25 September 2007

Another modest proposal

How not to answer clinical questions

No sooner had I put the final touches on my last post about physicians' overconfidence when trawling the web, than my colleague Michael Tennenhouse alerted me to an article in this month's Canadian family physician that should give patients pause (1). It's like what they say about ketchup: if you do not shake the bottle, none'll come and then a lot'll. Somebody please stop the flow.

Here is yet another example of what can go wrong when physicians set out to conquer the proverbial ocean of information without compass and sextant. Our Canadian crew, all family medicine residents at various Canadian universities, show more nautical expertise than the hapless landlubber in the Australian family physician (2). Unfortunately, they all could use some more lessons in navigation.

The Canadian article cites PICO also as the authors attempt to phrase an appropriate clinical question — although in my opinion their question leaves much to be desired. The Canadians then go on to complete a questionable search. On the surface things look reasonable. Yahoo! is not even on the map, and the authors are at least aware that "'Googling' your question does not necessarily lead to the best evidence-based answer" (p. 1535). PubMed is visited, "the obvious starting point." We then turn the rudder towards SUMsearch and TRIP.

A Quebec resource called InfoClinique is described as "excellent," but we are immediately told that it is available only to Laval University members. I tried it. It works slowly, searching a limited list of EBM sources (curiously, neither TRIP nor SUMsearch appears among its sources indéxées). When the results finally appear, the links are hidden behind lots of little x's. Recommending the unhelpful InfoClinique looks more like a well-meant attempt at Canadian two-solitudes inclusivism than a contribution to the authors' task.

Having steered around these shoals the authors glance briefly at textbooks, among which they include eMedicine, UpToDate, Dynamed and the resources to be found in MD Consult. They then sum things up by proudly describing how they answered their own question in under 10 minutes. Land ho!

While the general approach of the authors is acceptable, there are serious problems with their method. As my colleague pointed out to me, they choose mostly appropriate limits, but the actual search term selection, resulting from the poorly worked out question, is problematical. Their article nicely demonstrates some common mistakes made by many of our students and residents:

  1. Poor question building, especially with respect to identifying the problem and the intervention.
  2. The use of "red-herring" or unnecessary terms in the search strategy which artificially limit the results (in this case it's the term "expectant").
  3. Uncritical review of the search strategy and results.
  4. No use of MeSH in PubMed, or how to easily seek out the best MeSH from the most relevant retrieval, or how to use Related Article searching.
The authors argue in their second paragraph (p. 1535) that they have merely reproduced what most physicians would do given the case under consideration (premature rupture of membranes). They choose to use "the Internet alone" (i.e., free resources, except for InfoClinique). They choose to look at only "articles available without subscription, as we do not always have time to ask the library for an article when a patient is in front of us." Using a "variety of key words" their goal is "to mimic what is often done in practice." In other words, they are quite prepared to shortchange a patient, to flounder about on the web's treacherous expanse with poor preparation and ill-chosen search strategies, in order to save time and perhaps some expense. If this is truly mimicking what is done in practice, more's the pity.

In a recent post, T. Scott reiterates the point that physicians are drowning in information. Like most of us, he is "worrying about the health of the patients that are being served by physicians who have only the faintest notion of how to construct even the simplest search." As health sciences librarians it is "our responsibility to get [students] out of med school with a decent set of information management skills."

I couldn't agree more. So here is my other modest proposal. Let the aspiring family physician pass both these articles by. The Canadians and Australians have shown you how perilous the voyage can be in search of the best evidence. Seeing the library as little more than an obstacle in their way, they try to make the best of it on their own without capsizing. Sometimes they're lucky, in the same way that contestants in a bathtub race sometimes reach the finish line. But what furious paddling, what desperate flailing about all the while.

So pass these articles by. If you keep looking, however, surely something will, as they say, turn up.
"Precisely. It is evident to me that Australia is the legitimate sphere of action for Mr. Micawber! I entertain the conviction, my dear madam," said Mr. Micawber, "that it is, under existing circumstances, the land, the only land, for myself and family; and that something of an extraordinary nature will turn up on that shore."


1. Lacasse M, Lafortune V, Bartlett L, Guimond J. Answering clinical questions: What is the best way to search the Web? Can Fam.Physician 2007 Sep;53(9):1535-8. http://www.cfp.ca/cgi/content/full/53/9/1535

2. Qureshi F. How I use the internet. Aust.Fam.Physician 2007;36(7):538. http://www.racgp.org.au/afp/200707/17421