As mentioned in Step 1, it's a good idea to do some rough preliminary searching before you start your systematic approach. This will help with identifying key authors and keywords and come up with a very few articles which seem important to your topic. These key articles will become your "gold set" which you can not only use to identify search terms, but down the track can help you test your final search strategy. If these articles don't come up when you test your strategy in the databases, you must be missing something.

You can also use your gold set for snowball searching ​- following links to citations and reference lists of an article to find related articles.

Back in Step 1, you formulated your research question and may have used a search framework to help you do this. The framework (and the question itself) will direct you to the main two or three concepts you need to search. For each concept you will need to consider all the different terms that could be used to describe it. Use your own knowledge of the topic along with words and phrases from preliminary searching you have done to come up with these. Also consider some text mining, or the use of subject dictionaries to identify all possible search terms to use.

Remember to consider:

• both scientific and common terminology
• alternative spellings (such as British and American spellings, and possibly common misspellings)
• outdated and culturally-specific terms
• abbreviations or acronyms
• plurals
• database-specific controlled vocabulary or thesauri (eg. MeSH in Medline, CINAHL Headings in CINAHL)

For example, using the scenario we described earlier, “Does cranberry juice help reduce urinary tract infections in people living in aged care facilities?”

Note the use here of phrase searching (usually " ") and truncation symbols (most often *). Some databases use different symbols and some other techniques. Check our Database Help guide for specific instructions.

Once you've come up with as many specific terms as you can find and you are in the database, take one concept at a time, using OR between synonyms. At the end, link the big groups of each concept with the AND operator.

Usually you would not include every framework element in your search strategy. Here we have left out the Comparison element as there is no clear comparative intervention. Often the Outcome is left out because it comes out of all the found studies in any case, and may restrict the scope of the search. In many cases, two of the concepts you choose to search may come from the same framework element. The concepts for searching don't always align exactly with the framework - its purpose is to clarify your question and your concepts.

It's important for any literature search to consider the limitations of your search. For systematic reviews it is vital that you establish the boundaries very early on. Which studies will be excluded from the review and why? 

For a systematic review, information about the inclusion and exclusion criteria is usually recorded as a paragraph or table in the methods section (as well as in the Protocol).

Common criteria include:

• Time period - Perhaps there has been a previous review and this is an update, so included articles may be only those published since the previous review.
• Language - It is not usually necessary to arrange for translations of articles published only in other languages.
• Publication type - Usually you will only be looking for original studies, so you may consider excluding editorials and letters.
• Geography or setting - You may need to limit the review to a particular population group, or a particular setting (eg. school or hospital).
• Age groups - Many databases include particular age groups as subject headings or limits if this is relevant to your research question.
• Study design - Consider including specific study designs, such as controlled trials or qualitative studies, if that is applicable to your question. This could be a way of making a broad review more manageable.