Usually in systematic reviews a researcher will search enough resources to be able to state with confidence that the literature for the subject area has been comprehensively searched. However, there are no rules about how many resources should be included. Often the best test is to search a new resource for a limited component of a review to see if new items are identified.
The scope of your research question should be kept narrow so that a large number of irrelevant items do not have to be excluded from your result. If you are getting more than 3,000 items from the first database used with the search strategy for the review the scope of the research question or the structure of the search strategy should be examined closely.
The search engines to include in a systematic review will vary depending on the subject field and question being asked.
Several suggestions for databases to search are listed below. But we recommend you consult the Subject Research Guide in your field for a more extensive list. See http://unimelb.libguides.com/
Do not search both Medline and PubMed. Medline is a subset of the PubMed database platform. Medline (Ovid) searches are equivalent to searching all of PubMed.
Other versions of Medline (Ebsco, Web of Science) do not include a segment of PubMed. PubMed searches can be limited to this segment alone by searching for publisher [sb]. PubMed does not do proximity searching which can be important for reducing the search results found in systematic review searching. The University librarians can assist in designing a search to cover the contents of PubMed/Medline as completely as possible.
Extensive systematic reviews can include Web of Science Core Collection and/or Scopus. They may be used to follow up references and related articles of key included papers. Google Scholar may be used for a simplified one line version of the search strategy. Often the use of Google Scholar is restricted to a few pages of results.
Brainstorming keywords for a search strategy for a systematic review is essential. It is important when doing this that the terms used are relevant to the concept being searched for and do not cross conceptual boundaries. Each term should be tested in the major database to be used to identify what impact it is going to have on the search results received. Truncation may be used but again it is important to test each truncated term first to see that only expected terms are retrieved.
Be careful not to create an overwhelming list of terms as this will only increase the amount of scanning through irrelevant items to be done. The key to a good systematic review search strategy is lots of testing. Run the strategy on the primary database to be used and if you get a large number of items returned examine each item in the first 50 records carefully to identify the source of the terms you included. Try and identify major sources of irrelevant records and adjust the search strategy accordingly.
Words that have multiple meanings are especially prone to creating havoc with the search strategies and will often need to be searched with other terms to get the correct context of the term. For example, a search for the impact of the size of soft drink containers would include the term “can” which also appears in any sentence in the verb form (“can increase”, “can improve” etc.) Therefore the search would need to ensure that the search for can only returned items that also included related terms such as beverage or soft drink.
Be careful to test any acronyms that are being searched for. It is often not appreciated that acronyms will have many meanings in different subject fields and may often be the names of gene fragments or biochemical compounds as well as your chosen meaning. If you see very irrelevant items appearing after using acronyms then it may be necessary to search for the acronym together with other terms that will improve the context of the results.
Multiple terms covering the same concept are searched for with OR commands. If using Ovid databases, then a short cut is available for combining many separate searches. Search for OR/(first set)-(last set) (e.g. OR/1-15) to create a set containing the results of all of the included terms.
Most databases offer a thesaurus or list of available subject headings that can be allocated to an article by the author or indexer.
Use of the preferred term is a powerful and controlled way of directly accessing most if not all of the material within a field of study.The lists of preferred terms in Medline (MeSH), Embase (Emtree)) and Psycinfo are particularly extensive. Each is a hierarchical arrangement of broader terms, preferred and related terms, and narrower terms, designed to map the context and content of their respective fields.
In the OVID versions of Medline, Embase and Psycinfo the Search Tools feature offers a Map Term function for looking up the preferred term(s) for a given topic.
Take the term Acquired Brain Injury. You can map this to the preferred terms(s); check the definition of the preferred term in its scope notes; and "explode" to include all narrower terms.
The Medline MeSH tree is as follows. The preferred term is Brain Injuries, but this can be further subdivided.
The syntax to retrieve all uses of the term brain injuries AND its narrower terms is...
exp brain injuries/
Additional search limits - or filters - can be applied, such as limiting to English language, human or animal, gender, age group, whether peer reviewed, and type of study (such as review, therapy, diagnosis, prognosis, or causation-etiology).
The Emtree headings in Embase have a different arrangement. Acquired Brain Injury is a narrower term within Brain Injury.
Psycinfo has a different tree again, and Traumatic Brain Injury is the preferred term.
Boolean searching looks for the presence or absence of words in the fields of a record, or the text of a document.
Proximity searching allows better control of the relevance of concepts by adjusting their proximity to one another. If the concepts occur close together in a sentence or paragraph, the topics are more likely to be relevant than if they are widely separated.
Most database platforms offer proximity operators to specify word order and separation. Check the help system of the database you are using and look for proximity to find information on how to apply it correctly.
Replace # with the maximum number of words to occur between the two concepts.
Grey literature is unpublished material or has been distributed outside mainstream commercial publishing. It may include reports, theses, government and NGO publications, conference abstracts and proceedings, registries of clinical trials and prospective studies, and the results of hand searching or corresponding directly with authors.
Sources and/or repositories of grey literature include: