Systematic Review: Identify search terms, initial search & gold set

Developing a search strategy is the process of converting your research question into a format that the database can interpret. Databases work by matching the search terms that you enter to the titles, abstracts, and subject headings in the records of items in the database. In order to locate the relevant literature you need to use the same terms that are used in relevant sources.

For each identified concept in your research question, make a list of relevant keywords and subject headings.

Keywords

• Keywords are simply the terms used within an article. A database will generally search for keywords in the title and abstract fields, and may also search other fields of the database record.
• It is important to include alternative spellings and synonyms for your keywords to retrieve all articles on your topic.
• Author assigned keywords may also be searchable depending on the database being used.

Subject headings

• Subject headings are internationally recognised index terms that can be used to describe what an article is about. They are assigned after an article is published, and are based on the topic of the article (they may not be words used in the title or abstract).
• A database will only search for subject headings in the subject field of database records. You can also browse subject headings and see broader or narrower terms in the "Tree".
• Subject heading searching increases relevance and recall, and when combined with keyword searching is considered best practice for search strategies when reviewing the literature. 
• Subject headings are often unique to a particular database, so you will need to look for appropriate subject headings in each database you intend to use. They are not available for every topic so only use them if specifically relevant.
• Not all databases will have subject heading searching so ensure that your keywords are robust enough to stand alone in databases that don't index by subject headings.

As part of the process of developing a search strategy, it is recommended that you keep a master list of search terms for each key concept. This will make it easier when it comes to translating your search strategy across multiple database platforms. 

Once you have identified your key concepts, identify the words most likely to have been used in the published literature on this topic. It is important to develop a comprehensive range of terms for each discrete concept entailing a combination of subject headings and a wide range of keywords/phrases for each concept. 

You can scan or  ‘text mine' the papers in your gold set for useful search terms.

• Check if the author has provided a list of key terms that describe the content of the article, these may appear below the abstract or in the database record. Also look for relevant keywords or phrases that have been used in the title and abstract fields of the article.
• Use a title search to locate the gold set papers in a key database such as Ovid Medline or Embase and then use the database record to scan for subject headings and useful keywords or phrases.

Text mining tools you may want to try out are:

• PubReMiner Enter a simple search string or PMIDS (PubMed IDentifier) from gold set to determine high frequency words and subject headings
• Termine Upload a pdf, URL, or copy and paste an abstract to highlight frequency of words and phrases
• MeSH on Demand Enter abstract or text to identify subject headings.

As part of documenting the SR process, it is recommended that you keep a master list of all search terms for each key concept. Create a concept map to list the subject headings and key terms that you have identified. This list can be quite extensive and it can be helpful to use an Excel spreadsheet. This will make it easier when it comes to translating your search strategy across multiple database platforms.

It is important to start by doing some preliminary searching, or scoping searches, before you develop your search strategy. This type of searching usually involves a series of very targeted searches, perhaps looking for your main keywords in the article title. Conducting a number of scoping searches will help you to understand the availability of existing literature in your area of interest, and will enable you to build a 'gold set' of relevant references.

To be included in your gold set, an article cannot simply be background reading, it MUST be one that you will include in your review. Ensure that you have a reasonable number of papers in your gold set; try for at least 5-6 papers but preferably more.

Locating the gold set:

• Results of preliminary searches from key databases such as Google Scholar, Scopus, Ovid Medline and Pubmed. Developing a search strategy should always begin with preliminary searching to ensure that your question is feasible and hasn't already been answered in the literature.
• Key papers recommended by subject experts or your supervisor, or team members on your review.
• Citation searching ('pearl growing' or 'snowballing') techniques to locate further relevant references cited in or by key papers. VOSviewer and Connected Papers are just two of the tools you could use to conduct forward and backward citation tracking.
• References used in similar systematic reviews.

Reasons for the gold set:

• To help you identify relevant search terms and/or subject headings:
  Look at the article within a database paying close attention to the 'searchable fields' such as title, abstract, and keyword or subject heading fields to identify relevant terms used in the literature.
• Testing your search strategy:
  When an article in your gold set is not retrieved by your search strategy, you can revise your search accordingly in order to include it.  Always check that the article is indexed by the database you are searching, then see what keywords can be incorporated into your search strategy so that the article is retrieved.