Special Feature
Searching Medline with PubMed
By Leah Anderson, MLS
Editor’s Note: The growth of the Internet has greatly expanded the reach of medical researchers. This article provides some basic insights into Medline and PubMed, http://www.ncbi.nlm.nih.gov/PubMed. A future article will provide searching tips for use in PubMed.
A mandate from congress in 1997 allowed free access to Medline for anyone with an Internet connection. The recipient of the mandate was the National Library of Medicine (NLM), the producer of the Medline database and also a government agency within the National Institutes of Health. NLM fulfills this mandate by providing access to Medline with the PubMed and Internet Grateful Med (IGM) search interfaces. Both are developed and maintained by NLM, and both are accessible via the World Wide Web. This article provides an overview of PubMed and the features that make it unique relative to other interfaces available for searching Medline. Some of these differences are subtle while others are significant, but all have an effect on how well you retrieve the information you are seeking.
Searching Medline is not always free. While NLM produces the Medline database and provides free access to it, it also licenses the database to commercial companies, which, in turn, create other search interface programs for Medline. Much of the time, these other programs come at a cost. Some familiar programs are Ovid, SilverPlatter, Knowledge Finder, and Melvyl. Most medical schools purchase a license to provide access to Medline over their networks using these or other applications. Only NLM, as a government agency, was given the congressional mandate to make access to Medline free. Other web-based medical sites may offer "free" Medline simply by linking to PubMed.
The Database Difference
PubMed is a database separate from the traditional Medline database. It was created by combining the Medline database with a bit more. Thus, everything found in Medline will be found in PubMed but the converse is not true.
Deciding which references will appear in Medline is a tightly controlled process. Currently, articles from around 4000 journals are indexed in Medline. However, not every journal has every article included in Medline. Some journals are selectively indexed, meaning that if they contain an article that is not related to biomedicine, it is not included. For most selectively indexed journals, PubMed does not make this distinction. For example, the journal Science is selectively indexed in Medline. However, if you search for a geological article from a recent issue of Science in PubMed, you will find it. If you search for that article with other Medline interfaces (such as Ovid), you will not find it.
Many publishers have made agreements with NLM to provide references and abstracts to their journals electronically to NLM as soon as an issue is published. This basic information is transferred directly into the PubMed database without any of the usual indexing processes. This greatly increases the currency of PubMed over the traditional Medline database searched with programs such as Ovid or Knowledge Finder, which often have a lag time of 2-6 months for references to appear from recent issues.
Creating PubMed Links
In return for providing electronic references, publishers allow NLM to create PubMed links directly to their web site and to the full text of the article, if available. However, one has access to the full text of an article only if 1) the publisher is providing access to everyone for free such as through a free trial period; or 2) an individual has paid the publisher for full text access or access comes with membership; or 3) the library with which an individual is affiliated has paid for full text access to the journal.
One of the hallmarks of the Medline database is the application of controlled medical subject headings (called MeSH) to each record. MeSH provides a standardized medical vocabulary that can be instrumental in searching the database well. NLM has created a controlled vocabulary that also acts like a thesaurus; it is constructed to provide relationships among terms. These features can be used to search more effectively.
The Value of MeSH
As an example of the value of MeSH, one MeSH heading will be designated as the official subject heading for a disease that is known by several different names. All the names will be mapped to that one official heading. Thus, putting in one name will pull up the articles that discuss the same condition but call it something else. Another example is variant spellings. Medline’s subject headings are based on American spellings. Articles discussing "haematology" will be assigned "hematology" as a subject heading. Thus, anyone searching hematology will pull up articles with the British spelling as well.
Because PubMed includes the Medline database, it includes MeSH and will continue to do so. However, there is a lag period before subject headings are added to new records. It may take up to six months for subject headings to appear for a citation. Furthermore, some records in PubMed will never get subject headings. Nonbiomedical articles from selectively indexed journals will never have subject headings added though the basic record will always remain in the PubMed database.
Boolean Logic
The search engines behind many software interfaces designed to search Medline are based on Boolean logic. In such a system, one searches for term A and term B and term C. One can also exclude terms. The addition of the controlled subject heading vocabulary aids searching tremendously by helping out with variant spellings, multiple names, and in many other ways. In the end, Boolean searching is very mechanical; the terms used must be present anywhere in the record. Context and meaning are often lost in a Boolean system. The application of subject headings often cannot always address these problems. Thus, the number of irrelevant articles retrieved will increase. For example, "cultures" refers to groups of people as well as laboratory tests but all records with this term, regardless of the meaning intended, will become a part of the search. Causal relationships among terms can also be lost. An article discussing how caffeine causes miscarriage will have the same terms and subject headings as an article saying caffeine does not cause miscarriage.
The PubMed search engine tries to overcome many of the limitations of strict Boolean searching. When searching PubMed, you will find that each citation has a link called "See Related Articles." The group of citations pulled up with this link has been predetermined to be similar to the original citation through the calculations of a complex set of algorithms. The algorithms compare the citation to all others in the database, searching for similarities among them in the text of the title, abstract, and MeSH. The algorithms include the proximity of words to each other in the title, abstract, and subject headings, and how often a word appears in a record. Weights are scored using this method, and the citations with the most weight are considered as similar. When you find an article that is exactly what you are seeking, click the "See Related Articles" link, and similar articles will appear. (Ms. Anderson is Medical Librarian for the Health Sciences Library, Sequoia Health Services, Redwood City, CA.)
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content