Primary Literature - PubMed FAQs

What are MeSH terms?

Pubmed searches its database in several different ways. A basic understanding of this is important for formulating successful searches. The main language used by Pubmed is the MeSH (Medical Subject Heading) term. This is a hierarchical tree of terms that classifies everything that could possible be of interest to us into a series of headings and subheadings. Each time a paper is posted on Pubmed an employee of the National Library of Medicine classifies the paper using all of the MeSH terms that he or she thinks are appropriate to it. We are therefore somewhat at the mercy of the people performing this classification, although in general they do an excellent job and the results are reliable.

Ideally we should use MeSH terms when we perform a search so that we are speaking the same language as Pubmed. As with radiology requisitions, if you know precisely what question is being asked then you have a better chance of getting the right answer.

When you formulate a search on Pubmed without using MeSH terms, the computer at the National Library of Medicine tries to match the words you have chosen with what it takes to be the best available match for this term from the MeSH tree. This introduces another level of uncertainty and probably also irrelevance into your search.

How do I find the MeSH terms for the subject I am interested in?

If you want to search a topic using MeSH terms there are several ways of finding out the relevant MeSH terms.

1. We have found that one of the easiest ways to find MeSH terms is to see what MeSH terms were used when the Pubmed staff indexed other papers on the topic of interest. Use the Clinical Queries page to retrieve a small number of (hopefully) relevant papers. Click on one or two of these and choose the "Citation" view rather than the default "Abstract" view. This allows you to see the MeSH terms that have been used to index that paper and will often suggest MeSH terms that you might not otherwise have thought of. The problem with this approach is that you then have to type the MeSH term precisely into the query box, with [MeSH] after the term. It is easy to make a mistake doing this, particularly with more complex terms.

2. Use the MeSH Browser. This feature on the sidebar of the main Pubmed page allows you to enter a term and Pubmed will then tell you if it is a MeSH term or will alternatively suggest a list of possibilities for you to choose from. Once you have done this, Pubmed gives you the option of adding the term to your search strategy using a variety of Boolean operators. This avoids the need for you to type the term accurately into the query box and also avoids any worries about syntax.

In practice a combination of these approaches often works best, starting with the first method and using the second to construct your query.

How do I input these terms into my search?

Once you have figured out the MeSH terms that you want to search, you then have to get them into the query box in a form that Pubmed can understand. One approach is to type each term in yourself, taking care to get the spelling and syntax correct. This can be difficult, particularly for a complex search. If there are lots of "AND" and "OR" terms in your search, then you have to group terms using brackets, which can also become confusing.

There are a couple of ways we can make this easier.

1. MeSH Browser: We can use the MeSH Browser to insert terms into the search as described above. This avoids any problems with spelling or syntax.

2. Stepwise approach: We can break our search into different parts and combine them at the end. In a classical EBM search we will have several alternative terms in each column of the PICO question that need to be linked using the "OR" operator. We also have 3 or 4 columns, across which we need to use "AND" as a linking operator (see section on asking an answerable question). It is easy to see the potential for an unmanageable number of brackets if we try to phrase this in one big question.

We can start by searching the terms in one column of the PICO question, linked by "OR". Using the "Preview" button we can see how many articles this would retrieve. We then clear the search box and search the terms in the second column, again linked by "OR" and using the "Preview" button. We do the same for the third and fourth column. By clicking on "History" we can now see a list of each of the 4 previous searches, numbered #1 to #4. The next step is to enter the following string into the search box: #1 AND #2 AND #3 AND #4. Pubmed then combines the results of each search and retrieves only articles that contain MeSH terms from all 4 columns.

A word of caution is necessary here. The indexing of articles is not always perfect. Therefore it is often useful to browse through the results of different combinations (eg #1 AND #2) so as not to exclude important articles that were not comprehensively indexed. If the article has not been indexed with MeSH terms from all 4 columns of your search, then you will not retrieve that article.

How do I avoid getting lots of useless stuff from my search (applying limits)?

There are some easy tricks to reduce the amount of useless data in the results of your search. Under the query box in Pubmed there is a button called "Limits". When you click on this you get several options for limiting your search. We routinely check the following limits:

English (with apologies to those of you whose first language is not English!) - even if the abstract is in English it is unlikely that you will be able to appraise the paper if you cannot read the whole article.

Abstracts only - you need some information to decide whether it is likely to be worthwhile trying to get the actual paper.

Humans - applies to most of the literature that we have been interested in so far

Publication Date - this is optional. For example, you may only be interested in results with newer MR sequences and therefore can exclude old data by selecting the dates you want searched.

Type of study – For therapeutic (interventional) questions, you can first restrict your search to the highest level of evidence only (meta-analysis of randomised controlled trials). If this is negative, you can restrict your search to the next level - ‘randomised controlled trials.’ Failing this, search for ‘Controlled trials’ and so on. Thus you avoid retrieving studies with weaker designs. The hierarchy of trial design is described in more detail in the appraise section of this site. The table of ‘Levels of Evidence’ from the Oxford University / NHS website, to which a link is provided, illustrates the hierarchy nicely.

How do I save my search strategy (so that I can prove to someone else that my search was comprehensive or so that I can run exactly the same search at a later date to see if there is new evidence)?

There are some good reasons why you might want a precise record of your search strategy (ie what you entered into the query box). When you come up with a ground breaking evidence based paper for publication or policy for your department, one of the first rules of EBM is that you must be able to prove to others that your search for the best evidence was appropriate and comprehensive. If you subsequently decide to return to the same topic at a later date to see if there is any new evidence then you need to be able to recall the strategy. There are different ways to do this.

The easiest way is to print out the "History" page when you are finished your search. This will give you a listing of what you have done and is particularly useful if you have followed the stepwise approach to building a search described above. One problem with this is that to repeat the same strategy you have to manually type in all the same queries again, but this is not usually a problem once you have a good record of the search.

An electronic method of saving your strategy from the "History" page is to use the "Save as" command on your browser and just make sure to type ".html" after the filename (for example, "search1.html"). You can open this saved file at any time using your browser without having to be online.

There is a more elegant way to save your search strategy. The "Cubby" is a feature on the sidebar of the Pubmed site that is specifically designed to allow you to save search strategies and to use them at a later date to see of any new evidence has been published. You have to register to use this feature (registration is free). The problem with the Cubby is that it will not remember any search that is built in a stepwise way using the terms "#1" or "#2" etc. It can only be used for searches built with the entire search strategy in the query box at one time. This can be daunting for many 4-part EBM questions.