How To Find Keywords For A Medical Literature Review

Map PICO terms, add MeSH and synonyms, test in PubMed, and refine until top hits match your topic.

Finding the right search terms can make or break a medical literature review. Good keywords bring the most relevant studies to the top, cut screening time, and make your methods reproducible. This guide shows a clear way to move from a clinical question to a database-ready strategy that works across PubMed, Embase, and CINAHL.

You’ll start with a clinical framing like PICO, turn those concepts into controlled vocabulary and free-text terms, stitch them together with Boolean logic, and tune until top results match your topic.

Finding Keywords For A Medical Literature Review: A Step-By-Step Plan

Start by framing your question. These quick prompts help you name core concepts before you search.

Model	Best Use	Prompt You Can Fill
PICO	Therapy or diagnosis questions	Population: __; Intervention: __; Comparator: __; Outcome: __
PEO	Risk or aetiology questions	Population: __; Exposure: __; Outcome: __
SPIDER	Qualitative questions	Sample: __; Phenomenon: __; Design: __; Evaluation: __; Research type: __
PCC	Scoping reviews	Population: __; Concept: __; Context: __
ECLIPSE	Health policy questions	Expectation: __; Client group: __; Location: __; Impact: __; Professionals: __; Service: __

Turn The Question Into Seeds

Write a short sentence that captures the review aim, then pull out 2–4 concept buckets from it. Under each bucket, list plain-language words users and authors would write. Keep plurals, hyphenations, and UK/US spellings in the same line so you remember to capture them later.

Quick Seed List Example

Hypertension treatment in adults → buckets like “hypertension,” “adults,” and “treatment.” Under “hypertension,” jot hypertension, high blood pressure, and the abbreviation HTN. Under “treatment,” collect therapy, drug therapy, and some class names you expect to see.

Map Seeds To Controlled Vocabulary

Databases tag records with an official thesaurus so similar ideas group together. In biomedicine, the most widely used is MeSH, the Medical Subject Headings from the U.S. National Library of Medicine. Open the MeSH Browser in a new tab and check the preferred term, scope note, tree position, and narrower terms you may want to include or turn off.

Record the preferred Descriptor (e.g., Hypertension) and note useful Qualifiers (e.g., drug therapy, diagnosis).
Copy entry terms listed under the heading; these are built-in synonyms you can reuse as text words.
Decide if you need to “explode” the term to include narrower headings or to switch that off for precision.

For Embase, repeat the step in Emtree; for CINAHL, use CINAHL Headings; for PsycINFO, use the APA Thesaurus. The concept stays the same: pair each bucket with the database’s own subject terms and keep them next to your free-text list.

Expand With Synonyms And Variants

Free-text catches what indexing misses, records not yet indexed, and brand-new terms. Build a neat block of synonyms for each bucket using these quick tricks:

Spelling: tumor/tumour; fetus/foetus.
Morphology: treat, treating, treatment, therapeutic → treat* or therap*.
Abbreviations: HIV, COPD, HTN, RCT. Pair abbreviations with their long forms at least once.
Phrases: use quotes when a phrase matters, like "blood pressure".
Proximity: use a distance operator in PubMed’s title/abstract field when words can drift, like "resistant hypertension"[tiab:~3].

Now you have two layers per bucket: controlled terms and text words. You’ll join synonyms with OR inside a bucket, then join buckets with AND.

Build And Test In PubMed

PubMed supports both MeSH and text-word searching. Use a simple pattern for each bucket, then combine:

( "Hypertension"[mh] OR "Hypertension/drug therapy"[mh] OR (hypertension OR "high blood pressure" OR HTN)[tiab] ) 
AND 
( "Antihypertensive Agents"[mh] OR (antihypertensive* OR ACE inhibitor* OR beta blocker* OR diuretic* OR ARB* OR "calcium channel blocker*")[tiab] )

Tips that save time:

Check “Search Details” in the search builder to see how PubMed translated your words.
Field tags matter: [mh] for MeSH, [majr] for Major Topic, [tiab] for title/abstract, [pt] for publication type, [dp] for date.
Turn off explosion when needed with [mh:noexp] or pin a MeSH/Subheading pair like Hypertension/drug therapy[mh:noexp].
Wildcards (*) and quotes skip automatic term mapping, so balance them with MeSH blocks to keep recall strong.
Proximity works in title/abstract with the format "term1 term2"[tiab:~N].

Run a test first. Open the top 20–30 hits. If the first page isn’t on target, tweak synonyms, add or relax a Subheading, or swap a phrase for a proximity query. Keep edits small and test again.

Best Ways To Find Medical Literature Review Keywords

Adapt Your Terms Across Databases

Each platform has its own quirks. Copy your PubMed blocks into a text editor, then create versions for the other databases you’ll search. Keep the concept buckets identical; only the field tags, thesaurus labels, and operators change.

Embase (Ovid or Elsevier): swap MeSH for Emtree (e.g., Hypertension/), use platform proximity like adj3 or NEAR/3, and keep truncation.
CINAHL: map to CINAHL Headings and use MH for subject terms.
PsycINFO: map to the APA Thesaurus; proximity is often NEAR/x.
Scopus/Web of Science: no thesaurus; rely on carefully built text-word blocks and proximity.

Document Your Strategy So Others Can Reproduce It

Keep a live record while you search. A simple table with the database name, platform, date, full search string, and result counts saves hours later and backs transparent reporting. Export your final strategies into an appendix or a shared repository.

Note which filters you used and why (study design, language, or date). Save a copy of the unfiltered strategy as well, since filters change over time and can hide eligible studies when applied too early.

Quality Checks And Quick Fixes

Before you lock the strategy, run through this short list:

Precision check: scan the first 50 results. If many are off-topic, tighten the problem bucket or add a method term like randomized controlled trial[pt].
Recall check: pick five must-include papers and confirm your query finds all of them. If any drop out, add the missing term or switch a Subheading.
Parentheses: verify every OR block is wrapped before you AND the buckets together.
Field tags: confirm each term has the right tag in each database.
Noise words: trim vague terms like therapy without a context when they flood the set.

If you’re still getting odd results, open “Search Details” in PubMed to see how the system interpreted your words. Autoparsing can split phrases or map a word to a different concept than you intend; a quick tag or quote usually fixes it.

Crosswalk Of Common Databases

Use this quick crosswalk to keep syntax tidy when you move a strategy from one place to another.

Database	Controlled Vocabulary	Notes
PubMed/MEDLINE	MeSH	`[mh]`, `[tiab]`, proximity with `[tiab:~N]`; wildcards and quotes skip mapping
Embase	Emtree	Proximity like `adj3` or `NEAR/3`; use Emtree explode settings
CINAHL	CINAHL Headings	`MH` for headings; proximity varies by platform
PsycINFO	APA Thesaurus	Text fields plus `DE`/`MH` tags; proximity often `NEAR/x`
Scopus / Web of Science	—	No thesaurus; rely on text fields, quotes, and proximity

Two Reusable String Patterns

Copy and tailor these starter blocks. Swap in your topics and class names.

Drug Therapy For A Condition

( "Condition"[mh] OR "Condition/therapy"[mh] OR (condition OR "condition phrase")[tiab] )
AND
( "Relevant Drug Class"[mh] OR (class name* OR generic* OR brand* OR pharmacologic group* )[tiab] )

Helpful Tools That Speed Up Keyword Work

MeSH on Demand: paste an abstract to get suggested MeSH terms you can inspect in the Browser.
Built-in PubMed aids: use the search builder for field-tagged blocks, then copy the full string to your notes.

Where To Double-Check Syntax And Headings

When in doubt, verify against official help pages. The PubMed Help guide lists field tags, wildcards, phrase rules, and proximity format. For headings and scope notes, open the MeSH Browser. For full search design and documentation standards, see the Cochrane Handbook.

Decide What To Include And Exclude Early

Keyword work isn’t only about finding words to add. It’s also about deciding which angles don’t belong. List a few clear exclusions next to each bucket before you start testing. For hypertension therapy, you might exclude pregnancy if that isn’t your population, or veterinary terms if your review applies to humans only.

Add a small NOT block only when truly needed, and keep it tight, e.g., NOT (animals[mh] NOT humans[mh]).
Prefer positive filters where you can. “Adults[MeSH]” avoids a long list of paediatric terms in a NOT block.
Keep exclusions temporary during early tests. Once recall looks solid, re-apply them and recheck the first page of results.

Handle New Terms And Preprints

Indexing lags behind publishing dates, and author wording shifts over time. To catch early records, include a text-word block that doesn’t depend on subject headings. Pair this with date limits only when your protocol calls for it, and prefer a wide window during early runs.

If your topic has new slang or brand names, add them to the text block. Keep a list of new words you notice while screening, then feed those back into the next test run. This loop keeps recall high without blowing up the set.

Extend Beyond Journals When Needed

Many reviews need more than database hits. Policies, theses, and trial registries can matter. After you finish the core databases, push the same keywords into a trials register and one or two grey sources that suit your topic. Keep the same buckets and reuse your text-word blocks where the site allows a detailed search form.

Save the links you try and the date you ran them. Even a line in your notes—site name, query, and record count—keeps your reporting tight.

Common Mistakes And Easy Repairs

Nesting errors: OR terms not grouped. Fix by wrapping every synonym set in parentheses before adding AND.
Overuse of broad words: words like therapy without context. Add a class name, mechanism, or MeSH/Subheading pair.
Ignoring plural forms: rely on truncation where safe, or list both forms if truncation would clip too much.
Stopping after one database: repeat the plan across sources so you don’t miss records that aren’t in MEDLINE.
Not checking “Search Details”: always inspect how the system mapped your words; adjust with tags or quotes.

Name, Version, And Store Your Strings

Give each strategy a short name and a version number, then save it in a shared folder along with the date, database, platform, and record counts. If you tweak only one bucket, bump the minor version (e.g., 1.2 → 1.3). If you change buckets or add a new database, bump the major version (e.g., 1.x → 2.0).

At write-up time, paste the full strings into an appendix. Readers can see exactly how you searched, and you can rerun the same strings when you update the review.

Build Your Own Reusable Template

A lightweight template speeds repeat work and keeps teams consistent. Here’s a simple layout you can copy into your notes app:

Review aim: ____________________________
Databases & platforms: _________________
Date ranges: ___________________________
Bucket A heading terms: ________________
Bucket A text words: ___________________
Bucket B heading terms: ________________
Bucket B text words: ___________________
Bucket C heading terms: ________________
Bucket C text words: ___________________
Filters (kept separate during testing): ____________
Exclusions (if any): ___________________
Top five sentinel papers to check: _____

Fill this once, then reuse it across projects. You’ll spend more time screening good records and less time wrestling with syntax.

Where To Double-Check Syntax And Headings

Bring It All Together

Great keywords come from a simple rhythm: define the concepts step by step, pair each with the right subject headings, build generous text-word nets around them, then test and trim. Save every version, label what changed, and show the full strings in your methods. With that habit, your screening pile shrinks, your write-up stays clear, and anyone can repeat your search later without guesswork.