PubMed Search String Builder

PubMed can interpret an untagged query through Automatic Term Mapping, often called ATM. That behavior is useful for exploratory searching, but it also means the text typed into the search box is not necessarily the full expression executed. This builder makes its main choices explicit: free-text synonyms carry [tiab] and controlled headings carry [mh]. An explicitly fielded query should be reviewed as its own strategy; do not assume it inherits every mapping or expansion of an untagged phrase.

The title/abstract block searches where authors describe their work. It can recover recent records awaiting indexing, papers using newer terminology and records outside the part of PubMed covered by MeSH indexing. Put alternative expressions of the same concept inside one parenthesized OR group, then connect necessary groups with AND. The hypertension-and-exercise example begins with two concept groups. Keeping comparator and outcome optional can protect recall when those details are inconsistently named in abstracts. Eligibility screening remains the place to check details that cannot be searched reliably.

Quoted phrases and terminal asterisks need inspection. A phrase can behave differently when it is not found in PubMed’s phrase index, and wildcard expansion changes mapping behavior. Wildcard roots need enough leading characters; a broad stem can expand far beyond the intended idea. PubMed’s own proximity form is also distinct: an unordered title/abstract proximity expression can use a quoted term sequence with a distance tag. Ordered or wildcard-containing proximity suggestions are adapted to phrases here and flagged, rather than borrowing adjN or W/n from another database.

Treat the generated strategy as an AI-assisted draft for reviewer confirmation. A well-formed Boolean expression can still miss a field’s preferred terminology, retrieve the wrong meaning of an abbreviation, or make an unnecessary outcome mandatory. Start with a small set of known relevant papers. Check whether each paper would be retrieved by the population and intervention blocks separately, then by the combined strategy. If a paper is missed, inspect its title, abstract and indexing to find the cause before adding more synonyms. More terms are useful only when they represent the intended concept.

Keep eligibility criteria separate from search constraints. A language restriction, human-study filter or study-design filter can remove recent or incompletely indexed records. Apply only justified limits using the destination interface controls, and record exactly what you selected. The saved concept strategy can carry structured limits and exclusions, but unsupported settings appear as adaptation notes and are not silently applied. Excluding records with NOT deserves particular care: a relevant study can mention the excluded population or treatment in its background.

Copy or download the edited query after testing it. Keep a search log alongside the plain-text export with the database and host, coverage segment, vocabulary release where available, date and time, exact executed string, selected filters, result count and platform warnings. The generated text alone is not a complete reproducibility record. Session storage keeps this tab’s draft and interface edits, but it is not a permanent project archive. Ask a librarian or information specialist to review a consequential review search before final execution; this tool does not certify recall, precision or methodological adequacy.