A systematic review is called "systematic" because it is highly process-oriented and seeks to approach all elements of the review in a structured, formal way. The process should be transparent and repeatable, meaning that others can use your process to repeat it for themselves. This is equally true of searching the literature. For more information on the Cochrane approach to searching, please see Chapter 4 of their Handbook. I have also created a Summary of Chapter 4 in the Cochrane Handbook. Additionally, it is worth consulting the PRISMA-S Guidelines for Searching, which provide a 16-item checklist for reporting how your search was conducted.
Having formed an answerable question, in Step 1, you now have the components necessary to approach your search. In literature searching these "components" are referred to as "facets." Facet searching helps you break complex queries down into a few key ideas that you will then cast into terms to use when searching one or more databases.
For example, in the PICO question from Step 1, the question is "Should antibiotics be prescribed for children with Acute Otitis Media?" Because the question was structured using PICO, we can see the facets plainly: antibiotics, children, Acute Otitis Media. These terms will be our facets during our search. These facets can be modified, for instance, which antibiotic? Or are you interested in all of them? Which age group of children? Teens? Toddlers? Or all children? But the main outline of the search strategy is plain.
The next step in the process may be to decide which type of study you're interested in finding. For some questions, the best article type is one that reports on a Randomized Controlled Trial (RCT). In ranking the quality of evidence that emerges from various study designs, randomized controlled trials are at the apex of what is considered reliable and effective. In some cases, you may want to run what is called a "scoping search" first. This type of search will give you a broad idea of what results you may find and which type of article and study design will best answer your question. There are, in fact, at least fourteen different types of reviews as identified by Maria Grant in her 2009 article. This scoping review will do a number of things: first, as previously stated, it will provide you with a sense of what type of article and which publication is best suited to your question; second, a scoping review will help you see if your question has already been proffered; third, you will be able to see gaps in the literature that you may be able to fill.
For the Cochrane approach to developing a search strategy, see Chapter 4.4, Designing search strategies. When performing an initial search, you will want to identify the key facets (concepts) and find all of their various synonyms in the literature. One key first step, is to examine the thesauri used by key databases.
In the United States, MEDLINE/PUBMED is the main database used to search the medical literature. A more international database is Elsevier's EMBASE, and much of the content from that database can be found in SCOPUS, but not all. An additional database to consider for the medical literature is Web of Science. If you are searching a topic that is multidisciplinary in nature, you may have to consider CINAHL or other Nursing and Allied Health databases, PsychINFO or similar databases for the psychology literature, or other databases for Social Sciences. Case Western Reserve University libraries provide access to well over 500 research databases in a variety of fields.
Medline/PubMed uses Medical Subject Headings (MeSH), to control the vocabulary used in searches. MeSH is a standardized set of subject headings that are assigned to most articles in MEDLINE by human beings -- that is, they are conscientiously assigned terms, not automated or computer indexed. Subject headings and thesauri attempt to control the terminology assigned to an article so that the one term assigned collects articles using all other variations of that terminology in the literature.
For example, the use of the MeSH term "neoplasms" ensures that the majority of variants are picked up (cancer, cancers, tumor, tumors, malignancy, etc.) For the example, for the Venekamp article we've been referring to, one key term in the MeSH database is Otitis Media. The entry term provides the following information: the term heading, a scope note, subheadings, alternative entry terms (which can be gleaned for related terms), and the subject tree showing broader and narrower terms. The entry terms include "Middle Ear Inflammation," and the narrower terms include "Mastoiditis," "Otitis Media with Effusion," etc.
In PubMed, once you've found a MeSH term, you can add the term to the Search Builder and then Search PubMed:
Set 1: "Otitis Media"[Mesh], yields 25,640 results. Then we continue to add terms.
Set 2: "Anti-Bacterial Agents"[Mesh], 437,334 results.
Set 3: #1 AND #2 (using Boolean operators to combine concepts), 3,582 results.
Then, we'll limit our results to Children. Which gives us 2,449 results.
The above is a very general example. For each MeSH heading, Entry Terms should be taken from the MeSH database, and searched in the Title/Abstract fields; as well, Gold Standard articles, or ideal articles, should be gleaned for additional terms that can be added to the search. For example, for Set 1 above, you should also include: glue ear[tiab] OR otitis media[tiab] OR OMP[tiab] OR AOM[tiab] OR ear infection[tiab] ear ache[tiab], and so on, with all major concepts.
The search above should be repeatable in PubMed, should you wish to run it. But this is a basic search, and it does not take into account any of the variant Entry Terms that you will find in a the MeSH database, it does not take into account any variant terms not listed or Author Keywords. Nor is this search particularly narrowed or honed, for instance, you may have a particular antibiotic in mind, and the MeSH term is way too broad. Or, perhaps, you're only interested in pre-school children, not teenagers, etc.
It is important to understand that the process of literature searching is iterative. As you get results, you should discover new terms that you can add to your search statement which will improve the precision of your results. It is also important to spend time learning to use the database that you are going to search. MEDLINE provided by EBSCO (a publishing company) is going to be different than MEDLINE through the publisher OVID, which will be different from PubMed. It is important that you understand and become familiar with the tools you are using when conducting a search. It is also recommended that you first build your search in a separate document, like Microsoft Word, or some other program, and use that to document and archive your search query/search process. You can then alter and refine your search in the document, and cut and paste it into the database you are using to quickly run and re-run your searches.
Once you are satisfied that your search query is where you want it to be, then you will run your search in the databases that you have identified: PubMed, EMBASE, SCOPUS, etc., and download the results into a citation manager.
When you have completed your search and the results are satisfactory, you will download the citations, usually with abstracts, into a software application to manage them. There are well-known citations management tools, such as EndNote, RefWorks, Zotero, Mendeley, etc. These software applications are generally designed to aid in the creation of bibliographies and footnotes for papers on which you are working. However, they also have some features/functions which are desirable for the purposes of systematic reviews, including deduplication of matching citations -- especially important if you are searching multiple databases.
Citation management software, however, falls short when it comes to the more advanced features needed for full systematic reviews, which include the need to share access amongst multiple reviewers, the ability to select or reject articles (and denote the reason for the rejection), machine learning which will aid in the identification of studies as you approve them, and much more. For a full list of potential software applications that can be used, see the Tools & Resources section of this LibGuide, or look at Chapter 4.6.6.1 in the Cochrane Handbook (Software for managing the selection process), or look at the website, Systematic Review Toolbox.