Article type
Year
Abstract
Background: While preparing a systematic review on "antibiotics duration for treating uncomplicated, symptomatic lower urinary tract infections in elderly women", we compared different search engines of the MEDLINE database.
Methods: We applied the same (simplified) search strategy the same day (25.01.2000) to four different MEDLINE search engines. The total number of retrieved references, the number of unique and common references as well as the number of references included in our review were calculated.
Results: GratefulMed retrieved not only the highest total number of references but also the greatest number of articles relevant to our systematic review (see table). Ovid and Knowledge Finder showed the greatest concordance of references (599 identical), Knowledge Finder and GratefulMed the lowest number of same references (528). Comparing the four search engines, 294 references were found by only one provider, 8 by two systems, 77 by three, and 524 were retrieved by all four search engines. Provider References identified Unique references References included in the review after reading the titles n=339 abstracts n=170 papers n=58 GratefulMed 831 265 109 72 37 PubMed 607 0 63 45 27 Ovid 600 0 62 45 27 Knowledge Finder 599 (633)* 29 58 42 26 * Knowledge Finder showed only 599 of 633 most relevant references.
Conclusion: Users of the MEDLINE database should be aware that the results of their search do not only depend on the search strategy but also on the search engine used. No clear explanation for the differences in retrieval rate could be found in our study, although differences in the way the search terms are processed (sophistication of the use of boolean operators; mode of mapping to MESH terms) may be contributing factors.
Methods: We applied the same (simplified) search strategy the same day (25.01.2000) to four different MEDLINE search engines. The total number of retrieved references, the number of unique and common references as well as the number of references included in our review were calculated.
Results: GratefulMed retrieved not only the highest total number of references but also the greatest number of articles relevant to our systematic review (see table). Ovid and Knowledge Finder showed the greatest concordance of references (599 identical), Knowledge Finder and GratefulMed the lowest number of same references (528). Comparing the four search engines, 294 references were found by only one provider, 8 by two systems, 77 by three, and 524 were retrieved by all four search engines. Provider References identified Unique references References included in the review after reading the titles n=339 abstracts n=170 papers n=58 GratefulMed 831 265 109 72 37 PubMed 607 0 63 45 27 Ovid 600 0 62 45 27 Knowledge Finder 599 (633)* 29 58 42 26 * Knowledge Finder showed only 599 of 633 most relevant references.
Conclusion: Users of the MEDLINE database should be aware that the results of their search do not only depend on the search strategy but also on the search engine used. No clear explanation for the differences in retrieval rate could be found in our study, although differences in the way the search terms are processed (sophistication of the use of boolean operators; mode of mapping to MESH terms) may be contributing factors.