skip navigation
student.bmj.com

PubMed for beginners

Using PubMed to search for papers that you need can initially be a daunting task.Mohammad Al-Ubaydli takes you through the basics of using this valuable internet resource

Prof was angry today because none of us knew what percentage of stroke patients get epilepsy. He wants us to look up the answer by the next supervision, but I cannot find it in any of the textbooks. What should I do?

One option is to visit PubMed. This is a website. It has abstracts of much of the modern biomedical literature. If someone has done research that gives the answer to your question, you will probably find the abstract of the paper on PubMed.

Rings a bell now

PubMed contains a huge amount of information, so you need to learn a few things to find your way around. But after that, the site is simple to use.

Open your computer's web browser, and type in "www.pubmed.gov" in the address bar at the top (fig 1). In the top right corner is the logo of the National Library of Medicine in the United States which runs the website. The top left shows the logo of the National Center for Biotechnology Information. The centre has databases on gene sequences and proteins, entire textbooks, and of course PubMed.




The website helpfully defaults to "Search PubMed for" something, a little below the logos. That something is whatever you type in. For example if you type in "stroke patients with epilepsy" and click the Go button, the website searches the database for that phrase.

I found lots of papers, but I don't have time to sift through them all

The phrase "stroke patients with epilepsy" is a bad one for PubMed. As you can read in the second grey rectangle in fig 1, PubMed finds 645 papers, most of which are not quite what you had in mind. It is a little like filling out a blood form at your hospital by asking for "blood tests." The haematology department will probably refuse the request because it is not specific enough, and the more diligent haematologists will phone you to tell you off.

PubMed is forgiving: it searches for whatever you ask it to. The answer you are looking for will probably be somewhere in those 645 papers. But being specific gets you the right answer quickly.

But my question was perfectly clear

Not quite. Firstly, you asked about "stroke." Did you mean "cerebrovascular accident," or did you mean the verb "to stroke"? Secondly, did you want papers that were about strokes, or papers that mentioned the word stroke somewhere in the paper? Thirdly, you mentioned "patients with epilepsy." Did you want data from trials counting the percentage of patients who went on to develop epilepsy, or did you want anecdotal case reports in which a patient with stroke developed epilepsy? And so on.

So how can I ask more specific questions?

You should use cerebrovascular accident instead of stroke. And because epilepsy is unambiguous, you can enter "CVA epilepsy." After you press Go, you see that you are down to 443 papers (fig 2). It is a little easier to go through those, but there is one more trick. Click on 'Limit' just under the text of your question. It takes you to a page that allows you to narrow down your question in all sorts of ways (fig 3). In this case, you want to make sure that your answer comes from a trial. Click on Publication Types and choose Randomized Controlled Trial. Then click Go







Great, only one paper

And it looks rather promising as the title is "Incidence and predictors for post-stroke epilepsy."1 Click on the title to read the abstract.

"Twelve patients (2.5%) developed PSE [post-stroke epilepsy] during 12 months," is a good start. But of course you should read the full paper to understand all the caveats. For example, the study involves a small number of patients, and you would need to know the error margins for the study. The BMJ has an excellent guide on this.2

To read the full paper, click the Full Text button. PubMed provides these buttons to take you to the paper on the publisher's web site. Every publisher has a different policy about access to its journals, but if the journal is well established, and you are using a computer in your medical school's library, the chances are that you will be able to read the full paper. That is because your medical school will have already paid the journal's publisher.

Is that all PubMed does?

I hope you think that PubMed is useful. The site has many more features that will help you to answer questions throughout your medical career. On the left hand side of the screen you will see Help and Tutorial. Help gives you a clear explanation of every feature, and Tutorial is a guided tour. You should definitely go through these--after your supervision with Prof.



Mohammad Al-Ubaydli visiting research fellow, National Center for Biotechnology Information, Bethesda, MD, USA
Email: mo@mo.md


studentBMJ 2004;12:45-88 February ISSN 0966-6494

  1. Lossius MI, Ronning OM, Mowinckel P, Gjerstad L. Incidence and predictors for post-stroke epilepsy. A prospective controlled trial. The Akershus stroke study. Eur J Neurol 2002;9:365-8.
  2. Greenhalgh T, Taylor R. How to read a paper. London: BMJ Publishing Group. http://bmj.bmjjournals.com/collections/read.shtml


Previous article    Return to top   
Printer friendly page    Download article PDF    Email this article to a friend