Article type
Year
Abstract
Background: Finding relevant information in databases is the cornerstone of evidence-based medicine. However, searching databases can be challenging. Recently, sensitive and specific search strategies were developed to find information in MEDLINE for geriatric medicine quickly and easily. These search strategies proved to have excellent sensitivity and specificity.
Objective: To test the operating characteristics of geriatric search strategies in the Cochrane Database of Systematic Reviews (CDSR).
Methods: A reference set was obtained from the CDSR. One assessor labeled all 6437 reviews and protocols published until March 2011 as either relevant to geriatric medicine or not. We translated the search terms from MEDLINE into Cochrane. Then we compared the reviews that were found with the search strategy with the labeled reference set obtained from the CDSR.
Results: Only 373 (6%) of the articles in the CDSR were considered relevant to geriatric medicine. The most sensitive search strategy had a sensitivity of 65%, a specificity of 76% and a number-needed-to-read (NNR) of 7. The most specific search strategy had a sensitivity of 45%, a specificity of 98% and a NNR of 1.4.
Conclusion: A small percentage of reviews in the CDSR contain information on the elderly. Compared to the performance in MEDLINE, the sensitive search strategy performed slightly worse in the CDSR. Although it found two thirds of the relevant reviews, there were a lot of false positive findings resulting in a higher NNR. The specific search strategy retrieved very few false positives. However, half of the relevant reviews were missed. Differences in the performance of the search strategies could be due to a lower prevalence of relevant articles in the CDSR as well as the use of different criteria for considering articles as relevant to geriatric medicine. By using the CDSR as another reference set we have proven that the geriatric search strategies are robust.
Objective: To test the operating characteristics of geriatric search strategies in the Cochrane Database of Systematic Reviews (CDSR).
Methods: A reference set was obtained from the CDSR. One assessor labeled all 6437 reviews and protocols published until March 2011 as either relevant to geriatric medicine or not. We translated the search terms from MEDLINE into Cochrane. Then we compared the reviews that were found with the search strategy with the labeled reference set obtained from the CDSR.
Results: Only 373 (6%) of the articles in the CDSR were considered relevant to geriatric medicine. The most sensitive search strategy had a sensitivity of 65%, a specificity of 76% and a number-needed-to-read (NNR) of 7. The most specific search strategy had a sensitivity of 45%, a specificity of 98% and a NNR of 1.4.
Conclusion: A small percentage of reviews in the CDSR contain information on the elderly. Compared to the performance in MEDLINE, the sensitive search strategy performed slightly worse in the CDSR. Although it found two thirds of the relevant reviews, there were a lot of false positive findings resulting in a higher NNR. The specific search strategy retrieved very few false positives. However, half of the relevant reviews were missed. Differences in the performance of the search strategies could be due to a lower prevalence of relevant articles in the CDSR as well as the use of different criteria for considering articles as relevant to geriatric medicine. By using the CDSR as another reference set we have proven that the geriatric search strategies are robust.