2.4.3 Developing a Search Strategy

At the protocol stage, an initial search is conducted, usually in the early stages of the project. This search should cover at least one, and ideally two, of the major bibliographic databases relevant to the review question (e.g. MEDLINE/PubMed and CINAHL). This initial search assists the librarian or expert searcher to develop and refine the search strategy and ensure that the entire review team is familiar with the process, expectations and the management of resources.

In starting a review, the initial exploratory search is key to determining the potential number of citations that will need to be screened, and to check whether there are relevant papers that will meet the inclusion criteria. The first step is breaking up the review question into key concepts. During this process, the question may also be refined to be as specific as necessary to reach the objective. Therefore, a search is best developed in collaboration with the librarian and the research team at the beginning of the project (see Table 1).

Table 1: Example of a logic grid to break up a review question

For most review questions, it will be a combination of terms about the population, the topic of interest or condition, and the intervention or factors to be studied. Multiple mnemonics have been used to structure research questions, with the most familiar being PICO (Population, Intervention, Comparison and Outcome). However, JBI also uses others depending on the type of review question, including PCC (Population, Concept, Context), PICo (Population, phenomenon of Interest, Context), CoCoPop (Condition, Context, Population) or PIRD (Population, Index Test, Reference Test, Diagnosis). There are exceptions to this rule. A librarian experienced in evidence synthesis will be able to determine the best approach. Moreover, for each concept, consider terms across interprofessional fields (e.g. retention means very different things to the hospital’s human resources manager and a urologist), international spellings and synonyms (e.g. British terms vs. American terms such as anaesthesiology / anesthesiology) and historical terms (the definition of autism has changed over time). The following guidance gives examples.

Population and Topic of Interest terms: The population and topic of interest terms are usually the first concepts that come to mind when searching. If we follow the example set out in Table 1, our population is elderly patients with advanced cancer. In this particular instance, we may not need to specify the type of cancer, although these may be considerations as you consider your keywords and subject headings. You may need to test some examples of subject headings and keyword combinations to ensure the balance between sensitivity and specificity is right.

Intervention terms: Intervention terms should be specific and directly relatable to the question, such as the name of a medication or treatment regimen. In the example above, our ‘intervention’ is advanced directives and other forms of shared decision-making. As with the population terms, use subject headings where they are available and appropriate and combine with keywords. While searching for medication terms, remember to use both trade and generic names as well as subject headings when applicable. It is tempting to add the terms ‘intervention’ or ‘program’ to limit results down to intervention studies. Be cautious when applying vague terms such as these to reduce results, as this would significantly reduce numbers while potentially removing key articles.

Outcomes: For many reviews, outcome terms are not appropriate as part of the search (Shokraneh 2024). There are several reasons for this. First, authors do not always add the outcomes measured in their study in the abstract or keywords, so limiting to the use of these terms in the abstract or title search will needlessly reduce your results. Second, many outcomes have a variety of terms to express the concept, many of which are vague. For example, if searching for improved family relations, how will that be determined? An improved relationship between family members themselves, between family members and the ill loved one? Between family members and the care team? Many outcomes describe ‘best practice’ without giving a definition of how that is determined. Outcome searches can also inadvertently lead to bias by omitting results with outcomes that the searcher has not foreseen or otherwise thought to specify.

Tools such as concept maps, logic grids or other visualisation tools may make the process easier for all involved. Table 1 above is an example of a logic grid, which breaks down the question into the key concepts. The searcher can add synonyms, keywords or subject headings to each column.

Search filters: Search filters or search hedges are previously validated searches that can be applied to multiple topics. An example of a search filter might be the Economic Evaluations/Cost/Economic Models published by the Canadian Agency for Drugs and Technology in Health (CADTH), which for PubMed is:

This lengthy search filter can be added to a PubMed search to ensure a sensitive retrieval of articles included in PubMed on economic evaluation, economic models or costs, with the reassurance that previous reviewers have tested the results, thereby saving significant time for the searcher. Search filters should always be cited.

There are many search filters available for research design, such as the one listed above (‘Strings Attached: CADTH's Database Search Filters,’ 2019). Moreover, search filters are often tailored for specific geographical locations, such as low- and middle-income countries (LMICs) or Northern Europe. Sometimes, they are for specific population groups, such as paediatric populations, or Aboriginal and indigenous persons. Search filters can be found on the websites of several libraries, practice guidelines and health technology assessment organisations that are too numerous to list here. You can also find search filters within published reviews. A selection of search filter resources is available at:

• The InterTASC Information Specialists' Sub-Group Search Filter Resource: https://sites.google.com/a/york.ac.uk/issg-search-filters-resource/

• Strings Attached: CADTH’s Search Filters: https://searchfilters.cadth.ca/

• SIGN Search Filters: https://www.sign.ac.uk/using-our-guidelines/methodology/search-filters/

In addition to a search strategy for one database to include in your protocol, you will also need to list any additional databases and grey literature sources that you plan to search. Include both the names of the databases (e.g. CINAHL, PsycINFO) and their platforms (e.g. EBSCOhost, Ovid, ProQuest). You must also include the date searched, languages searched and any limits used.