Systematic Review: Search strategy, using limits & filters, test & review strategy

In order to run your search in a database you will need to combine the different concepts in your topic to form a search strategy.

We do this using Boolean operators to connect your terms. The main operators are AND and OR. Join the terms within a concept with the OR operator, then link the concept sets with the AND operator.

See the online tutorial for more detail: Combine the search terms using Boolean

The tutorial below demonstrates how to develop and perform systematic searches in Ovid Medline database. The techniques demonstrated in the tutorial can be applied to other databases.

Systematic searching for a review

Refining your strategy with search operators

Some useful refinements used in developing search strategies include:

• Phrase searching: Enclosing words in quotation marks tells the database to look for that specific phrase (i.e. words are next to each other and word order is maintained) which can lead to more precise results.  
• Truncation: Truncation is a way of expanding the search by looking for the various endings of a particular word. To use truncation, enter the root of the word with the truncation symbol at the point where differing word endings could start. Avoid truncation where the root word is too short or common as you may get excessive irrelevancies. Truncation operators used in the middle of a word (to replace zero or one letter) may be referred to as wildcards.
• Proximity: Proximity operators (also known as adjacency operators) enable you to specify how closely you want your search terms to be found in relation to one another. They are broader than phrase searching and allow you to adjust the precision of the results. To use proximity place the operator between your terms eg. postnatal ADJ4 depression

Searching in specific fields

The default search in a database will search across multiple key fields. You may wish to narrow your search by searching only in specific fields (such as title and abstract). You may similarly wish to enhance a search by including additional fields that would not be included in a default keyword search (such as country of publication, or institution).

Examples:

• chronic pain.ti,ab   (use a comma to combine multiple fields)
• (chronic pain OR recur* pain).tw,kf    (use brackets when combining terms within a field)

Field codes can differ for databases on the same platform. For example .tw. searches the title and abstract fields in Medline (identical to ti,ab), but searches title, abstract, table of contents and key concepts fields in PsycInfo.

In Ovid databases you can type the appropriate field code into your search, use the Search Fields tab to locate them and add to your search, or click the 'i' icon to the left of the database name to be taken to the field guide listing:

Database syntax may vary across database platforms.

Use the printable PDF tips sheet below to ensure that you select the correct operator for the database that you are using.

Database Limits

Many databases feature an inbuilt set of limiters that may be used to restrict your search. Typically, these can be selected by ticking a box to apply the limits from the database interface. For example, ticking Humans, or ticking English Language.

Limits should only be applied once your search strategy is complete. (Do not apply the same limit multiple times to different search lines).

The use of limits can result in the exclusion of relevant studies, so don't use them if the number of combined search results is manageable without them. If necessary, the use of a search filter in place of a limiter is preferred for systematic reviews.

Search Filters

Search Filters (sometimes referred to as hedges) are a combination of terms (subject headings and keywords) designed to filter search results to those with specific content. They are, in effect, a preset search strategy which can be incorporated into your search to restrict the results.

Search Filters are database/platform specific due to the inclusion of database-specific subject headings and syntax. For example, a study design filter to limit to RCTs, would be quite different for Ovid Medline, PubMed, and Ovid Embase.

Search Filters are often developed by experts and are the most effective way of restricting a search. However, all filters (including those that maximise sensitivity) introduce the possibility of missing relevant results, so don’t use them if the number of combined search results is manageable without them.  

Validated Search Filters
A search filter is ‘validated’ when it has been tested and validated against an independent sample ('gold set') of relevant articles, and the results published. If you make any changes to a validated filter (to translate it into another database for example), it is no longer a validated filter.

Inbuilt Search Filters
Often you will need to add a search filter line-by-line to your search strategy. Some databases however may have inbuilt filters (in addition to inbuilt limits). For example, the screenshot below shows the Cochrane Library's humans filter available in the Additional Limits menu in Ovid Medline:

Filters can be categorised into two main groups:

• By study type - methodological
• By topic or subject.

The following example is a filter used to identify randomised trials in Ovid Medline. This and filters for other databases can be found in the Cochrane handbook, Chapter 4. Search filters 3.6.

Box 3.d Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE: sensitivity and precision maximising version (2008 revision); Ovid format

1. randomized controlled trial.pt.
2. controlled clinical trial.pt.
3. randomized.ab.
4. placebo.ab.
5. clinical trials as topic.sh.
6. randomly.ab.
7. trial.ti.
8. 1 or 2 or 3 or 4 or 5 or 6 or 7
9. Exp animals/ not humans.sh.
10. 8 not 9

See also Cochrane search filters for Embase.com (see translation into Ovid Embase here) and for Cinahl Plus databases.

A systematic approach to searching requires testing the ability of your search to retrieve known articles that represent your research question. This testing process requires you to have a gold set of relevant articles that you want to see in your search results. The steps for testing on a gold set of articles is below:

• Once you have run your search strategy and have a result set, the aim is to ensure that this gold set of articles is included in the results.
• First, you need to see which gold set articles are included in the database you are using. If you are testing an Ovid Medline search strategy, the gold set you use can only contain articles that are indexed in Ovid Medline. The easiest way to check if the article is present is to do a search in the Title field to see if it is found.

• Once you have confirmed that the article is present, you need to see if the article is included in your search results. You can do this by combining your final search strategy results with the gold set article using AND.

• Do this for each article in your gold set. If an article is present in the database but not in your results, you know that the search strategy needs further refinement.

How many articles should you have in your gold set?
The more articles you have, the more evidence and confidence you will have in the strength and appropriateness of your search strategy. For example, testing on 10 articles will let you know that you are on the right track. Testing on 25 articles from a variety of sources will give you much better assurance that you are not going to miss relevant papers due to the chosen terms and parameters of your search.

Reviewing your search strategy can be made easier by using a checklist or guide to help you self-assess for comprehensiveness, effectiveness and errors.

• The PRESS checklist for search strategies is a structured tool that is widely used for peer review of electronic literature search strategies.