Library Guides: Systematic review guide: Designing a search strategy

Designing a search strategy: The "Master" strategy

While you will use several databases over the course of your searches, you should begin by designing one "master" search strategy. This is usually designed in and for Ovid MEDLINE. This is likely to be the search you have reported in your protocol, with some tweaks or improvements.
Do not worry about using other databases until you have completed and tested your master strategy and are happy with the number and relevance of results it is retrieving. It is essential you get this search right before continuing searching elsewhere. The master search is then used as the basis for the searches in other databases and changed as necessary to suit each database used. You will find advice on search strategy adaption (translation) in further on in Section 6 under "Adapting a strategy for other databases".

Designing a search strategy

MEDLINE, Embase, CINAHL, PsycINFO (and other databases) can be searched using a combination of two retrieval approaches. The first is text words or keywords (words occurring in the title, abstract or other searchable fields available in the database). The second is based on subject headings (standardised subject terms from a specific thesaurus or controlled vocabulary, assigned to an article record either by human indexers or automated indexing approaches).

A systematic search must include both keywords and subject headings (except for searches in databases without subject headings such as SCOPUS). Anything that may be missed by one method will hopefully be found via the other. For instance, recently published items may not yet have subject headings added to their article record, while keyword searching is vulnerable to issues such as the searcher missing or omitting synonyms, or there can sometimes be spelling mistakes in the the bibliographic information provided by journals to the database.

Approaches for identifying and searching with keywords and subject headings and how to combine them appropriately within a search strategy are detailed below.

Systematic Reviews: CRD's guidance for undertaking Reviews in Healthcare (pdf)
"Appendix 2: Example Search Strategy to Identify Studies from Electronic Databases" (p.243) details a professional systematic search strategy, explains how it was developed with worked examples and explanations throughout.

Keywords and Subject Headings

Section 3 of the Library Services Literature Searching eLeaning tutorial for Medicine and Biomedicine explains how keywords are searched in databases as well including how to utilise wildcards, truncation and proximity and adjacency operators in keyword searches.

Section 4 of the Library Services Literature Searching eLeaning tutorial for Medicine and Biomedicine shows where subject headings are indexed in databases, how to identify appropriate headings and include them in your search.

Section 5 of the Library Services Literature Searching eLeaning tutorial for Medicine and Biomedicine demonstrates how to combine your keyword and subject headings lines with Boolean commands AND / OR to construct both individual search concepts and a final comphrensive strategy.

Ovid MEDLINE search guidance

Ovid MEDLINE Tutorial
This tutorial guides you through searching using both keywords and subject headings, combining your search lines using Boolean commands like AND / OR and how to use more advanced search functions like truncation and wildcards, as well as how to find full text. Some material will help you search in a more systematic and structured way - for example creating an answerable, focused search question
Introduction to Ovid Databases
This guide explains the difference between Ovid databases, provides information about MEDLINE and guidance on search commands and techniques for MEDLINE and the other Ovid databases.

Text mining

Text mining can help you identify further search terms by scanning for recurrent words and phrases as well mapping text to subject headings and visualising your search. You will need a selection of relevant articles that you've already identified to "mine".

PubMed PubReminer
Will run your simple keyword search through PubMed and display the resulting abstracts as well as generating frequency tables. These tables will show you the journals in which your query is published the most, the authors which are most active in the field of your query, the most commonly used words in the titles and abstracts and MeSH (MEDLINE's Medical Subject Headings) related to the topic.
Yale MeSH analyser
Allows entry of up to 20 PMIDs (Pubmed's unique ID number for each of it's records) of relevant articles. It will create a grid that shows the MeSH headings assigned to them by PubMed.
PubVenn
Lets you explore PubMed using venn diagrams - showing a the relative size of the citation set for each term as well as how those sets interact.
MeSH on Demand
Will identify MeSH terms in your submitted abstract or manuscript and list similar articles.
Systematic Review Accelerator Word Freq
Takes a library of articles in XML format from EndNote and outputs the most common words/phrases that appear within that library.