How to Build a PubMed Search Strategy Using MeSH Terms: A Step-by-Step Guide 

Send two researchers the same clinical question to PubMed. There’s a good chance they’ll come back with different studies. Neither of them was intentionally “bad”. One typed in some words and pressed enter. The other one sat down and decided on the search: the subject headings, free-text terms, the string looked for, the date it was run.

That’s a serious issue if you are writing a systematic review or a scoping review. The wrong conclusions can be reached if a few important trials are missed. And if no one can run your search again, reviewers have no reason to believe it. So let’s make a nice one. At the end, you will be able to use MeSH terms with keywords, use Boolean operators and field tags to combine terms, and, with the help of a record, follow along line by line.

What that decision-making looks like inside a systematic review, from the PICO question through to the final included studies, is covered in how to write a systematic review and meta-analysis.

What is a PubMed Search Strategy

A PubMed search strategy is a precise way of formulating a search query to find the most relevant articles.

Consider a search strategy as the recipe for the results you are getting. It explains the subject headings you employed and the free text terms, how you connected them, which field tags you used, and any restrictions (and why you needed them).

Users often mix up PubMed, MEDLINE and MeSH. They aren’t:

• The National Library of Medicine (NLM) hosts the free search platform PubMed via NCBI.
• MEDLINE is NLM’s primary bibliographic database. The vast majority of PubMed is MEDLINE, but there are other records in PubMed.
• NLM uses a controlled vocabulary called Medical Subject Headings (MeSH) to index articles in MEDLINE.

Two words will also occur when someone makes a search evaluation. Sensitivity will refer to how many of the relevant papers you find are relevant to your search. Precision is the percentage of the information you retrieved that is useful. In systematic reviews, it’s typically SENSITIVITY that prevails. You would prefer to look at a dozen or more abstracts than you would miss a vital trial.

MeSH terms vs. keywords in PubMed

NLM tags an article with MeSH headings after it is added to MEDLINE, indicating the contents of the article. For this reason, a paper reporting a “heart attack” is still found under “Myocardial Infarction”[Mesh]. The keywords are used in more literal ways. They match with words in the title, abstract or author keywords.

Feature	MeSH terms	Free-text keywords
What they are	Standardized subject headings	Words and phrases used in records
Main advantage	Group related concepts under one preferred heading	Capture wording variations and newer terminology
Limitation	Missing from records that aren’t indexed	May retrieve irrelevant or differently used terms
Example	“Myocardial Infarction”[Mesh]	heart attack[tiab]

They are both necessary but not sufficient. MeSH rescues articles written with odd or old-fashioned wording. What keywords do that which MeSH can’t see: records awaiting indexing, records not in MEDLINE, and newer terms that do not have a heading. Ideally, a systematic review search will make use of both.

How to build a PubMed search strategy using MeSH terms 

To create a PubMed search strategy of MeSH terms, follow this procedure:

Step 1: Start by stating your research question in the PICO format

PICO is a simple method to dissect a clinical question:

• The range of people you study (population size).
• Treatment or exposure to something that causes harm.
• Condition (Comparison): what you are comparing it to
• Process: how you are measuring the outcome

We will continue in this guide with one question: In adults, is exercise effective to improve blood pressure? 

Adults = population; 

no exercise or usual care = comparison; 

exercise = intervention; 

blood pressure = outcome.

It is not necessary to use a search block for each letter of PICO, however. Terms such as “usual care” are used in a variety of ways, and searching for them tends to miss good studies. Two or three fundamental ideas are a good starting point.

How to turn a clinical problem into a properly structured PICO question, and which variants to use for observational or qualitative studies, is covered in the PICO framework for research questions.

Step 2: The main concepts and synonyms need to be identified

Now write all the possible descriptions that an author could use for each concept. Generally, it can be done with a simple table:

Concept	Synonyms and variations
Exercise	physical activity, aerobic training, resistance training, walking
Blood pressure	hypertension, high blood pressure, systolic pressure, diastolic pressure

Be sure to not forget abbreviations, plurals, US and UK spellings. A helpful hint: raise two or three papers you already know you need to review. Look at their titles, abstracts and author keywords, as well as their MeSH terms on their PubMed pages. There’s almost always at least one term or two that you haven’t thought of.

Step 3: Search for the right MeSH headings

The third step is to seek the appropriate MeSH headings.

Go to the MeSH Database (found on the PubMed home page). Enter your term and then click the record that corresponds to it. There are a couple of points to consider:

• The term you are looking for, such as the preferred heading.
• Entry Terms: synonyms that lead to that heading and are good additions to your list of keywords
• The tree structure with wider and narrower headings encircling it.
•  The subheadings (e.g. therapy or diagnosis)

As an example, after typing in “heart attack,” the database redirects you to Myocardial Infarction.

Step 4: Check whether the heading fits your question 

To make sure that your heading matches your question.

Take precautions when using similar headings. There are two different MeSH headings: “Exercise” and “Exercise Therapy”. As is the medical term “Blood Pressure” itself, and “Hypertension” the disease. Each record is preceded by a scope note (short definition). 

Read it first before you commit.

How to combine MeSH terms and keywords in PubMed 

One more thing. MeSH is updated annually, so refer to the current vocabulary rather than using headings that were used in a search five years ago.

To use MeSH terms and keywords in PubMed, you must add an ‘AND’ between them. Using MeSH terms and keywords in PubMed requires an ‘AND’ between the two.

Once you have your headings and synonyms, it’s time to organize them into concept blocks, one concept block per concept. Boolean operators perform two functions here. They connect the words within the blocks, and they connect the blocks to the other blocks.

To combine synonyms with OR, you can use the OR operator.

Enter any of your terms in PubMed, and OR will return records. Is a part of a block:

heart attack OR myocardial infarction OR coronary thrombosis

Each synonym you enter using the OR operator expands the scope of the results.

Use AND to bring together concepts.

AND is the opposite of that. It only returns papers that match both It and exercise AND blood pressure gives you papers that match both. The new AND block reduces your results.

Use NOT cautiously

No record that mentions the excluded term is thrown out, and good studies can be lost this way easily. If you’re going to eliminate pediatrics from your job, try searching for hypertension NOT children. Any study that included both adults and children is lost if it doesn’t answer your question, too. The default setting for most systematic reviews is NOT out; in cases of screening, this is preferable since many studies will not be relevant.

 Use PubMed field tags

Field tags are terms that direct PubMed to a place to search for a term:

•  [Mesh] searches MeSH headings (and, if you like, the narrower headings that are found below them).
• [tiab] will search the titles of published articles, abstracts and author keywords.

If you’re searching for phrases, enclose them in quotation marks, such as “physical activity”[tiab]. Next, add parentheses around each concept block, which will help PubMed to combine the elements in the order you intended.

Beware of Automatic Term Mapping (ATM) also. PubMed implicitly reinterprets terms without tags. It may include MeSH headings or split up a phrase into separate words. If you don’t like the results that PubMed returns, look at the Search Details section on the Advanced Search page; it tells you exactly what you put and what they got back.

A PubMed search strategy Example for a systematic review 

Let’s take our exercise and blood pressure question and put it into action, and make each block a concept:

“`

(

  “Exercise”[Mesh]

  OR exercis*[tiab]

  OR “physical activity”[tiab]

)

AND

(

  “Hypertension”[Mesh]

  OR hypertens*[tiab]

  OR “blood pressure”[tiab]

)

“`

Line by line:

• “Exercise” [Mesh] fetches articles that have been indexed under the narrower terms under Exercise.
• Exercise*[tiab] looks for exercise, exercises, exercising, and exercised in titles and abstracts.
• “Physical activity”[tiab] is a common synonym added as an exact phrase.
• Hypertension[Mesh] will retrieve articles that are indexed under the narrower terms of Hypertension and its subheadings.
• Hypertens*[tiab] captures the word hypertension, hypertensive and related terms.
• The term ‘blood pressure'[tiab] will include studies of blood pressure in individuals without a diagnosis of hypertension.

Why is there an OR in the blocks? Any one term suffices to demonstrate that an article is about that concept. Why is there an AND between them? A relevant paper must have to address both exercise and blood pressure.

Be aware: This is a starting point. Not validated and will not catch all. Before employing in a real review, verify that the current MeSH headings and entry terms are accurate, test it on papers you know should show up, and review Search Details. You may also include “Blood Pressure”[Mesh] or additional types of physical activity such as “resistance training”[tiab] or a population block for adults or a study design filter depending on your protocol.

The Advanced PubMed Search Techniques

Once the basics are comfortable, there are a few tools that are worth learning:

• MeSH explosion. Narrower headings for each MeSH term are automatically retrieved by PubMed. Search “Hypertension”[Mesh], and you’ll similarly see headings such as Hypertension, Pregnancy-Induced. Use [mh:noexp] to get only the heading, e.g., “Hypertension”[mh:noexp].
• MeSH subheadings. These focus a heading down to one approach to a given topic. For instance, if you search for “Hypertension/therapy”, you’ll see articles focusing on therapy for hypertension. Be gentle with them when performing systematic reviews, because the results are dependent on how each indexer marked each article.
• Truncation. Word endings are located by the asterisk, as in exercis*, which includes exercise, exercises, and exercising. PubMed requires a minimum of four characters before the asterisk. When a term is truncated, it also turns off Automatic Term Mapping for that term.
•  Advanced Search Builder. Under the search bar, choose “Advanced. You can then create a search field by field, view the history, merge previous searches (such as #1 AND #2), and open Search Details for any of the searches.

How to improve and validate your search strategy 

To enhance and test your search strategy, use the tools and techniques outlined below:

Wait to run your final search until you’ve tested it. Here are the steps you’ll want to take in a checklist:

• Collect some articles that you are aware of and ensure that all articles show up when you search. If one is missing, read the title, abstract and MeSH terms to determine what you missed.
• Check the details of the search and ensure that each term was interpreted as you intended by PubMed.
• Review spelling, parentheses, Boolean operators and field tags.
• Skim the first couple of pages of results. Lots of garbage, or obvious papers missing? Find the term responsible.
•  Include synonyms that you haven’t thought of, but if you have a reason for using them.
• Request a medical librarian or information specialist to review it if it is important, like in a systematic review.

For this review, there is a standard tool. The PRESS 2015 guideline (Peer Review of Electronic Search Strategies) provides a checklist to help reviewers identify conceptual or technical problems that might lead to missing or incorrect information, such as missing synonyms, using the wrong operator, or misusing a subject heading.

Also, avoid narrow searches, such as by publication year, language, or article type. The limits can exclude studies, and they can introduce bias into the review if they have not included a methodological criterion.

How to write and cite a reproducible PubMed search

Take notes during search. Recalling what to do afterwards is ineffective, if not impossible. The simplest search log is:

Field	What to record
Database/platform	PubMed
Search date	Exact date the search was run
Search string	Complete query as executed
Filters	All limits applied
Results	Number of records retrieved
Updates	Dates and changes for later searches

If it is time to publish, include the full search strategy used for each database, in the exact form used. You need to provide a list of keywords. There is an extension to the PRISMA statement called PRISMA-S which establishes what should be included in that report. Remember, PRISMA-S is a reporting project. Even if the search is well documented, it can fail to identify studies if the search itself is inadequate.

The PRISMA-S checklist and explanation paper are available from the EQUATOR Network PRISMA-S page.

Common Mistakes That Can Be Prevented:

• Using only a few keywords from MeSH terms
• When using the word “or,” use synonyms with AND.
• Do not use parentheses around concept blocks
•  Adding restrictive filters too early
•  Assuming every PubMed record has been indexed with MeSH
• Not saving the final search string and date you performed it

 Frequently asked questions

Q1. How to identify the appropriate search terms in the MeSH database?

Find your key concepts in the MeSH Database and read the scope note, entry terms and tree for each concept. Then go to PubMed and look at some articles in the area, and see what MeSH terms they were assigned. That is where you might find headings that you otherwise wouldn’t.

 Q2. Is it possible to search the database PubMed without using the terms MeSH?

You can. The [tiab] free-text search is fine alone and will be required for new records and any records that have not previously been indexed. If you are conducting a systematic review, however, you will achieve greater coverage with the combination of free text and MeSH.

 Q3. How to perform a PubMed search using PICO?

Use PICO to select your key ideas, and then find the blocks of MeSH terms and key words for each. If the studies of one element (such as comparison) were excluded, then leave it out.

Q4. Is there a need to apply filters in a systematic review?

Only if such are warranted by your review question and protocol. Regardless of what filters you employ, please report them.

 Q5. Can PubMed be used as a sole source of information for a systematic review?

Rarely. The type of databases to search will depend on the question, but the majority of systematic reviews also search databases such as Embase and the Cochrane Library as well as trial registers when applicable.

Next steps

After a few times, it becomes an everyday occurrence. Define the question. Identify appropriate MeSH headings. Add free-text synonyms. Make the blocks, try out the search and jot it down. Now repeat the Exercise and Blood Pressure example above and try to change the exercise to something that fits your topic, and then go through the above checklist. A next step is to take advantage of NLM’s free PubMed and MeSH tutorials.

Disclaimer:

Articles published by American Academy of Research & Academics are prepared by our team using information from direct experience, publicly available resources, and educational references. AI tools may be used to assist with drafting, proofreading, and formatting; however, all content undergoes review and approval before publication.
The information provided is intended for educational purposes only. Requirements, policies, and processes may change over time. Readers should consult official sources for the most current information.