BACKGROUND: Systematic review articles support the advance of science and translation of research evidence into healthcare practice. Inaccurate retrieval from medline could limit access to reviews. OBJECTIVE: To determine the quality of indexing systematic reviews and meta-analyses in medline. METHODS: The Clinical Hedges Database, containing the results of a hand search of 161 journals, was used to test medline indexing terms for their ability to retrieve systematic reviews that met predefined methodologic criteria (labelled as 'pass' review articles) and reviews that reported a meta-analysis. RESULTS: The Clinical Hedges Database contained 49 028 articles; 753 were 'pass' review articles (552 with a meta-analysis). In total 758 review articles (independent of whether they passed) reported a meta-analysis. The search strategy that retrieved the highest number of 'pass' systematic reviews achieved a sensitivity of 97.1%. The publication type 'meta analysis' had a false positive rate of 5.6% (95% CI 3.9 to 7.6), and false negative rate of 0.31% (95% CI 0.26 to 0.36) for retrieving systematic reviews that reported a meta-analysis. CONCLUSIONS: Inaccuracies in indexing systematic reviews and meta-analyses in medline can be partly overcome by a 5-term search strategy. Introducing a publication type for systematic reviews of the literature could improve retrieval performance.
BACKGROUND: Systematic review articles support the advance of science and translation of research evidence into healthcare practice. Inaccurate retrieval from medline could limit access to reviews. OBJECTIVE: To determine the quality of indexing systematic reviews and meta-analyses in medline. METHODS: The Clinical Hedges Database, containing the results of a hand search of 161 journals, was used to test medline indexing terms for their ability to retrieve systematic reviews that met predefined methodologic criteria (labelled as 'pass' review articles) and reviews that reported a meta-analysis. RESULTS: The Clinical Hedges Database contained 49 028 articles; 753 were 'pass' review articles (552 with a meta-analysis). In total 758 review articles (independent of whether they passed) reported a meta-analysis. The search strategy that retrieved the highest number of 'pass' systematic reviews achieved a sensitivity of 97.1%. The publication type 'meta analysis' had a false positive rate of 5.6% (95% CI 3.9 to 7.6), and false negative rate of 0.31% (95% CI 0.26 to 0.36) for retrieving systematic reviews that reported a meta-analysis. CONCLUSIONS: Inaccuracies in indexing systematic reviews and meta-analyses in medline can be partly overcome by a 5-term search strategy. Introducing a publication type for systematic reviews of the literature could improve retrieval performance.