Saturday, November 11, 2006

Refining clinical questions for more effective and relevant search results

The level of detail in an information request from a user, whether in received in-person the library, by email or telephone, or on clinical rounds, is not always as high as we might like. If the question is imperfectly defined or lacks clarity in intent, it can be very difficult to be certain that you are retrieving the "right" information from the literature, which also will likely cause problems to the requestor if he/she receives results that differ from the situation tthe information was intended to address. If you receive a fairly broad question, how do you refine the question?

Some have suggested the PICO format (Patient or Problem, Intervention, Comparison, Outcome) as one means of refining a clinical question (more information on PICO at bottom of post).

Consider this situation -- the question in the current case study may have originally looked something like "When should colistin be used in the intensive care unit?" How well does this question fit into the PICO format?

Patient/Problem: ICU patients
Intervention: colistin
Comparison: not defined
Outcome: not defined

If you received this question on the reference desk from the clinician treating the patient considered in the case study, rather than on bedside rounds where you'd be hearing the details of the situation and would have a better idea of why colistin is being debated, how would you work with the user to refine this query?

From a searching perspective, it seems as if the Patient/Problem portion of this query may be a key initial target for refining this question. There are a multitude of reasons why a patient might be in the ICU, but a somewhat more limited list of reasons why a clinician might consider administering an antibiotic such as colistin. We know that, because of the increasing problems of antibiotic resistance, particularly in the intensive care setting, clinicians are working to reserve antibiotic administration for situations in which it is truly needed. Asking the requestor a brief question such as "What conditions might you consider treating with colistin in the ICU?" would likely be very useful in understanding this question in the absence of additional patient details.

You also may quickly query the user considering the Comparison and Outcomes elements of the format: "Are you weighing colistin against another treatment option?" and "Are there particular outcomes are you particularly interested in (e.g. radiologic confirmation of infection resolution, mortality, morbidity, ventilator days, length of stay)?"

Refining the question to include a focus on Acinetobacter pneumonia in this case improves the search results dramatically.

A PubMed search based on the initial, broad question may look something like, retrieving about 90 citations:

(colistin OR colistimethate) AND (critical care OR intensive care OR intensive care unit OR critical illness OR ICU) (link to these search results)

Focusing on patients with Acinetobacter pneumonai using the strategy employed within this month's case leads to this strategy, retrieving about 30 citations:

colistin[mh] OR colistin[tiab] OR colistin[substance name] OR colistimethate[tiab] OR colistimethate[substance name]) AND (Acinetobacter[mh] OR Acinetobacter infections[mh] OR Acinetobacter[tiab]) AND (pneumonia[tiab] OR pneumonia[mh]) (link to these search results)
The first search fails to retrieve approximately two-thirds of the citations retrieved by the search focused on Acinetobacter pneumonia, a striking lack of overlap that may have a significant impact on the search results you might provide the requestor with. Refining the query with a few quick follow-up questions for the clinician can enable a much more productive and efficient, and likely much more relevant, searching process.


A few additional resources considering the use of PICO for clinical question refinement:
- A group of library and information sciences students from the University of British Columbia have drafted a few graphical representations of this structure that may be useful in picturing how this format might be used in structuring and refining a search query (via Dean Giustini's Google Scholar blog).

- The National Library of Medicine also has developed an interface for searching PubMed that operationalizes this format.

- An article by Booth, O'Rourke, and Ford (Bull Med Libr Assoc. 2000 July; 88(3): 239–246) discusses the impact of PICO structure use on the pre-search reference interview, including a form for mediated search requests based on PICO elements.

- The Univerity of Washington's HealthLinks has a very practical guide for translating clinical questions into PICO format, complete with practice questions.

0 Comments:

Post a Comment

<< Home