Making literature searches easier: a rapid and sensitive search filter for retrieving randomized controlled trials from PubMed

Abstract

To develop and test the sensitivity and precision of a rapid and simple search filter (RSSF), suitable for busy clinicians wanting to find randomised controlled trials (RCTs) in PubMed. Ideally it should retrieve all the RCTs, but as few irrelevant studies as possible, and be easy to use. The RSSF consisted of the search term 'Randomized Controlled Trial' limited to the Publication Type field. Journals that published the highest numbers of diabetes RCTs between 2000 and 2005 were identified, and then handsearched in order define a set of known RCTs. The sensitivity of the RSSF was tested by measuring the proportion of the known RCTs retrieved, and the precision by checking the proportion of the retrieved studies which were RCTs. The RSSF was compared to a highly sensitive search strategy (HSSS) developed for PubMed. Embase was checked for trials not in PubMed. Sixteen journals were found to contain half of all published RCTs in diabetes. 820 diabetes RCTs were identified by handsearching. Measured against these, the RSSF gave a sensitivity of 96.0% (95% CI, 94.8% to 97.1%), and a precision of 93.6% (95% CI 91.7% to 95.0%). Compared to the HSSS, the RSSF reduced the filtering required by 87%. An Embase search for diabetes RCTs found 36 (2.1%) not in PubMed. A rapid simple search filter for PubMed can find almost all diabetes RCTs, while excluding most studies not required, thereby greatly reducing the time cost of searching and filtering results, and of searching other databases.