Cochrane review authors should contact their Trials Search Co-ordinator before starting a search.
For most Cochrane reviews, the search structure in most databases will be comprised of a subject search for population or condition and intervention together with a methodological filter for the study design, such as randomized trials.
For searches of CENTRAL, do not apply a randomized trial filter and do not limit to human.
Avoid too many different search concepts but use a wide variety of synonyms and related terms (both free text and controlled vocabulary terms) combined with ‘OR’ within each concept.
Combine different concepts with ‘AND’.
Avoid use of the ‘NOT’ operator in combining search sets.
Aim for high sensitivity and be prepared to accept low precision.
Do not apply language restrictions to the search strategy.
Searches designed for a specific database and service provider will need to be ‘translated’ for use in another database or service provider.
Ensure awareness of any retracted publications (e.g. fraudulent publications), errata and comments.
For identifying randomized trials in MEDLINE, begin with the sensitivity-maximizing version of the Cochrane Highly Sensitive Search Strategy. If this retrieves an unmanageable number of references, use the sensitivity- and precision-maximizing version instead.
For update searches, where possible, separate database files should be selected and searched separately for the MEDLINE-indexed records and the non-indexed in-process records.