Systematic reviews of test accuracy should search a range of databases to identify primary studies

J Clin Epidemiol. 2008 Apr;61(4):357-364. doi: 10.1016/j.jclinepi.2007.05.013. Epub 2007 Oct 15.

Abstract

Objective: To estimate the yield from searching a range of bibliographic databases and additional sources to identify test accuracy studies for systematic reviews.

Study design and setting: We examined eight systematic reviews and their database searches: MEDLINE, EMBASE, BIOSIS, Science Citation Index, LILACS, Pascal, and CENTRAL. We used studies included in each systematic review as the "gold standard," against which yield was estimated. For each database, we classified studies in each gold standard set as being (1) included in the database and identified by searches, (2) included and not identified, and (3) not included in the database.

Results: No search identified all studies in any gold standard set. EMBASE, Science Citation Index, and BIOSIS contained studies that were not on MEDLINE. Over 20% of studies in the gold standard sets were not identified by searching MEDLINE. Six studies on LILACS were not on any other database. Eight gold standard studies were not included in any of the databases, and a further 22 were not identified by the electronic search strategies.

Conclusions: Systematic reviews of test accuracy studies should search a range of databases. Even searches designed to be very sensitive, that do not use study design filters, can fail to identify relevant studies.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Databases, Bibliographic
  • Evidence-Based Medicine / methods*
  • Humans
  • Information Storage and Retrieval / methods
  • Information Storage and Retrieval / standards*
  • Meta-Analysis as Topic
  • Predictive Value of Tests*
  • Research Design
  • Review Literature as Topic*