Help

The number of results will be determined by your question and the number of studies that have been conducted in that area. Preliminary (scoping) searching is important to help gauge possible number of results. You may need to change your question or modify your inclusion/exclusion criteria if the numbers are beyond the capacity of your review team.

The comprehensive/exhaustive style of search required for systematic searching will find significantly more records than will remain after screening (typically more than 99% are excluded).

• search result numbers in older reviews may be misleading as publication volumes have substantially increased in recent years
• how big is your review team – how many people will be available to do the screening?
• how much has been published in this area? More published = higher numbers
• how focussed is your question?

The short answer is: it depends!

Some of the factors to consider include:

• the type of review you are undertaking (systematic, scoping, rapid, umbrella, etc.)
• expectations – your disciplinary area, where you plan to publish, your review team, manuals/guidelines you're following
• whether your topic is multidisciplinary or focused on a single discipline area
• how many are reasonable to search and at what point is it likely to be a case of diminishing returns?
• are you planning to include Grey Literature & Other Sources, e.g. theses and conference proceedings?

No one database can cover all the literature for your research question. While databases may claim to comprehensively cover the literature in a specific area, e.g. nursing, pharmacy, or education, there are many different databases out there which may contain unique records.

The Where to search page can help you decide where to search - or you are welcome to consult a librarian.

See Run your search on other databases for information on translating searches.

Refer to Finding the evidence: key search tips for tips on systematic searching and to see common search syntax required for key databases.

You don't need to search both! We recommend using MEDLINE (Ovid) rather than the freely available PubMed as MEDLINE via Ovid gives you much greater control over your search and is specifically designed for systematic searching.

PubMed is underpinned by a search algorithm and is better designed for busy researchers looking for a few results giving the most relevant research, rather than a comprehensive search for everything on a topic.

• Ovid's Medline compared to PubMed - 2021 article produced by Wolters Kluwer (Ovid)
• How to compare Ovid MEDLINE & PubMed - 2019 PowerPoint presentation produced by Wolters Kluwer (Ovid)

Prior to searching additional databases, have you:

• located all relevant subject headings
• searched for all relevant keywords and word variations
• checked scope notes for subject headings and included synonyms under the ‘Used for’ heading
• reviewed other published reviews (and their strategies) in the area
• spoken with your review team or supervisor
• determined whether the number of results is manageable for your review team? If not the question and/or inclusion criteria may need to be adjusted
• had the search critiqued/peer reviewed (by supervisor, review team, librarian)
• tested that the search is locating test/sample articles?

Remember that your draft MEDLINE search may need to be updated as you map across to other databases and discover additional terms.

See Run your search on other databases for information on translating searches.

Your MEDLINE search should be mostly finalised before you consider translation, to save time and prevent errors. Some changes may need to be made if you're translating to another database with an extensive thesaurus, such as Embase, ERIC and SPORTDiscus, as you may discover additional terms that then need to be incorporated back into your 'master' MEDLINE search. For this reason, we recommend translating to databases without thesauri last.

PRISMA-S states 'Database limits are not recommended for use in systematic reviews, due to their fallibility'. Ideally this is left for screening because of the risk of losing potentially relevant material e.g. due to incorrectly or poorly indexed database records. Some strategies (such as 'Humans' search filters) have been developed to address concerns with the 'checkbox' limit options in databases.

For example, very new records in PsycInfo will not be found if an English language limit is applied as no language information has been included at this point.

Language bias has also been raised as a concern with using 'English Language' limits, with recommendations including listing excluded non-English language studies that might otherwise appear to meet inclusion criteria, and using tools such as Google Translate (Pieper & Puljak 2021). For pragmatic reasons, however, a review group might choose to use certain database limits, in which case this needs to be clearly reported and justified with limitations acknowledged.

While not required by all databases and interfaces, capitalisation of search operators (aka connecting terms/Boolean operators) is required by some – and it can be difficult to work out which! For this reason we suggest capitalising these by default. It has the added benefit of helping to clearly distinguish your operators from your terms.

There are a few possible reasons that you are seeing zero results for a search line on the Ovid platform (MEDLINE, Embase, PsycINFO, etc.). If this happens, please check the following:

• Have you made a spelling mistake or a syntax mistake?
• Have you translated your subject headings? MEDLINE uses Medical Subject Headings (MeSH), whereas Emcare and Embase use the Emtree, and APA PsycINFO uses the Thesaurus of Psychological Index Terms. As a result, subject headings will vary between these databases, even though you’re on the same platform (Ovid). You need to check your subject heading lines when translating your search between databases and update to the appropriate term(s). See the Run your search on other databases page for more information.
• Check your keyword lines – are you using correct field codes within the database?

Of course, some terms may not be present within the database. You may need to reconsider your search terms, or whether the database is appropriate for your topic.

Available 'filters' (pre-made search strategies) can be valuable in designing your search. If these have been developed using a gold standard and validation, reported performance measures such as sensitivity and precision will only apply if you use the strategy in the published form. Due to frequent changes in database interfaces and thesauri, however, you may find a published filter can't be used in that form and choose to make minor changes.

Cite any published strategies used and report any adaptations made (PRISMA 2020 requirements). Alternatively, you can use published strategies as sources of inspiration for your own strategy which you will then need to independently performance test. See the Search filters page for more.

Many databases have specific, unique features that offer advantages in crafting a 'best balance' search strategy. They may also be more up-to-date.

For example, a multidisciplinary database might advertise that it includes 'all the records from MEDLINE' but does not provide the option to search using the MeSH thesaurus; using the more sophisticated features of MEDLINE via Ovid may locate relevant records you would otherwise have missed.

'Portals' such as the WHO's International Clinical Trials Registry Platform (ICTRP) are limited to searching on the standard fields that all contributing sources share, and there can also be a delay before records are loaded.

PRISMA 2020 lists the following as ‘essential elements’: date last searched; database name (e.g. MEDLINE); interface (aka ‘platform’) you accessed the database through (e.g. Ovid); database date coverage where available (e.g. 1946 to January 19, 2022). This is so that other researchers can accurately reproduce your search strategy, and can draw informed conclusions about how comprehensive it is.

Libraries have differing subscriptions to databases via a range of interfaces, so being specific about both the database and the platform is good practice; for example, state: 'ERIC (ProQuest)' not just 'ProQuest'. If you search larger collections which have varied parts (depending on your selections, library subscriptions, and changes made by companies over time) state the parts you searched, e.g.: Web of Science (Science Citation Index Expanded – 1975-present, Social Sciences Citation Index – 1975-present); ProQuest Sociology Collection (Applied Social Sciences Index & Abstracts (ASSIA) (1987 - current), Sociological Abstracts (1952 - current)).

Current guidance is unclear on how to manage and report all aspects of your 'other sources' search (i.e. sources and methods other than traditional literature database searching).

Key considerations:

• be transparent about what you do, such that your approach can be understood and reproduced, and note any limitations
• you will need to weigh principles of best practice (such as independent dual screening) against practical considerations

Potential barriers/issues:

• you may not be able to 'batch export' records from a source (e.g. website or search engine search) into a reference manager to allow tracking and screening
• the number of records and resources available may make manual creation of records in a reference manager unfeasible
• accurate numbers of results for searches may be unclear or unavailable, making it impossible to report number of results identified

Decisions to make:

• how will you 'bound' (limit) the search to ensure it is feasible? For example, you could establish pre-specified limits to your search (such as stating that the first x number of results/pages/issues will be screened))
• will you record the source in the 'databases and registers' or 'other methods' section of the flow diagram? If a source has the features of a traditional literature database - such as advanced search, clear search results numbers, refining options, and batch export allowing you to easily track, report, export, and screen - you could report it in the 'databases' section, even if the material it includes is mostly 'grey' (e.g. ProQuest Dissertations and Theses; Web of Science Conference Proceedings Citation Index)
• how will you report numbers in the flow diagram? The PRISMA 2020 flow diagram template has no provision for 'first pass' under 'Identification of new studies via other methods', implying that these studies are not part of the standard screening workflow. Rethlefsen et al. 2022, however, recommend reporting numbers of records 'identified' separate to numbers of reports assessed as 'eligible', where feasible. This distinction will ideally be clear in your flowchart or detailed in the supplementary material

For more see Record and report