How To Do A Search Strategy For A Systematic Review In Medicine | No-Nonsense Steps

Guide / By

Build a reproducible plan: frame the question, map terms to headings and text words, search multiple sources, log steps, and report with PRISMA.

Medical questions rarely fit in a single keyword. A solid search strategy turns a messy topic into a traceable set of steps that any reviewer can repeat. This guide sets out practical moves that save time and reduce bias.

Search Strategy For A Systematic Review In Medicine: Step-By-Step

Systematic Review Search Steps And Evidence You Keep
Step What You Do Evidence You Keep
1. Frame the question Write a clear review question using PICOS (Population, Intervention, Comparator, Outcomes, Study design). Protocol text; inclusion and exclusion notes.
2. Register the plan Record the protocol in a public place. Registration ID; version date.
3. Pick sources Choose core databases and registers that fit the topic and study designs. Full list of sources with platforms.
4. Map concepts List concepts and synonyms. Add spelling variants, acronyms, and drug or gene names. Concept table with term lists.
5. Add subject headings Find MeSH, Emtree, and other controlled terms for each concept. Heading lists with scope notes.
6. Build logic Combine headings and text words with Boolean, proximity, phrase marks, and truncation. Full search strings per database.
7. Pilot test Run small tests; check if known sentinel studies appear. Seed set list with PMIDs or IDs.
8. Translate Adapt syntax for each platform; keep the concept logic the same. Translations with date stamps.
9. Add grey sources Search trial registers, preprints, theses, and guidelines. Links and query notes.
10. Use filters wisely If you must, apply tested study design filters and only safe limits like language or year when justified. Named filters; rationale.
11. Run and export Export all fields, abstracts, and IDs in a dedup-friendly format. RIS or XML files; export logs.
12. Deduplicate Remove duplicates with clear rules; keep a record of matches. Dedup report; match pairs.
13. Peer review Ask an information specialist to run a PRESS check. Completed checklist.
14. Re-run before close Repeat the searches near submission to catch new studies. Top-up date and counts.

Define The Question With PICOS

Start with a plain sentence that states who, what, compared with what, and which outcomes matter. Turn that sentence into PICOS fields. This keeps scope tight and drives term picking later.

Map Concepts To Subject Headings And Text Words

Use controlled vocab where it exists, then add free text to catch new or unindexed records. For diseases and procedures, grab MeSH or Emtree headings and check scope notes. For newer topics, include brand names, gene symbols, lay terms, and spelling variants. Add British and American spellings, hyphenation, and common abbreviations.

Build Boolean And Proximity Logic

Group synonyms with OR. Link separate concepts with AND. Use quotation marks for exact phrases when the platform allows it. Add truncation with care so you don’t pull noise. In Ovid and Embase, proximity operators help join terms that sit near each other. Keep logic readable with line numbers or indents.

Ovid MEDLINE (sample respiratory syncytial virus vaccine string)
1. Respiratory Syncytial Viruses/ OR Respiratory Syncytial Virus, Human/
2. (RSV OR "respiratory syncytial").ti,ab,kf.
3. 1 OR 2
4. Vaccines/ OR Immunization/
5. (vaccin* OR immuni#ation).ti,ab,kf.
6. 4 OR 5
7. 3 AND 6
8. randomized controlled trial.pt.
9. 7 AND 8

Pilot Test Against A Seed Set

Before you scale up, run the draft in one database and check whether known sentinel studies appear. If they don’t, find which concepts block recall and loosen them.

Choose Databases And Registers

Pair a biomedical engine with topic-specific sources. MEDLINE and Embase are a common duo for drug and device questions. CENTRAL helps for randomized trials. Add CINAHL for nursing and PsycINFO for mental health.

Include Grey Literature And Trial Registers

Add clinical trial registers, theses, conference abstracts, and preprint servers. This narrows publication bias and catches studies still in progress. Keep queries repeatable, then export counts and links.

Handle Filters And Limits Safely

Study design filters from trusted groups are safer than ad-hoc strings. Date or language limits need a reason in the protocol. Avoid “human” limits that can hide animal-only indexing quirks. When in doubt, search broad and screen.

Run The Searches And Export Data

Export with full citation fields and record IDs to make deduping straightforward. Save the exact files you exported, the number of records pulled per source, and the date.

Document Everything For Reproducibility

Paste each full strategy, platform, and date into an appendix. Add the PRISMA flow with counts by source and the reasons for exclusions at screening. Readers should be able to re-run your exact steps without guesswork.

Peer Review With The PRESS Checklist

A second set of eyes catches gaps in headings, missed synonyms, syntax slips, and unsafe limits. Ask a librarian or search specialist to run a PRESS check and sign off before you lock the strategies.

Maintain Version Control And Re-Run

Name strategy files with dates and versions. Set calendar reminders to re-run searches before final data extraction and again before submission. Add new records to the library and document the top-up in the PRISMA flow.

How To Build A Medical Systematic Review Search Strategy: Common Pitfalls

Relying Only On Free Text

Free text can miss indexed records. Pair it with the right subject headings. Check the explosion rules and subheadings so you don’t lose scope.

Forgetting Variant Spellings And Acronyms

Spellings shift across regions and journals. Add British and American forms, hyphen variants, and common acronyms. Watch for overlap with unrelated terms and patch with proximity or field tags.

Over-tight Phrase Searching In PubMed

Phrase marks in PubMed can switch off automatic term mapping. When that happens, you can miss mapped MeSH and narrower terms. Use phrase marks only where they add precision you can’t get another way, and check the details page to see how PubMed read your query.

Unclear Study Design Logic

Homemade “trial” strings often miss actual randomized trials. Use tested filters from trusted sources or the publication type field where available. Keep design terms separate from the PICO concepts so you can tune recall.

Heavy Use Of Limits

Stacking many limits can hide good studies. If a limit trims too much, move that choice to the screening stage and log the reason there.

Searching One Database Only

No single source holds the full set. Pair at least two major biomedical databases and add subject sources as needed. Back that up with trial registers and citation chasing.

Skipping Grey Literature

Posters, theses, and preprints often include early results. A short, repeatable set of queries in those sources fills gaps and shows that you cast a wide net.

Poor Translation Across Platforms

Each platform has its own quirks. Field tags, proximity, and truncation differ. Keep a master concept grid, then translate with care and test each line before you run the full string.

Database And Register Menu

Where You Search And What Each Source Adds
Source Why Use It Notes
MEDLINE (PubMed or Ovid) Core biomedicine with MeSH indexing. PubMed adds in-process records; Ovid offers rich field tags.
Embase (Elsevier) Strong drug and device indexing with Emtree. Includes many conferences; proximity tools help.
Cochrane CENTRAL Trials index drawn from multiple sources. Useful for randomized trial questions.
CINAHL Nursing and allied health coverage. Great for practice and setting terms.
PsycINFO Mental health and behavior studies. Good for scales and diagnostic terms.
ClinicalTrials.gov and WHO ICTRP Registered and ongoing trials. Use short, concept-level queries.

PRISMA Reporting, Data Management, And Export

Log record counts by source on the day you run each search. Save raw exports and your dedup rules. During screening, track reasons for exclusion with terse labels. When you write the methods, include the full strings in an appendix and add a PRISMA flow that matches your counts. Align dates and numbers across the text, flow chart, and appendix.

Write Strategies People Can Read

Structure long strings. Use line numbers, comments, and spacing so another person can follow the logic. Keep one master file per database and date.

Balance Recall And Precision

A search that pulls every record wastes time; a search that misses half the field risks wrong answers. Tune a single concept at a time. If you need more recall, add synonyms and loosen proximity. If you need more precision, switch off truncation or tighten proximity.

Deduping That Won’t Bite You

Pick one tool and write your match rules. PMID, DOI, title, and first author pairs work well. Keep both the log and the dropped IDs.

Templates And Handy Snippets

Concept Grid Starter

Concept A (Disease): MeSH headings; Emtree headings; text words and acronyms
Concept B (Intervention): Headings; device or drug names; text words
Concept C (Study design): publication types; validated filters
Limits: year or language only when justified; never by impact factor or country
Sources: database list; registers; grey sources
Outputs: file types; fields; counts by source

Last Checks Before Submission

  • Run a PRESS check by a second searcher and keep the signed form.
  • Re-run the full strategy within four weeks of final screening.
  • Update the PRISMA flow and match all dates and counts.
  • Attach full strategies, platforms, and run dates in an appendix.
  • State databases and registers by name in the abstract and methods.

With a traceable plan, tested logic, and clean reporting, your search stands up to scrutiny and saves the team hours during screening and synthesis. Follow these steps and your search will be clear, thorough, and easy to audit later on. No shortcuts needed.

Helpful links you can cite in your methods: the Cochrane Handbook; the PRISMA 2020 flow diagram; and the PRESS checklist.