The role of MEDLINE, EMBASE and CENTRAL in systematic literature searches of Cochrane Reviews

Article type
Authors
Klerings I1, van Noord M2, Mittermayer T3, Nussbaumer-Streit B1, Wagner G1, Dobrescu A4, Gartlehner G1
1Cochrane Austria
2Medical Center Library & Archives, School of Medicine, Duke University
3Ludwig Boltzmann Institution for Health Technology Assessment
4Genetics Department, Victor Babes University of Medicine and Pharmacy Timisoara
Abstract
Background: A high-quality systematic review should be based on a comprehensive literature search. The Methodological Expectations of Cochrane Intervention Reviews stipulate that CENTRAL, MEDLINE and EMBASE need to be searched, regardless of topic. Since literature identification and selection is a very time-consuming part of the review process, having a well-founded understanding of the yield of these databases is important, both when choosing additional information sources and when streamlining the search process for rapid reviews.

Objectives: Our objective is to assess how searches in CENTRAL, MEDLINE and EMBASE lead to included references in Cochrane Reviews. We seek to examine the yield of these databases in systematic searches on a range of topics, and estimate the impact of limiting the number of databases used in rapid reviews. We want to better understand if non-retrieval of studies is due to database choice, limitations of the search strategy, or the type of publication (i.e. grey or published literature) included in the systematic reviews.

Methods: As part of a larger methods project of the Cochrane Rapid Review Methods Group (1), we randomly chose 60 Cochrane Reviews on 5 topics: cardiovascular disease, cerebrovascular disease, osteoarthritis, chronic respiratory conditions, or mental health. For each review, we identified the information sources used in the original searches (e.g. databases, hand search). We checked to see if included references were indexed in CENTRAL, MEDLINE or EMBASE and also assessed the share of grey literature among the references. We re-ran the searches to verify that the indexed references were retrieved by the reported search strategies. We will assessed the database coverage and the recall of the search strategies (per database and cumulatively) for each review individually and based on topic.
Results will be available at the Summit.

Conclusions: Will be available at the Summit.

(1) Nussbaumer-Streit B, et.al. Assessing the validity of abbreviated literature searches for rapid reviews: protocol of a non-inferiority and meta-epidemiologic study. Sys rev. 2016 Nov 22;5(1):197.