Recently on MEDLIB-L (Schmidt comment, Fishel comment) there has been some controversy over the National Library of Medicine's proposal to delete 42 subheadings in MEDLINE. According to the announcement (reprinted by Schmidt), "The National Library of Medicine (NLM) has developed a restructuring of subheadings (qualifiers) within the Medical Subject Heading (MeSH) vocabulary. The new, streamlined list of 42 replaces the existing complement of 83 topical qualifiers." Here are the proposed changes (reprinted from MEDLIB-L).

The announcement states that many of the two million daily PubMed searchers are untrained users who did not receive the "intensive training in all aspects of the vocabulary and its use." The reduction of subheadings is an attempt to help first-time or infrequent users have the greater likelihood of a successful search retrieval. "The goal of MeSH continues to be a vocabulary that is useful, reproducible, and understandable."

I disagree with NLM's rational for eliminating the subheadings. I would guess that 90% of the "first-time" or "infrequent" PubMed users aren't even using the subheadings. Heck, I would say that the majority of your "first-time" or "infrequent" users aren't even searching using MeSH. If I was a betting librarian (huh, good blog name) I would say that the majority of new/infrequent PubMed users are doing Google type searches within PubMed. You know what I am talking about, phrases with quotations, incorrect boolean, etc. Considering they are probably doing this type of searching, I highly doubt they would even know that are subheadings. Eliminating the subheadings would do nothing for these type of searchers. In fact if you want to use that same rationale, why not eliminate MeSH. I am pretty certain that many of the new PubMed users aren't using MeSH and infrequent users are often confused by it. I don't care what NLM's goal is, MeSH is not useful, reproducible, nor understandable by the new or infrequent PubMed user, that is why they have generic looking search box that does the behind the scenes mapping of search terms.

I agree with Whitneydt that 83 subheadings is an awful lot of subheadings, and perhaps Cindy Schmidt's own subheading proposal would be more appropriate for NLM to consider. There should be some definite clean up and deletion of some subheadings. But NLM's rationale for this is flawed, unless they have statistics or some other sort of proof to indicate that this change is indeed needed.

What are your thoughts? If you wish, comments may be sent to Stuart J. Nelson, MD, Head, MeSH Section, NLM nelsonst@mail.nih.gov before April 10, 2006.