Searching for diagnostic test accuracy studies: an application to screening for open angle glaucoma (OAG)

Tags: Poster
Fraser C, Mowatt G, Siddiqui R, Burr J

Background: A systematic review of glaucoma diagnostic test accuracy was undertaken as part of a health technology assessment of the effectiveness of OAG screening. The assessment also required systematic reviews of epidemiology, disease progression and economic evaluations. Extensive literature searching, using separate sets of searches for each review, was undertaken. For diagnostic test accuracy, databases searched included MEDLINE, EMBASE, Science Citation Index, BIOSIS, the Cochrane Central Register of Controlled Trials (CENTRAL), HMIC (Health Management Information Consortium) and full text searching of selected ophthalmology journals. In total, 4839 reports were identified of which 636 were selected for full text assessment with 100 meeting the initial inclusion criteria.

Objectives: To assess the contribution of a MEDLINE/EMBASE search, incorporating a diagnostic filter, in identifying studies of diagnostic test accuracy.

Methods: The MEDLINE search strategy included a filter, to identify diagnostic test accuracy studies, that had been previously developed and validated against gold standards derived from handsearching and was used in conjunction with search terms relating to glaucoma and the relevant diagnostic tests (n=14). The diagnostic filter was translated for EMBASE.

Results: The inclusion of the diagnostic filter reduced the number of hits in the MEDLINE/EMBASE deduplicated search by 69.9% (from 13,870 to 4159). Of the 97 relevant reports, indexed in MEDLINE and/or EMBASE, 89 were retrieved by searching and screening of the MEDLINE/EMBASE results. Three were excluded by the search (two due to the filter) and five were missed from the screening of titles and abstracts. These losses were attributable to poor reporting or absence of abstracts. Additional reports were identified from other searches for the review (seven) or in the course of identifying reports for the epidemiology review (four).

Conclusions: Despite extensive searching, only six reports were not identified by the MEDLINE/EMBASE search (three not indexed, three missed). The filter considerably reduced the retrievals with the loss of only two reports attributable to its use.