Do published search filters to identify diagnostic test accuracy studies perform adequately?

Abstract

To assess the performance of published search filters in finding diagnostic test accuracy studies. Diagnostic test accuracy search filters were identified by searching medline, our own files and by requesting unpublished filters from colleagues. We applied the filters to a case study review of diagnostic test accuracy studies for urinary tract infections (UTI) in young children. The included studies with records in medline formed the gold standard. The performance of the filters in finding those gold standard records was assessed. We identified twenty-three diagnostic test accuracy search filters for use with medline. The case study systematic review of UTI included 179 studies of diagnostic test accuracy, of which 160 were available in medline. The filters showed a wide range of sensitivities (range: 20.6% to 86.9%) and precision (range: 1% to 9.4%). Our results broadly support those reported in two other studies. The search filters tested do not offer an adequate trade-off between sensitivity and precision to be used to identify studies for systematic reviews. However, there are methods available to explore whether search filters are viable based on an objective statistical analysis of the text and indexing used in records.