Comparison of two different search strategies in identifying literature for a diagnostic test accuracy review of Down's Syndrome screening

Tags: Oral
Alldred S, Hampson L, Neilson J, Alfirevic Z, Deeks J

Background: It is well known that identification of all relevant studies, and documentation of search strategies so that they can be reproduced is what distinguishes a systematic review from a traditional literature review. Literature regarding diagnostic tests is often fragmented and difficult to gather. The Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy (CHSRDTA) (draft version) recommends that search strategies should primarily focus on terms for index tests and the target condition. It also recommends that methodology filters for diagnostic accuracy studies should be avoided if possible. The justification for this approach is that indexing for diagnostic tests is poor, and that the emphasis should be on comprehensiveness/sensitivity rather than precision in identifying papers.

Objectives: To determine whether retrieval of papers suitable for inclusion in a systematic review of Down's syndrome screening is adversely biased by the application of a methodology filter for diagnostic accuracy studies within the search strategy.

Methods: Search strategies were constructed and run in MEDLINE and EMBASE. The first strategy focussed around terms in

three categories - the target condition, the diagnostic tests in question and the study population. The second strategy used a methodological filter for identifying diagnostic accuracy studies. Papers suitable for inclusion were identified according to methods stipulated in the handbook (CHSRDTA). The two strategies were compared to see whether there was any significant difference in terms of identified papers and the sensitivity and specificity was calculated.

Results: For the comprehensive search the yield was 5947 from MEDLINE and 5658 from EMBASE. The yield from the methodological filter search was 4697 from MEDLINE and 4380 from EMBASE. Screening of the abstracts is in process. The sensitivity and specificity of the search using a filter will be compared with the original search and results will be presented.