PECODR Research Plan


Hypothesis: Additional indexing of medical information using the controlled vocabulary of PECODR elements would facilitate access to this group of manuscripts and enhance the ability of health professionals to find answers to their questions.

Research Plan: Taxonomy Identification and production of controlled vocabulary
The first phase of this research will be to identify existing taxonomy for the PECODR elements. We will then use colleagues in the department of family medicine and colleagues in the field of clinical information retrieval to confirm that we have all relevant taxonomy. In the likely situation of some PECODR elements not having a validated taxonomy we will generate these and validate them using a Delphi approach (Jones and Hunter 1995). This process requires fifteen to twenty volunteers to help with an iterative validation of the taxonomy using e-mail to correspond. They will first be asked to exclude, add to, or clarify the taxonomy terms. This information from the Delphi correspondents will be aggregated to produce a revised taxonomy. This will then be sent back to the Delphi group for one or two more cycles of revision depending on the amount of agreement within the group.

The second important question is how the enhanced indexing system can benefit the search. First, the identification of the PECODR element will help solve the ambiguity problem (a term can be used in more than one element as in the example given above). A clear identification of the PECODR element  can prevent a term from wrongly matching another one in a different element.  Second, with the semantic relations created between terms in the controlled vocabulary, it will be possible to match more elements in an article with elements in the question, leading to a higher relevance in the information displayed by the search. In the prototype that we will develop, instead of re-indexing the whole PubMed database, we will use a lightweight approach, which re-rank the conventional PubMed search results according to the additional PECODR taxonomy approach. Those results that match the question on the basis of this new technique will be ranked higher. This approach has been used by Demner-Fushman et al (Demner-Fushman and Lin 2007) and Zhang et al (Zhang, Shen et al. 2004) and proved to be effective.
During the development phase, the indexing method will be tested using a sample of actual questions of health professionals’ taken from question databases (Ely, Osheroff et al. 2007) and a sample of the NLM database from several years. In addition we will include a set of articles recognized by clinicians as highly important and that have been republished in abstract form by the American College of Physicians Journal Club and Evidence Based Medicine Journal.

PECODR Index search engine assessment:  To determine if the PECODR enhanced index is more effective than default search engines with respect to three outcomes: 
(1) Relevance – results match PECODR elements in question
 (2) Importance - contain a results with high citation counts (adjusted for year of publication)
 (3) Impact - clinically important articles for clinicians