Article type
Year
Abstract
Introduction: One of the key principles of a systematic review is the identification and consideration of all relevant trials. In 1994, the Cochrane Collaboration developed an optimal search strategy for the retrieval of controlled trials from MEDLINE using Ovid and SilverPlatter (Dickersin et al, 1994). This search strategy has been used by the Collaboration to identify thousands of controlled trials now available for consideration in systematic reviews. PubMed provides free web based searching of MEDLINE and PREMEDLINE. Because of a number of features, including the provision of more timely access to citations, reviewers and others will increasingly turn to PubMed as a means of identifying trials.
Objective: To develop an optimal search strategy to retrieve controlled trials using PubMed.
Methods: A comparison of the results from PubMed and MEDLINE (as accessed via Ovid3.0 using the Cochrane optimal search strategy) was completed to determine the relevant software and terminology differences. Differences in retrieval were examined and the test strategy revised. The test strategy was then compared to a "gold standard". The gold standard consisted of hand and electronic search results for 11 general US health care journals for 1970, 1981 and 1992 previously analysed as part of a pilot study of the sensitivity and yield of hand and MEDLINE searching. Because the Publication Type tag CONTROLLED CLINICAL TRIAL was introduced in 1995, hand and electronic search results for the same 11 journals for 1996 were added to the gold standard.
Results: Rather than simply translating the Cochrane optimal search strategy for MEDLINE, PubMed specific features necessitated the development of a new strategy. For instance, in PubMed the use of 'adj' or 'near' is currently not available and result sets cannot be combined in the same manner as possible with Ovid or SilverPlatter accessed MEDLINE. Finally, there have been changes in the MeSH tree structure that have occurred since the development of the original MEDLINE strategy that were incorporated into the design of the test strategy. Comparison of the test strategy with the gold standard is currently being completed and results will be available for presentation at the Colloquium.
Discussion: Failure to identify all relevant trials for a systematic review could result in biased findings. An optimal search strategy for the retrieval of controlled trials was developed for use with PubMed increasing the ability of reviewers to identify all possibly relevant trials.
Objective: To develop an optimal search strategy to retrieve controlled trials using PubMed.
Methods: A comparison of the results from PubMed and MEDLINE (as accessed via Ovid3.0 using the Cochrane optimal search strategy) was completed to determine the relevant software and terminology differences. Differences in retrieval were examined and the test strategy revised. The test strategy was then compared to a "gold standard". The gold standard consisted of hand and electronic search results for 11 general US health care journals for 1970, 1981 and 1992 previously analysed as part of a pilot study of the sensitivity and yield of hand and MEDLINE searching. Because the Publication Type tag CONTROLLED CLINICAL TRIAL was introduced in 1995, hand and electronic search results for the same 11 journals for 1996 were added to the gold standard.
Results: Rather than simply translating the Cochrane optimal search strategy for MEDLINE, PubMed specific features necessitated the development of a new strategy. For instance, in PubMed the use of 'adj' or 'near' is currently not available and result sets cannot be combined in the same manner as possible with Ovid or SilverPlatter accessed MEDLINE. Finally, there have been changes in the MeSH tree structure that have occurred since the development of the original MEDLINE strategy that were incorporated into the design of the test strategy. Comparison of the test strategy with the gold standard is currently being completed and results will be available for presentation at the Colloquium.
Discussion: Failure to identify all relevant trials for a systematic review could result in biased findings. An optimal search strategy for the retrieval of controlled trials was developed for use with PubMed increasing the ability of reviewers to identify all possibly relevant trials.