Email updates

Keep up to date with the latest news and content from BMC Medical Informatics and Decision Making and BioMed Central.

Open Access Research article

Developing optimal search strategies for detecting clinically sound and relevant causation studies in EMBASE

R Brian Haynes12*, Monika Kastner1, Nancy L Wilczynski1 and the Hedges Team

Author Affiliations

1 Health Information Research Unit, Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, L8N 3Z5, Canada

2 Department of Medicine, Faculty of Health Sciences, McMaster University, Hamilton, Ontario, L8N 3Z5, Canada

For all author emails, please log on.

BMC Medical Informatics and Decision Making 2005, 5:8  doi:10.1186/1472-6947-5-8

Published: 22 March 2005

Abstract

Background

Evaluating the existence and strength of an association between a putative cause and adverse clinical outcome is complex and best done by assessing all available evidence. With the increasing burden of chronic disease, greater time demands on health professionals, and the explosion of information, effective retrieval of best evidence has become both more important and more difficult. Optimal search retrieval can be hampered by a number of obstacles, especially poor search strategies, but using empirically tested methodological search filters can enhance the accuracy of searches for sound evidence concerning etiology. Although such filters have previously been developed for studies of relevance to causation in MEDLINE, no empirically tested search strategy exists for EMBASE.

Methods

An analytic survey was conducted, comparing hand searches of journals with retrievals from EMBASE for candidate search terms and combinations. 6 research assistants read all issues of 55 journals indexed in EMBASE. All articles were rated using purpose and quality indicators and categorized into clinically relevant original studies, review articles, general papers, or case reports. The original and review articles were then categorized as 'pass' or 'fail' for scientific merit according to explicit criteria in the areas of causation (etiology) and other clinical topics. Candidate search strategies were developed for causation, then run in a subset of 55 EMBASE journals, the retrievals being compared with the hand search data. The sensitivity, specificity, precision, and accuracy of the search strategies were calculated.

Results

Of the 1489 studies classified as causation, 14% were methodologically sound. When search terms were combined, sensitivity reached 92%. Compared with the best single-term strategy, the best combination of terms resulted in an absolute increase in sensitivity (19%) and specificity (5.2%). Maximizing specificity for combined terms resulted in an increase of 7.1% compared with the single term but this came at an expense of sensitivity (39% absolute decrease). A search strategy that optimized the trade-off between sensitivity and specificity achieved 81.9% for sensitivity and 81.4% for specificity.

Conclusion

We have discovered search strategies that retrieve high quality studies of causation from EMBASE with high sensitivity, high specificity, or an optimal balance of each.