For a systematic review, the search must be reproducible: another researcher runs your string and obtains your number of results. Typing terms into the box does not meet that standard.
Understand MeSH
PubMed indexes articles against Medical Subject Headings, a controlled vocabulary assigned by human indexers. Searching a MeSH term retrieves articles about that concept regardless of the words the authors used.
Find the right heading in the MeSH database first. “Heart attack” maps to Myocardial Infarction; searching the phrase alone misses everything indexed under the formal term.
Combine MeSH with free text
MeSH indexing lags publication by weeks or months, so recent articles are missed by a MeSH-only search. Always combine both:
("Myocardial Infarction"[Mesh] OR "myocardial infarction"[tiab] OR "heart attack"[tiab])
The [tiab] tag searches title and abstract, which is more precise than an untagged search across every field.
Build with Boolean structure
- OR within a concept — every synonym for the same idea
- AND between concepts — each PICO element as its own block
- NOT sparingly — it excludes relevant records more often than authors expect
Build each concept block separately, check its yield, then combine. A search that returns four results usually has one over-restricted block, and building in pieces shows you which.
Useful field tags
[tiab]title and abstract ·[ti]title only[Mesh]exact heading ·[Mesh:NoExp]without narrower terms[au]author ·[ta]journal ·[dp]date of publication*truncation, as inrandom*for randomise, randomised, randomisation
Filters, and their risk
Publication-type and date filters are legitimate if pre-specified. Language filters introduce bias and should be justified. Never apply a filter you did not declare in the protocol.
Document it for the paper
Record the database and interface, the complete search string exactly as run, every filter, the date searched, and the number of results. This goes in a supplementary file — PRISMA item 7 requires it.
Then search at least two more databases. PubMed alone is not sufficient coverage for a systematic review in almost any field.