The National Library of Medicine's new search system, PubMed, gives us a powerful new interface to Medline. With links to "related articles" and publishers' Web sites, PubMed offers search, refinement, display, and delivery features previously unavailable. Along with IGM, PubMed is now accessible without charge worldwide. Our users are hearing about it, and we will all need some experience with PubMed to respond to their questions. There is extensive online help and documentation-click on "overview" or "help." This new system does not look or feel like Elhill, Grateful Med, or commercial engines.
There are at least three ways to begin a search in PubMed: basic, advanced, and clinical queries. Your most obvious option is to use the search box on the PubMed homepage (basic search). You enter words or phrases, press the return key (or click on "search"), and PubMed displays search results in brief form. Clicking on an author's name takes you to a full record. You may use your browser's navigation capabilities to return to the brief display, print, or move elsewhere. Although this "basic" search seems intended for simple queries, the documentation explains how to use it with truncation, qualifiers, even complex Boolean searching.
Clicking on Advanced Search takes you to a slightly more complex input screen. Here you can more easily limit the term(s) you enter (e.g., MeSH, textword, author, UI). You also have a detailed "neighbor" capability (PubMed calls this "list terms mode"). If you change the mode default from "automatic" to "list terms," PubMed will perform the equivalent of an Elhill "neighbor" command. From the resulting display, you can select one or more terms to search. PubMed displays the number of citations retrieved in "current query." Click to "retrieve" (in brief form) all citations which match your search strategy. You can choose to limit by publication date and/or set the number of citations per page before you click "retrieve." You can also choose to narrow your search by using "add term to query." The other option on this page is to "modify current query" which allows you to use other Boolean operators, although using the current version of this form is somewhat confusing.
The third way to enter your search is to click on Clinical Queries. Building on work by Haynes and others at McMaster University, this approach to PubMed uses stored search strategies (filters) to locate studies which focus on therapy, diagnosis, etiology, or prognosis of clinical conditions. You will want to read the background information and review the fascinating "table for clinical queries using research methodology filters." This is a very interesting application of the work on evidence-based medicine and filtering.
Some of the most powerful features of PubMed appear after you get an inital retrieval, regardless of the input form you've used. PubMed has precomputed sets of related articles for each article in the database. The relationships are based on statistical properties of words used in the title, abstract, and MeSH headings. Your initial search strategy does not need to be overly detailed, since you will have the opportunity to select the "best" article(s) from your retrieval set and link to logical neighbors.
PubMed provides many other links. As its name implies, NLM is working with publishers in this endeavor. When possible, PubMedprovides links to the Web sites of publishers so that users can retrieve the full documents. Obviously, licensing agreements will be important, but we can expect the number of participating publishers will grow. Collaboration with publishers also benefits NLM, since formatted data supplied electronically by publishers will speed the input process.
PubMed is an outgrowth of the Entrez retrieval system developed by the National Center for Biotechnology Information (NCBI) at NLM. Originally the system was used for sequence and structural records in molecular biology. Searching for related DNA or protein sequences is fundamental to scientists in these fields. You can use PubMed to search GenBank, genome, or structural databases directly or you may link to sequence data from the bibliographic records retrieved in your PubMed search. records retreived in your PubMed search.