Skip to Main Content

Search filters

Search filters overview

Search filters (also known as 'hedges') are pre-made search strategies you can incorporate into your search, saving you time.

Filters:

  • are usually either methodological or topic-based
  • can also be referred to as clinical queries or optimal search strategies
  • may be formally 'validated' to meet a minimum level of article retrieval sensitivity and precision.

This video from Yale University provides an introduction to pre-made search strategies and suggestions of where to find validated filters.

Video Length: 10:57

  • Building systematic search strategies is complex and time consuming.
  • It can be worth searching for pre-made search strategies that match your concepts.
  • Pre-made search strategies are known as 'filters' or 'hedges'.
  • They may also be called 'optimal search strategies', 'quality filters', or 'clinical queries'.
  • Topic filters are designed to find resources relevant to a specific topic.
  • Methodological filters are designed to find specific study design types.
  • These can save time, but there are things to consider before deciding to use existing filters.
  • Some filters are expert informed, using scientific validation techniques.
  • Validated sets have been examined to determine their sensitivity and specificity.
  • If a filter has a sensitivity of 99%, it will retrieve 99% of all articles known to be relevant in a gold standard test set.
  • A filter with high sensitivity will be very comprehensive, but will require you to filter out some irrelevant results.
  • If a filter has a specificity of 99%, it filters out 99% of articles that are known to be irrelevant in the gold standard test set.
  • A filter with high specificity will be extremely focused, but may miss some relevant results.
  • In most cases, you will end up choosing a filter that has a balance of sensitivity and specificity.
  • Not all filters are validated - this doesn't mean they are poor quality, but you will have to carefully evaluate the filter yourself.
  • If a filter has been validated, consider:
    • the sensitivity and specificity score
    • the database used for validation
    • if you will need to remap the filter to a new database
    • the subject area of the gold standard test set
    • the date of the validation
  • For details about where to search for filters, please continue reading the information on this page.

More information

Filters: considerations and commonly used

Consider whether the search filter is:

  • current - subject headings and field codes can change over time
  • relevant to your review topic and inclusion criteria
  • comprehensive and transparent
  • validated - what methods were used to develop it?
  • created by an expert information manager/librarian or research group
  • cited - search filters should be cited when used. If a filter has not been cited many times, this may indicate issues with quality/acceptability
'All search filters and all search strategies are compromises and an assessment of the performance of filters for your own research should always be made.' - 'What is the ISSG Search Filter Resource?', ISSG Search Filters Resource

 Use a checklist or ranking tool:

There are a couple of approaches to limiting to randomised controlled trials: search filters, and machine learning tools such as that available via Covidence.

Using search filters is only recommended for certain databases, and where numbers of results prohibit reasonable screening loads. It is not needed for Cochrane Library which is already limited to trials in humans.

The Cochrane search filters (provided for PubMed, MEDLINE via Ovid, Embase via Ovid and CINAHL via EBSCOhost) have the humans and RCTs filtering built in together; this can be separated out and only the appropriate section applied if needed for your review.

See the Sources of filters box on this page for other available filters for randomised controlled trials (e.g. CADTH, SIGN).

The Cochrane Randomized Controlled Trial (RCT) classifier is a machine learning tool integrated into Covidence that tags records as 'Possible RCT' or 'Not RCT'.

The following are possible approaches for limiting your results to human studies, without excluding those where both humans and animals are mentioned, and addressing limitations in 'Humans' database limits (typically a checkbox or menu option, where available).

For the purposes of demonstration line number 10 is taken as the final search line with all of your concept sets combined.

MEDLINE via Ovid
10   your final set of results before applying limits (e.g. topic sets 3 AND 9)
11   10 NOT (exp animals/ NOT humans.sh.)
Source: Cochrane Handbook for Systematic Reviews of Interventions, Technical Supplement to Chapter 4: Searching for and selecting studies, Section 3.6.1 The Cochrane Highly Sensitive Search Strategies for identifying randomized trials in MEDLINE

Embase via Ovid
10   your final set of results before applying limits (e.g. topic sets 3 AND 9)
11   (rat OR rats OR mouse OR mice OR swine OR porcine OR murine OR sheep OR lambs OR pigs OR piglets OR rabbit OR rabbits OR cat OR cats OR dog OR dogs OR cattle OR bovine OR monkey OR monkeys OR trout OR marmoset$1).ti. AND animal experiment/
12   Animal experiment/ NOT (human experiment/ OR human/)
13   11 OR 12
14   10 NOT 13
Source: Cochrane Handbook for Systematic Reviews of Interventions, Technical Supplement to Chapter 4: Searching for and selecting studies, Section 3.6.2 Search filters for identifying randomized trials in Embase

CINAHL via EBSCOhost
10   your final set of results before applying limits (e.g. topic sets 3 AND 9)
11   MH animals+
12   MH (animal studies)
13   TI (animal model*)
14   S11 OR S12 OR S13
15   MH (human)
16   S14 NOT S15
17   S10 NOT S16
Source: Cochrane Handbook for Systematic Reviews of Interventions, Technical Supplement to Chapter 4: Searching for and selecting studies, Section 3.6.3 Cochrane CINAHL Plus filter

While it is relatively common to have English language as an inclusion criterion, the best practice recommendation is to manage this in the screening rather than applying it during the search by limiting/refining the search to English language. The following support and expand on this:

Can't find a quality, recognised filter for your database and platform/interface?

In this case you can either:

  • not use a filter for that database and leave it to the screening (most sensitive approach)
  • create your own filter drawing on other strategies, testing and the review team's expertise

The second option is most commonly adopted when the number of records is problematic for screening given your available resources.

Note: If you edit or update a validated filter, or apply it outside its tested database or subject area, it is no longer considered validated – so you cannot guarantee the level of sensitivity and/or precision.

Guidelines and standards

'If published approaches such as search filters…or search strategies from other systematic reviews, were used, cite them. If published approaches were adapted…note the changes made.' - PRISMA 2020 Explanation and Elaboration, p. 7


Other standards

Sources of filters