How to Turn a PICO Question into a Search Strategy

A PICO question is a clinical question broken into Population, Intervention, Comparison, and Outcome. A search strategy is the translation of those four elements into terms a database can actually match. The work between the two is conceptual: you decide which elements are searchable, which improve retrieval, and which should be left out of the query so they do not silently exclude relevant studies.

This guide walks through that translation for clinicians and medical researchers building a focused evidence search. It uses one worked example, gives diagnostic checkpoints at each stage, and ends with stop and escalation rules. The aim is a reproducible strategy you can defend in a guideline or scholarly methods section.

Cochrane's guidance is to plan the search from the review's eligibility criteria. That ordering matters. If you begin with keywords, you tend to search for the words you already associate with the topic. If you begin with eligibility criteria, you search for the studies you would actually include.

Write the PICO elements as inclusion criteria first. For each element, note whether it is a hard requirement or a desirable feature. A hard requirement belongs in the search. A desirable feature may belong in screening instead.

Checkpoint 1: Can a colleague read your PICO and predict which studies you would accept? If not, the search cannot be better than the question.

Take this question: In adults with community-acquired pneumonia treated as outpatients, does a short antibiotic course compared with a standard course reduce treatment failure?

• Population: adults with community-acquired pneumonia managed as outpatients
• Intervention: short-course antibiotic therapy
• Comparison: standard or longer-course antibiotic therapy
• Outcome: treatment failure

This is a deliberately narrow example. The population has a care setting, the intervention is a duration concept rather than a named drug, and the outcome is a single clinical endpoint. Narrow examples expose the translation problems clearly.

For each element, build a concept block: the idea plus the terms that express it. Then decide how the block will enter the query.

The population block carries the disease and the setting. "Community-acquired pneumonia" is a phrase with established subject headings. "Outpatients" is a setting concept that may appear as free text, as a study design feature, or not at all in the abstract. Treat the setting as a term to try, not an assumption.

"Short course" is a duration concept. It may be expressed as a number of days, as "short-course," or as a named regimen. The comparison may be described as "standard," "longer," or simply as the comparator in a trial. Duration concepts are the hardest to capture with a single term, so this block needs the widest term variation.

"Treatment failure" may appear as a defined composite, as readmission, or as a clinical deterioration. If the outcome is not indexed consistently, it is often better used as a screening criterion than as a search block.

Checkpoint 2: For each block, write down whether it is (a) reliably indexed, (b) free-text only, or (c) better handled at screening. Blocks in category (c) should usually be dropped from the query.

Not every PICO element belongs in the search string. Adding a block narrows the result set, and a narrow block built on inconsistent terminology can remove relevant studies before you ever see them.

A practical rule:

• Keep population and intervention blocks when both are central and well expressed.
• Keep the comparison block only when the comparison is a defining feature of the question.
• Drop the outcome block when the outcome is reported under many labels; screen for it instead.
• Drop any block whose terms you cannot state without guessing.

For the worked example, population and intervention are strong blocks. The comparison is a duration concept and can be folded into the intervention block. The outcome is better handled at screening.

Checkpoint 3: Run the strategy with and without each optional block. If removing a block adds many irrelevant records but no new relevant ones, the block was doing useful work. If removing it adds relevant studies, the block was too narrow.

MeSH is NLM's controlled and hierarchical vocabulary for indexing and searching biomedical information. Preferred headings, entry terms, scope notes, qualifiers, and tree positions can be used to expand or focus a PubMed strategy. That hierarchy is the main reason to use MeSH: a single heading can retrieve records indexed under narrower concepts.

Free-text keywords capture what the heading misses: new terminology, drug names, and phrasing used in titles and abstracts. Cochrane recommends combining free-text terms with subject headings, which is the standard structure for a sensitive strategy.

A workable pattern for each block is: MeSH heading OR its entry terms OR free-text synonyms, combined with OR inside the block, then combined with AND across blocks.

PubMed supports field tags, phrase searching, filters, Advanced Search, saved searches, Similar Articles, citation lookup, and links to full text when available. Field tags let you decide whether a term must appear in the title, the abstract, or the indexing. Restricting a term to the title is a strong narrowing move; use it only when you are confident the term is consistently used there.

PubMed's search details reveal how terms were translated and combined. This is the single most useful diagnostic in the whole process. If a phrase was split, a term was mapped to an unexpected heading, or a block was combined incorrectly, the search details show it.

Checkpoint 4: Open the search details and confirm that every block you intended is present, that OR and AND are grouped as you planned, and that no term was mapped somewhere you did not expect.

Before you trust a strategy, test it against studies you already know should be found. Build a small set of known relevant papers, including at least one that is indexed and one that is recent. Run the strategy and check whether each known study appears.

This is the most direct test of sensitivity. A strategy that misses a study you already know about will miss others you do not know about.

When a known study is missing, work backwards:

1. Open the record and read its MeSH terms and abstract wording.
2. Compare those terms with your blocks.
3. Identify which block excluded it.
4. Broaden that block, or move that concept to screening.

Repeat until the known set is retrieved. Then re-check precision, because broadening a block will bring in more noise.

Checkpoint 5: Record the known-study set and the result of each validation run. This record is what makes the strategy reproducible for a guideline or methods section.

Use this checklist for each new PICO question.

A PICO translation is iterative, and the iteration is easier when search and reading happen in one place. WisPaper's scholarly search interface is the natural starting point for running concept blocks and inspecting what comes back. For readers who want a public entry point before signing in, the Scholar Search page explains the paper-finding workflow.

Once candidate studies are identified, save them so the known-study set and the final included set stay together. My Library is where papers can be saved or uploaded and organized. That matters for validation: your known-study set should be a stable list you can rerun against, not a set of browser tabs.

When the question grows into a full review, the search and screening work can be carried in Research Projects, which is the area for creating and managing a project. For questions you want to ask directly of the papers you have collected, Library QA supports asking questions based on papers uploaded to My Library or added from search results.

Cochrane recommends documenting the search process, deduplicating records, screening in stages, linking multiple reports of one study, and rerunning searches before publication. Each of these has a search-strategy consequence.

• Documentation: save the final string, the database, the date, and the number of records. Save the search itself where the database supports saved searches.
• Deduplication: expect the same study to appear from multiple sources; deduplicate before screening.
• Staged screening: screen titles and abstracts first, then full text. Your outcome block, if dropped from the query, is applied here.
• Linked reports: one study may have several reports. Link them so a single study is not counted twice.
• Rerun before publication: a strategy validated months ago may need a refresh.

Checkpoint 6: Before you move to full-text screening, confirm that the strategy is saved, the record counts are logged, and duplicates are removed.

A validated search delivers studies, not conclusions. When you appraise what you retrieve, keep four judgments separate.

• Evidence level is about the design and strength of the body of evidence for the question.
• Applicability is about whether the studied population, setting, and intervention match your question.
• Indirectness is about how far the available evidence sits from your exact PICO elements, for example a different care setting or a broader population.
• Study limitations are about the conduct of individual studies.

These are different judgments and should be reported separately. A high-level study can still be indirect for your population; a directly applicable study can still have important limitations. Keeping them apart prevents a strong design label from being read as a conclusion about your patients.

Use these rules to decide when to stop refining and when to escalate.

Stop refining when:

• All studies in your known-study set are retrieved.
• Removing optional blocks no longer adds relevant records.
• The search details match your intended structure.
• The strategy is documented and saved.

Escalate when:

• A known relevant study cannot be retrieved after two rounds of broadening. Escalate to a librarian or information specialist.
• The question spans multiple databases or grey literature. Cochrane recommends using multiple sources; a multi-source search benefits from specialist support.
• The PICO itself is unstable, meaning the population or comparison keeps changing. Return to Checkpoint 1 before searching further.
• The evidence you retrieve is consistently indirect for your population. Escalate the question for review rather than stretching the search to compensate.