Article type
Year
Abstract
Background: Embase is a key database to be searched when undertaking systematic reviews. It contains unique content but can suffer from high levels of indexing which can retrieve large numbers of irrelevant records. Using the option to focus (major) the EMTREE subject headings may help to improve the efficiency of searching.
Objectives: To identify the proportion of relevant records retrieved by searches using some or all major EMTREE subject headings in a search, rather than all (i.e. non-major) subject headings.
Methods: The relative recall approach was used to select reviews produced since 2010 by a range of agencies including Cochrane. We reran the search strategies reported in those reviews, varying the use of major EMTREE headings, to identify the impact of the changes on the retrieval of the known relevant records (included studies) in the SRs. The terms were majored in the intervention concept, the population concept and both concepts. The total number of records retrieved and the number of included studies retrieved were recorded and the sensitivity and precision of each strategy was calculated, along with the percentage decrease in sensitivity and precision from the original strategy for each of the amended strategies. We calculated the mean, median and ranges for the groups of strategies available.
Results: A total of 463 SRs were collected; 50 were randomly selected addressing a variety of topics. Focusing the intervention headings (39 SRs) reduced mean sensitivity from 71.6% to 68.8% and improved mean precision from 1% to 1.1%. Focusing the population heading (39 SRs) reduced mean sensitivity from 74.3% to 67.4% and improved precision from mean 0.9% to 1.3%. Focusing both concepts (34 SRs) reduced sensitivity from a mean of 72.8% to 63.9% and improved precision from mean of 0.8% to 1.5%.
Conclusions: This analysis of 50 SRs of widely varying topics found that focusing the EMTREE terms for the intervention, population or both, could achieve small improvements in precision, but should only be attempted when researchers are confident that their strategies are sensitive.
Objectives: To identify the proportion of relevant records retrieved by searches using some or all major EMTREE subject headings in a search, rather than all (i.e. non-major) subject headings.
Methods: The relative recall approach was used to select reviews produced since 2010 by a range of agencies including Cochrane. We reran the search strategies reported in those reviews, varying the use of major EMTREE headings, to identify the impact of the changes on the retrieval of the known relevant records (included studies) in the SRs. The terms were majored in the intervention concept, the population concept and both concepts. The total number of records retrieved and the number of included studies retrieved were recorded and the sensitivity and precision of each strategy was calculated, along with the percentage decrease in sensitivity and precision from the original strategy for each of the amended strategies. We calculated the mean, median and ranges for the groups of strategies available.
Results: A total of 463 SRs were collected; 50 were randomly selected addressing a variety of topics. Focusing the intervention headings (39 SRs) reduced mean sensitivity from 71.6% to 68.8% and improved mean precision from 1% to 1.1%. Focusing the population heading (39 SRs) reduced mean sensitivity from 74.3% to 67.4% and improved precision from mean 0.9% to 1.3%. Focusing both concepts (34 SRs) reduced sensitivity from a mean of 72.8% to 63.9% and improved precision from mean of 0.8% to 1.5%.
Conclusions: This analysis of 50 SRs of widely varying topics found that focusing the EMTREE terms for the intervention, population or both, could achieve small improvements in precision, but should only be attempted when researchers are confident that their strategies are sensitive.