Searching and Screening for Systematic Reviews

A systematic review is defined by its search. If the strategy misses relevant studies, no amount of careful analysis afterwards can recover them, and reviewers judge the search on whether another team could run it again and get the same records.

A sound strategy combines controlled vocabulary, such as MeSH in MEDLINE and Emtree in Embase, with free-text terms for each concept in the question, joined with Boolean operators and adapted to each database's syntax. Health reviews usually search at least MEDLINE, Embase and CENTRAL, adding subject databases where relevant: CINAHL for nursing and allied health, PsycINFO for mental health, SPORTDiscus for exercise science. Trial registers and other grey literature reduce the risk that only positive results are found. Peer review of the strategy against the PRESS checklist catches errors before they propagate.

Screening then narrows thousands of records to the included set. Duplicates are removed first, which is harder than it sounds when the same study appears with different formatting across databases. Titles and abstracts are screened by two reviewers independently, followed by full texts, with disagreements resolved by discussion or a third reviewer and reasons for exclusion recorded at the full-text stage. Tools such as Rayyan, Covidence and EPPI-Reviewer make this traceable.

Reviewer comments on searches tend to be specific. A single database, a strategy published only as a list of keywords, no date of last search, no trial registers, or a flow diagram whose numbers do not add up. Search strategies also date quickly: a search more than twelve months old at submission is often sent back for an update, so it pays to plan the timeline with a final re-run in mind.

The notes in this topic cover translating a question into a strategy, documenting searches to the PRISMA-S standard, handling deduplication, calibrating screeners, and producing a PRISMA flow diagram whose numbers reconcile.