Skip to Main Content

Systematic review guide: 2. Develop a search strategy

Search strategy

Systematic review searches need to be as extensive as possible to maximise recall and limit potential bias as a result of an incomplete evidence base.  However, it is necessary to strike a balance between sensitivity (comprehensiveness) and precision (relevance)

Sensitivity - the number of relevant results divided by the total number of relevant results in existence 

Precision - the number of relevant results identified divided by the total number of results identified 

Increasing the comprehensiveness of a search will reduce its precision and will retrieve more irrelevant results. However, 

" a conservatively estimated reading rate of one or two abstracts per minute, the results of a database search can be screened at the rate of 60–120 per hour
(or approximately 500–1000 over an 8-hour period), so the high yield and low precision associated with systematic review searching may not be as daunting
as it might at first appear in comparison with the total time to be invested in the review."
(Cochrane Handbook for Systematic Reviews of Interventions, 2022, Section 4.4.3)

Once you've developed an answerable question using a relevant framework, spend time identifying all possible synonyms and related terms for each concept.  It is important not to overlook this stage in the search process as you want to ensure your search retrieves as many relevant records as possible.

How do I do this?
  • Brainstorm on a search plan.  There is a template below.
  • Check the keywords, subject terms or MESH terms used in other research papers or systematic reviews. 
  • Use text mining tools to analyse the frequency of terms used in relevant texts.  Here are some examples of popular text mining tools: PubMed Reminer, JSTOR text analyzer, TerMine, RAYYAN
  • Use search filters to get additional terms to incorporate in your search strategy.  Search filters are pre-made search strategies on specific topics.  Here are some examples of popular search filters: PubMed Clinical Queries, Lowitja Institute Lit.Search and the Campbell Collaboration.
  • Check searchRxiv for shared search strategies.

Controlled vocabularies (such as the medical subject headings (MESH) used in Medline and EMTREE headings used in EMBASE) provide an organised approach to the way knowledge is described. 

They are extremely important as they ensure uniformity in the indexing of publications included within a database. Using the same terminology throughout a database creates consistency and precision and helps you to find relevant information no matter what terminology the author may have used within their publication.

Indexing is usually a manual process. Databases such as Medline employ specially trained indexers to read the full-text of each publication then identify all of the concepts covered within the article. These concepts are then translated to the controlled vocabulary used within the database. It is the indexer’s job to ensure that each concept included in the article are identified and assigned a term.

Each database may use different subject headings to describe the same concept.

Here is an example of subject headings for the term "Type 1 diabetes" in the following databases:

Database  Subject heading
Medline (OVID) Diabetes Mellitus, Type 1
EMBASE (OVID) insulin dependent diabetes mellitus
PsycINFO (OVID) Diabetes Mellitus
CINAHL (EBSCO) Diabetes Mellitus, Type 1

Medline, EMBASE, EMCARE, AMED, PsycINFO and CINAHL databases provide a search option to “explode” terms. Exploding terms will broaden/expand your search as they retrieve indexed records for a term, plus other terms which are a derivative (more specific, narrower terms) of the search term. Exploding search terms provides a fast way to find related concepts in a single search.

For example, exploding the term "Diabetes Mellitus, Type 1" in MEDLINE will include records indexed with the MESH headings Diabetes Mellitus, Type 1 and Wolfram Syndrome.

NOTE: Clicking on a MESH heading will display its tree, including the exploded terms.

Keyword searches are extremely important when conducting systematic reviews, and should be used in combination with the relevant subject headings within each of your database searches:

  • Authors may not describe their methods or objectives well and indexers are not always experts in the subject areas
  • The available indexing terms might not correspond to the terms the searcher wishes to use. Cochrane Handbook for Systematic Reviews of Interventions (2011, Section 6.4.5)

The definition of ‘truncation’ is to shorten or cut-off at the end. Truncation is used in database searches to ensure the retrieval of all possible variations of a search term.  All databases allow truncation, but the symbols used may vary, so it is best to check the database help for details.

Databases usually allow words to be truncated either at the end, or internally:

  • Truncating a word at the end ensures that all variations of the word, beginning with a specific root, will be retrieved. This is particularly useful for retrieving singular and plural versions of words.

Be careful not to truncate terms too early, or you may retrieve a high number of irrelevant documents.

Most databases use an asterisk (*) to find alternate endings for terms. For example:

therap* will retrieve words such as:

therapy, therapies, therapist, therapists, therapeutic, therapeutics, therapeutically etc.

  • Truncating within a word ensures that any variations of spelling of a word can be retrieved. For example, pediatrics or paediatrics

Internal truncation is available in some databases, allowing you to search for alternate spellings of words - extremely useful when searching for American and English spellings of words.

For example, using the OVID databases (MEDLINE, EMBASE, PsycINFO, etc), a question mark included within a word can designate zero or one character in that place:

colo?r will retrieve either:

colour or color


Boolean Operators

Boolean operators enable you to link terms together, either to widen a search or to exclude terms from your search results.



Nesting uses brackets ( ) to group complex searches. For example:

diseases AND (chronic OR acute)

will retrieve records containing 

chronic diseases OR acute diseases OR both 

It is also useful to use when combining synonyms. For example: (heart attack OR myocardial infarction OR cardiovascular stroke)


Proximity Operators

Proximity or adjacency searching is similar to using Boolean operators in that you are specifying relationships between 2 or more terms. However, proximity searching allows you to specify the proximity of words to each other.

Some databases allow you to search for words within a specified number of words from each other. For example, physician adj3 relationship will retrieve records that contain the words physician and relationship with up to 2 words in between, in any order. 

physician adj3 relationship

will retrieve 

"physician patient relationship" or "patient physician relationship" or "relationship of the physician"

Proximity operators vary across different databases.  Refer to the quick search guide for health databases for further details. 

You should next think about the limits you intend to apply to your search.

Criteria Questions to ask Advice from the Cochrane Handbook for Systematic Review Interventions (2022, Section 4.4.5)
Time period Will your review be restricted by year of publication, or is it important that you cover all years? "Date restrictions should be applied only if it is known that relevant studies could only have been reported during a specific time period, for example if the intervention was only available after a certain time point."
Language Should you restrict to English language publications only? Language restrictions should not be included in the search strategy as there is a risk of introducing bias.  
Publication type Are you restricting your search by publication type? Restrictions of publication types, such as excluding letters or comments, are generally not advised as they may contain important additional information not reported elsewhere.
Geography Are there any geographic considerations to include in your search strategy?  For example, if you were researching Chinese herbal medicine you would need to consult Chinese literature.

Search filters (also known as hedges) are pre-made and pre-test search strategies designed to retrieve specific search results. They are often developed for specific databases and use a combination of subject headings, keywords and truncation. They can be useful for checking additional terms for your search strategy and limiting results by study design or clinical query type etc. 

Always evaluate a search filter before using it to determine if if suits your specific needs.

Testing your search 

Searching is an iterative process, so testing your search will help you to find the right balance.  Some factors to consider include: 

  • Could the use of proximity operators improve the search?
  • Are the keywords used in the search appropriate?
  • Can you identify terms that are adding to irrelevant results?  Is there a risk of missing relevant results if these terms are removed from your search?
  • Are the subject headings used in the search also represented by relevant keywords?
  • Has the search strategy been adapted for relevant databases, especially those that don't use subject headings?
  • Have you identified a set of highly relevant articles?  Do all of these articles get retrieved in your search strategy?  This will help to determine the effectiveness of the search.

 Use this checklist to review your search strategy.  The Smart Searching testing module also has some useful tips.  

When to stop?

There are no hard and fast rules but these factors such as time, effectiveness of the search and whether continued searching yields any significant new results could influence your decision. 

You should rerun your search preferably within 6 months prior to publication to identify any new research to include in your review. One way to do this is to update the date limits of the last relevant line of your saved searches. The attached document - "codes for updating database searches" - shows how to do this in the most commonly searched databases.  

Once your updated search is run, import your results to a new group in Endnote e.g. "updated search". Deduplicate your Endnote library.  Records remaining in the 'updated search" group are the newly identified and potentially eligible studies. 

If potentially eligible studies are identified but not included in your review, they "will need to be reported as references under "studies awaiting classification..." (Cochrane Handbook for Systematic Reviews of Interventions, 2022, Section 4.4.10)