Wednesday, April 08, 2009

PubMed searching

For point of care searching PubMed has been supplanted for many users, in the interest of time, by UptoDate and other filtered resources. Nevertheless, for certain searching tasks, I continue to find PubMed’s unique features useful. Occasionally I find useful tips to share with readers.

A question that often arises is “How useful is a free text search in PubMed?” Free text searching is fine for some searching purposes but does not take full advantage of PubMed’s features. I experimented by comparing a free text search with a search using Boolean operators for the topic “heparin induced thrombocytopenia”. The results?


Free text:





Boolean:




Identical.

You can find out exactly how the search was done by clicking the details tab which, in the case of our searches, reveals the exact same strategy for both:





(No other popular medical search engine that I know of allows you to do this, by the way).

These strategies, however, yield 2908 hits. This illustrates the disadvantage of free text searching which is that it usually produces an unmanageable number of citations. A more restrictive search strategy is often needed. One method is to restrict search terms to the title field. Without the title restriction PubMed searches throughout the entire citation including the abstract. The results will include many articles in which heparin induced thrombocytopenia is incidental to the main topic---articles you probably don’t want. By applying the title restriction you confine your results to articles that have all three words in the title. Type [ti] after each search term:




1363 is still too many citations. Additional strategies will be explored in future posts.

1 comment:

A Medical Librarian from Toronto said...

For this search, you might consider using the MeSH database to ensure a more accurate search (though free text will sometimes get it right). Try, for example: "Heparin"[Mesh] AND "Thrombocytopenia/chemically induced"[Mesh] This will yield 2140 citations.

At the point of care, this is a bit much. You might consider focusing your search topic. Consider using the "Clinical Queries" function in PubMed. Are you asking a question of etiology? If so, the same search string used in clinical queries with etiology narrow chosen will yield 84 citations, and we haven't even limited to English language or by date! Hope this is helpful.