at least a few that you want to look at more closely. If this doesn’t happen, then you may want to think about going back to your search terms and refining them.
There are two basic types of questions that you can apply EBP skills to: research questions and clinical questions. A clinical question is one that you might have if you’re in clinic or with a patient, and you want a quick answer where someone has already done the research for you. For these types of questions, you would use want to do a quick search in the “tertiary” literature, so something from a prefiltered resource like a Cochrane review or evidence from a drug database like Micromedex. This will give you information that a group of experts has already digested for you, and you can get a quick answer that will hopefully be relevant to your case.
If you have a slightly longer timeframe and you’re interested in doing some research yourself, or if, perhaps, you have a more obscure question where the answer isn’t readily available from a tertiary resource, or you need background information, then you’ll want to do a more in-depth search in one of the unfiltered sources, like PubMed, Embase or Scopus. These will give you primary literature, or studies, and secondary literature, or reviews and meta-analyses of these studies, which you can then evaluate yourself.
Once you’re clear on your PICO and your clinical question, and you’ve decided on a source, how do you go about actually searching that source? There is more information on searching individual sources in the other modules, but these are some general tips.
The first tip is to use variant word endings or truncation. This is particularly important when you’re doing textword or keyword searching, because when you search for these, you won’t get automatic mapping. You also want to remember to be careful with truncation. If you search for prevent* you’ll get prevents, prevention, preventing, etc. However, when you’re truncating shorter words, you want to be careful that you’re actually getting what you want. If you search something like pharm* with an asterisk, PubMed will only search the first 600 word endings, in alphabetical order, so in this case, you’ll only get to pharmacodependency. You won’t get pharmacotherapy or even pharmacy. If you want to search for rat studies and you type in rat*, you'll also get terms like rate, rates, ration, and rather, none of which are relevant.
You also want to remember to use synonyms, so you’d want to search stroke OR brain infarction instead of one or the other. That way, if one article says stroke and one says brain infaction, you’ll still get both. This is also a good thing to remember when you’re moving your search between databases, because PubMed might prefer one term and Embase may prefer another.
In addition to using keyword searching, you also want to make sure you search for controlled language terms like MeSH or Emtree before you start your full search. Controlled vocabulary like MeSH or Emtree is a list of all of the important concepts for each article or entry in a database. It will use the same term each time, so even if you have two authors who have used different language to describe a concept, if you’re using MeSH to search, you’ll still pull both articles.
Generally, you also want to use your PICO to help build your search logic. So, first you would OR together all of your terms from your population, then your intervention, then maybe your comparator and/or your outcomes, depending on your topic. Once you have a list of “OR”ed synonyms, you can put that list in parenthesis and then AND your major concepts from the PICO.
Boolean Searching: Use the connector terms AND, OR, & NOT to structure your search query.
Boolean logic is a code format that tells almost all databases exactly what results you want to find. You can use it to tell the database how to limit or expand your search. The basics :
College AND High School
Combining search words with AND narrows the search.
College OR High School
Combining Search words with OR broadens the search.
Using your PICO, you can craft a sample search to find studies or reviews that answer your question. This is what your search history might look like:
|#1||stroke OR cerebrovascular accident OR apoplexy OR brain infarction||285888|
|#2||baby aspirin OR low dose aspirin OR low dose acetylsalicylic acid||6959|
|#3||clopidogrel OR plavix||12127|
|#4||#1 AND #2 AND #3||142|
|#5||(#1 AND #2 AND #3) Filters: Middle Aged: 45-64 years||44|
This search was done in PubMed, but you could use the same concepts to create a search in Embase or another database. In this instance, I ran my search concepts one at and time and then eventually combined them together in search #5.
So, Search #1 is a list of synonyms for the population: stroke or cerebrovascular accident or apoplexy or brain infarction. Search #2 has a list of terms for the intervention: baby aspirin or low-dose aspirin or a low-dose acetylsalicylic acid. Search #3 is a list of terms for the comparator: clopidogrel or plavix. Then, I started to combine my search terms together. Search #4 is a combination of all three of the searches, so #1 AND #2 AND #3. And, finally, I used an age filter for middle age, which is ages 45 to 64 in PubMed, to further limit my results to those that would be most relevant. This gives me a final total of 44 articles found.