Article type
Year
Abstract
Objective: To test the sensitivity of full text searching for the identification of randomized controlled trials.
Methods: Six journals (American Review of Respiratory Disease, Thorax, British Journal of Diseases of the Chest, Pediatric Pulmonology, Clinical Allergy and European Respiratory Review) were hand text searched, covering different years but within the span of 1970 to 1995. These six journals were then downloaded electronically from either Medline or Embase into a Pro-Cite database, and the Cochrane Airways Group electronic search for RCTs carried out. The set of trials obtained from the text searching was removed from the set obtained electronically, and the remaining references browsed within the title, abstract and keyword fields for possible additional trials. The full papers were obtained and the number of confirmed RCTs recorded.
Results: A total of 1101 randomized controlled trials were identified by the text searching of six journals. An additional 86 (7.8%) confirmed RCTs were identified by the electronic search.
Discussion: These results show that text searching has a high sensitivity, but that an electronic search may detect more RCTs. Overall the difference in sensitivity is small, yet in view of the need to trace the very highest number of randomized controlled trials for Cochrane reviews, it is of some concern. We recommend that text searching should not be viewed as the gold standard against which electronic searches are evaluated. A combination of text and electronic searching should be used.
Methods: Six journals (American Review of Respiratory Disease, Thorax, British Journal of Diseases of the Chest, Pediatric Pulmonology, Clinical Allergy and European Respiratory Review) were hand text searched, covering different years but within the span of 1970 to 1995. These six journals were then downloaded electronically from either Medline or Embase into a Pro-Cite database, and the Cochrane Airways Group electronic search for RCTs carried out. The set of trials obtained from the text searching was removed from the set obtained electronically, and the remaining references browsed within the title, abstract and keyword fields for possible additional trials. The full papers were obtained and the number of confirmed RCTs recorded.
Results: A total of 1101 randomized controlled trials were identified by the text searching of six journals. An additional 86 (7.8%) confirmed RCTs were identified by the electronic search.
Discussion: These results show that text searching has a high sensitivity, but that an electronic search may detect more RCTs. Overall the difference in sensitivity is small, yet in view of the need to trace the very highest number of randomized controlled trials for Cochrane reviews, it is of some concern. We recommend that text searching should not be viewed as the gold standard against which electronic searches are evaluated. A combination of text and electronic searching should be used.