Electronic Thesis and Dissertation Repository


Doctor of Philosophy


Epidemiology and Biostatistics


Dr. Amit Garg

2nd Supervisor

Dr. Mark Speechley

Joint Supervisor


Background: Clinicians frequently search PubMed to guide patient care. This study investigated the factors that impact successful searching and the utility of PubMed search filters.

Methods: A random sample of nephrologists was surveyed between 2008 and 2010; 160 valid responses were received (72% response rate). One group of 60 respondents was presented with the same two clinical questions, while the other 100 were each presented with a unique clinical question. The clinical questions were based on recently published systematic reviews. Respondents provided the search terms they would type into PubMed to address their clinical question(s). All physician-provided searches were executed in PubMed and outcome measures of sensitivity (proportion of relevant articles found) and precision (proportion of all articles found that are relevant) were calculated. Primary studies included in the reviews served as the reference standards of relevant articles. For the first group of respondents, the associations between the search query or nephrologist characteristics and search outcomes were investigated through multivariable regression modeling. For the second group, three types of filters were applied to the physician-provided searches: one designed to identify high quality studies about treatment (‘methods’), one designed to identify studies with renal content (‘content’) and one designed to limit searches to journals that publish renal evidence (‘journal’). Search outcomes of the non-filtered and filter-aided searches were compared using the Wilcoxon Signed-Rank test.

Results: Multifaceted searching (e.g. using MeSH, limits) improved sensitivity (rate ratio[RR]:2.6; 95% CI:1.4-5.0) and precision (RR:2.0; 95% CI:1.3-3.3). The addition of concept terms decreased sensitivity (RR:0.7; 95% CI:0.5-0.9), while increasing precision (RR:1.6; 95% CI:1.3-2.0). No associations were evident between nephrologist characteristics and sensitivity. However, physicians who previously received training in literature searching produced searches with better precision (RR:2.3; 95% CI:1.4-3.6). The combined use of the ‘methods’ and ‘content’ filters produced the largest improvement in precision with no change in sensitivity, compared to non-filtered searches (median difference:5.5%; 95% CI:2%-12%).

Conclusions: Use of multifaceted searching and filters can improve physician-provided searches in PubMed. Literature training curricula should adopt the findings from this study. Improved search performance has the potential to enhance clinical practice and improve patient care.