Skip to Main Content

Your browser is unsupported

We recommend using the latest version of IE11, Edge, Chrome, Firefox or Safari.

Internal Medicine Residents & Students: Class Notes - PubMed

Guide for teaching Internal Medicine Residents

MeSH - Using PubMed's Indexing to Search


     Me dical

     S ubject

     H eading


Mesh searching is Concept Searching. Concept searching helps us focus precisely on the most relevant articles and discard articles where a keyword might appear incidentally, but does not comprise the focus of the search.

MeSH is utilized to index or tag most articles in PubMed using the NLM classification system. When an article is published in a journal covered by PubMed, it is sent to the National Library of Medicine (NLM). At NLM, a subject matter expert or medical librarian analyses the article's contents and assigns subject headings or tags that reflect the article's key concepts. 

There is some delay in indexing, so if you want the latest articles, you will need to use keywords too. Very recent articles without MeSH indexing will not be retrieved by a MeSH search alone. Be aware that a small percentage of articles are never indexed with MeSH and must always be retrieved through keyword searching alone.

Many people don't realize how important MeSH is because many keyword searches automatically map to both  MeSH and Keyword searching. MeSH is especially important when a keyword has many meanings - example: Nursing.  There are over 80 concepts with the word Nursing included in them in the MeSH thesaurus. If you are interested in breastfeeding infants, unless you select the nursing concept related to Breastfeeding, you will retrieve many irrelevant articles.

MeSH searching allows users to further limit retrieval by applying Major Concepts and subheadings limits. These limiters should be used carefully since indexing is not always precise. For example, an article that might be quite interesting to you, may be indexed with a MeSH term at the non-major level, and you will not retrieve it if you specify the major-only level of MeSH. Generally, the basic MeSH level is a good way to search unless you have overwhelming results or are very clear that the major-only MeSH search is optimal.  


MeSH searching is also useful in cases where a MeSH term maps to numerous subordinate concepts that can expand a search - example: Contraceptive Agents recalls most of the contraceptive drugs articles in the PubMed database because it has dozens of subordinate headings. Reproductive Control Agents is an even more comprehensive term that recalls the articles associated with contraceptive agents and many others. Another example: Health Planning has an extensive list of subordinate terms that can amplify a search. If you do keyword only searching, you will not benefit from the subordinate concepts listed beneath a MeSH concept.

On the other hand. MeSH does not contain the term, "service-learning", which will retrieve over 500 citations when used as a keyword.


An example of a Diagnosis search in MeSH:

P: in an otherwise healthy seven-year old boy with a sore throat

I: how does the clinical exam

C: compare to throat culture

O: in diagnosing streptococcal infection?


Here is the search built in the MeSH Builder by searching for 1) "sore throat" (select the Microbiology subheading to retrieve "throat cultures" since this is not a MeSH term), 2) "clinical examination" (the MeSH term is Physical Examination), 3) and "Streptococcal Infections" (this is a MeSH term).  

Search -  ("Pharyngitis/microbiology"[Mesh] AND "Physical Examination"[Mesh]) AND "Streptococcal Infections"[Mesh])


Clinical Queries and Optimizing Searches With Filters

Clinical Queries - CQ is a valuable PubMed search interface that brings specialized filters to aid you in getting the highest levels of evidence for your patient oriented questions. If you are in a hurry to answer a clinical question, start with Clinical Queries in the first place.

From your PubMed search results page, select the Advanced link just beneath the search box. Look for the number of your search and select Clinical Queries from the More Resources pulldown at the top of the Advanced page page. When you are in Clinical Queries, you can just enter the search number (e.g., #56) and press enter. Alternatively, you can enter a search string. Once the search results are back, select the type of question and whether you want a broad or narrow filter. Note the link to the Haynes search filters at the bottom of the first column results.You can also search Clinical Queries by entering the search terms directly in the CQ search box. Select the link to the search results at the bottom of the first column to have your search registered in the MYNCBI home recent searches table.


Optimizing Your Search

1. A strong and streamlined rapid search -

a) Search by both MeSH and Keywords - but use the minimun number of concepts needed to identify your intended literature. If you do not retrieve many results, consider looking for additional keywords and MeSH terms (see the Pearling entry below). Search on MeSH terms only (e.g., tag reads [Mesh]) - in other words, use the subheadings or MeSH Major focus selections judiciously.

b) Consider using a PIC format - sometimes it is better to drop the O (Outcome) component of the PICO question if you want to be open to a wider variety of outcomes.

c) Use the Clinical Queries filters, limit by dates, limit to Core Journals...add filters to reduce the research results. Clinical Queries utilizes special filters developed by Dr. Haynes that allow you to maximize specificity or sensitivity by question type. Remember, unless a filter is critical, specifying it may reduce results in unanticipated ways - e.g., if the indexer neglected to add "Human" as a MeSH term, then a MeSH only search may not retrieve a citaton that is about humans, but not appropriately tagged!

2. Pearling or adding to your search terminology - You can add to your search term list by looking at the indexing for any articles that capture the content you are focused on. You can also look at the articles cited by or citing your prize article. This puts you in touch with the researchers who may be on the same investigative thread.

3. Systematic Reviews.  Note, systematic reviews often focus on clinical questions and are useful for immediate medical decision making. However, they are comprehensive reviews of the literature that must be scrutinized for timeliness. For example, a systematic review may have been published in 2013, but the underlying searches comprising the review may have concluded in early 2012 or before. Check the most current literature to be sure a significant study was not published after the search window of a systematic review closed.

Types of Questions - PICO, Background v Foreground

Background VS Foreground Questions -

Background questions - concerned with general information about a condition or disease - causes, symptoms, progression, accepted treatments, outcomes, prevention...

Foreground questions - patient-centered and specific to a patient and their contexts - comorbidities, demographics, life-cycle stages, life-style, environment, genetic background, epigenetic factors, psychological, sociological, behavioral, economic, anthropological, health delivery context....

The following 10-minute video provides an excellent introduction to the P.I.C.O. question and its place in the framework of evidence-based medicine:



The P.I.C.O question is a structured way of asking a clinical question about a specific patient. Link to UIC PICO tutorial page:


Example of a P.I.C.O. question formulation and search -

Dr. Jones is concerned about potential problems (e.g., methotrexate-related neutropenia / MTX-related neutropenia) related to prescribing methotrexate medication for Mr. Smith's rheumatoid arthritis. She has been noticing some mild dementia and wonders if noncompliance might present a problem....

Are patients with rheumatoid arthritis who are taking methotrexate and have dementia at a greater risk for developing methotrexate-related neutropenia than patients without dementia?

P: in patients with rheumatoid arthritis who are taking methotrexate

I: have dementia (risk exposure)

C: without dementia (controls without risk exposure)

O: have a greater risk of developing methotrexate-related neutropenia

Keyword Search -  (methotrexate related neutropenia) dementia (rheumatoid arthritis)

PMID:  23743624 "Do impaired memory, cognitive dysfunction and distress play a role in methotrexate-related neutropenia in rheumatoid arthritis patients? A comparative study"

** Note, this keyword search was translated in the background by PubMed to include a keyword search and a MeSH search! See the Search Details box on the right sidebar of the search page.  PubMed's translator used a MeSH term wherever one existed and when one did not exist, it appropriately grouped the keywords for multi-word concepts. (Sometimes it has trouble with this grouping, but generally it is brilliant)!