How should a PubMed search query be written?

A concrete task with RigorMed

1. Turn a medical question into an editable query and search record

   These are suggested task inputs and outputs to request, not a new execution, a customer outcome or a guarantee of completion.

   Materials to provide: A specific research question, key concepts and one or two known relevant publications.

   Requested outputs / delivery goal: A query draft and a concepts and synonyms table; a search record is compiled only after an actual test search.

2. An example request

   First list concepts and synonyms, draft a Boolean query to test in PubMed, and explain fields and filters. After testing, use actual Search Details, dates and PMIDs to compile the search record. Do not turn non-retrieval into a claim that no study exists.

3. What still needs review

   Direct searching and model-assisted close reading are separate steps. Citations or abstracts are not lawfully obtained full text; a formal review still requires a complete search protocol.

Break the question into searchable concepts

Suppose you want studies about hypertension and exercise. The table illustrates concept building; it is not a formal search strategy with validated recall.

A starting expression is (hypertension OR "high blood pressure") AND (exercise OR "physical activity"). Parentheses define the scope of OR; AND, OR and NOT are PubMed Boolean operators. This example only illustrates term combination. After entering it, open Search Details to check the actual translation and test recall with relevant papers.

PubMed applies Automatic Term Mapping to untagged terms. Quotation marks do not guarantee an exact phrase match: when a phrase is absent from the phrase index, the system may ignore the quotation marks and remap the terms. Do not infer what was searched solely from the characters you entered.

Six steps from an initial search to verification

1. Specify the question to answer

   State the target population, disease or exposure, intervention or comparison, outcome and study design. Decide which concepts must be searched and which are better assessed during manual screening. Clinical literature searching cannot rely on a disease name alone.

2. Find synonyms for each core concept

   Identify common English expressions in titles, abstracts and keywords of known relevant papers. When subject headings are needed, check their exact names and hierarchy in the official NLM MeSH Browser. MeSH and title/abstract free text can complement one another. Do not invent headings or assume newly added records have already been indexed with MeSH.

3. Combine broadly before narrowing

   Use OR within synonym groups and AND between core concept groups, preserving grouping with parentheses. NOT, excessive AND combinations, specific article types and date restrictions can exclude relevant work and should be used only with an explicit rationale.

4. Check Search Details and field scope

   Inspect PubMed Search Details after searching to confirm automatic mapping, phrase parsing and final field scope. To restrict terms to titles or abstracts, use official field tags such as [tiab]. Tags restrict the searched fields and turn off automatic mapping for that term; more tags are not inherently better.

5. Test known papers and boundary cases

   Choose one or two papers known to be relevant and check whether they are retrieved. Sample irrelevant results from the first few pages to judge whether synonyms or concepts should be added or removed. Record why papers were missed instead of considering the total result count alone.

6. Preserve a reviewable search record

   Record the database, complete query, execution date, actual filters, PMIDs checked and exclusion reasons. Before finalizing a manuscript or formal review, choose reporting guidance appropriate to the study type. Retain every search-method version and the reasons for changes.

Checklist before reporting search conclusions

To describe the current state of research, retain at least the following evidence:

• Are the research question and core concepts that must be searched clear?
• Have each concept's synonyms, abbreviations and official MeSH terms been checked?
• Do OR, AND and parentheses express the intended logic?
• Have automatic mappings, phrases and field scopes in Search Details been inspected?
• Are known relevant papers retrieved, and have reasons for missed records been analyzed?
• Are dates, filters, the full query and PMIDs retained?
• Are databases, registrations and unpublished evidence outside the search scope described as limitations?

'Not found' means only that the chosen database, date, query and filters did not retrieve a study. It does not establish that no research exists. A PubMed record also does not establish free full-text access, endorsement of quality or a clinical recommendation.

Frequently asked questions