18 September 2007

A modest proposal

Would this post sound more authoritative if I entitled it: "Predictive value of information literacy (IL) levels for recurrent mediocre search results (RMSR)?"

Last July I commented on a review article by Karen Davies on the information-seeking behaviour of physicians. The article appeared in the June 2007 issue of Health information and libraries journal. There were no real surprises in Davies' review. It's the old problem. Things are still in statu quo. Doctors have information needs and they have a wealth of resources to find the answers to their questions; but they do not have the time or the skills to get the best results from the tools at hand. Librarians have the time and training to provide the best resources and to meet physicians' information needs by using these resources in an expert manner; but we are often left waiting by the phone.

There are many physicians who are using the Internet intelligently to answer clinical questions and who have made the effort to get training in its proper use. And then there are the others who conform to all the bleak analyses of Karen Davies' article. What mystifies me is that not only are there physicians who don't get beyond "quick and dirty" searching or who simply don't know what they are doing, some even go so far as to hang that dirty laundry in public — even in the pages of prestigious medical journals. In an earlier post I discussed one such disaster, and now I must regretfully point out another.

When I came across the one-page "How I use the internet" in the Australian family physician, my first thought was, "Lucky patient you were. The first item that popped up in your doctor's web search was only a four-year-old article. Things could have been worse." The target of my scorn is a brief "practice tip" which, minus the illustrations, takes up half of page 538 in volume 36. Written by a general practitioner in Adelaide, it is an unintentionally humorous description of using the Internet to find information about discontinuing warfarin for a female patient who had suffered an episode of deep vein thrombosis. The author decides to conduct a "quick, computerised literature search using the Yahoo! search engine." He picks Yahoo! because the patient, a hockey player, is familiar with it, and he wants to allow her to be involved with her treatment.

In Yahoo! he enters the query, "Can d-dimer levels predict recurrent venous thromboembolism?" He justifies using this form of query by citing The well-built clinical question (an online tutorial developed by librarians Connie Schardt, Duke University Medical Center Library and Jill Mayer, University of North Carolina at Chapel Hill Health Science Library). "Entering key words would not help to construct a question and phrase it in such a way as to facilitate an answer," he writes. I had to read this sentence a few times, but its author seems to be claiming that entering a syntactically correct English-language question into a web search engine is preferable to a key word search. Cue Twilight Zone theme.

Our physician does not understand the difference between proper use of a web search engine and constructing a well built clinical question according to the principles of EBM. Like Captain Kirk consulting the ship's computer, he dutifully asks his question. And a fine one it is; but sticking to only this formulation of the question radically limits the possibilities for retrieval. It is an approach free of empirical duty. Had he consulted a librarian, our physician would have found that the skilled employment of key words in a Yahoo! search produces more recent articles on this topic, like one published just last year in the NEJM, which in turn cites a clinical trial that concludes: "The optimal duration of oral anticoagulant treatment in patients with idiopathic venous thromboembolism is still uncertain."

What our physician's quick and dirty method does produce as its first item is a JAMA article from 2003. Without further ado, he uses the information from this article to advise the patient, who — trusting soul — is "happy with the outcome of the search result." Aside from the problematic acceptance of Yahoo!'s ranking algorithm, why use only Yahoo! in the first place? Surely he must be aware of Google Scholar. A search that plugs the same question into that search engine turns up relevant articles from Circulation and Archives of internal medicine in the top ten. More importantly, why has he ignored PubMed, ClinicalTrials.gov, and other free EBM resources? And surely his young patient would have benefited more from being directed to something like the MedlinePlus article on thrombophlebitis or JAMA's own patient page on the same topic, which cites a 2006 JAMA article with new information on other measures for identifying patients at low risk for recurrent venous thromboembolism.

In his book The trouble with medical journals, Richard Smith quotes an aphorism that he valued highly as editor of BMJ: "Anybody is free to make a fool of himself in my journal" (p. 64). This was in the context of a discussion of BMJ's policy of allowing rapid, sometimes egregious, responses by email to published articles (a practice that was later tightened up by the editorial board). But in the case of the "practice tip" under review here, the editors must have read it and approved it. What were they thinking? This is a tip the way Swift's A modest proposal is social planning.

"The time constraints of a consultation" are cited as the excuse for this travesty of medical information retrieval. Our physician would have required less time to email the question to his local health sciences library and have them send the information. "How I use the internet" looks to me like a perfect demonstration of the dire need for improved information literacy for health care providers. Here is my modest proposal. If they truly wish patients to continue, as Mr. Micawber would say, in the enjoyment of salubrity, medical people and medical journal editors should view such amateurish squibs with greater suspicion. In short, they should not be published. Or next time, the patient may not be so lucky.