Appraisal of: Karimi S, Pohl S, Scholer F, Cavedon L, Zobel J. Boolean versus ranked querying for biomedical systematic reviews. BMC Med Inform Decis Mak. 2010; 10(1):58.
Boolean searching of biomedical databases is fundamental to the systematic review process but has the disadvantage of requiring the examination of all retrieved references, while the formulation of effective Boolean queries is complex, can require multiple iterations before a search strategy is finalised and can still fail to retrieve all relevant records. This study compared the performance of Boolean searching with ranked retrieval (in which results are ranked in order of relevance) when searching for a systematic review by a series of experiments. Investigations regarding ranked retrieval included: the effectiveness of different ranked query schemes; the effectiveness of exploiting metadata in MEDLINE records; the effectiveness of query expansion using MeSH; defining a stopping criterion for determining when to stop examining ranked documents; and combining Boolean and ranked retrieval in a hybrid search method. Results showed that ranked retrieval by itself was not viable for a search task requiring high recall such as for a systematic review, but that a refinement of the standard Boolean search process involving ranking within a Boolean result set can improve the overall search performance by providing early indication of the quality of the results, thereby speeding up the iterative query-refinement process. The authors conclude that an interactive query-development process using a hybrid ranked and Boolean retrieval system has the potential to contribute significant time-savings to the search process for a systematic review.