Use of study design search filters, search strategy quality and reviewers’ screening burden in prognosis, diagnosis and effectiveness systematic reviews




Poster session 2 Thursday: Evidence synthesis - methods / improving conduct and reporting


Thursday 14 September 2017 - 12:30 to 14:00


All authors in correct order:

Jordan JL1, Corp N1, van der Windt DA1
1 Research Institute for Primary Care and Health Sciences, Keele University, United Kingdom
Presenting author and contact person

Presenting author:

Joanne Jordan

Contact person:

Abstract text
Background: Search filters are used to limit records downloaded from databases to specific study designs when comprehensive searches are conducted. For example, Cochrane Highly Sensitive Search (HSS) filter is recommended for retrieving RCTs in Cochrane SRs. However, filters are not currently advised for prognosis and diagnosis SRs due to lack of precision and potential to miss important studies.
Aim: To evaluate and compare search filter use, reviewers’ screening burden (Number needed to Read (NNR)) and quality of search strategies in recently published prognosis, diagnosis and effectiveness SRs.
Methods: Purposive sample of 30 freely available, recent SRs (10 effectiveness (3 Cochrane SRs), 10 prognosis, 10 diagnosis reviews) in musculoskeletal diseases selected. Data were extracted on search details, number of records downloaded and studies included in SRs. NNR was calculated for each review. Available search strategies assessed independently by 2 information specialists using Peer Review of Electronic Search Strategies (PRESS).
Results: Overall, 5 diagnostic, 8 prognostic and 5 effectiveness SRs used search terms for study design/type. Cochrane HSS filter was the only published filter referenced in any of the 30 SRs and was used in 2 of the effectiveness SRs (both Cochrane SRs). NRR ranged from 4 to 1291 depending on comprehensiveness of search strategies, but did not appear to vary according to type of SR. However, this was difficult to judge as topic areas varied considerably. Quality of search strategies using PRESS checklist, and impact on NNRs, will be discussed during the conference.
Conclusions: Despite advice not to use filters for diagnostic and prognostic SRs, reviewers continue to restrict SR searches by using study design specific terms, but not use specifically designed published filters. In effectiveness SRs, where efficient RCT filters exist and are recommended, many reviewers are not using them. This highlights the need for designing and evaluating high-quality search filters in several areas, and better implementation of the established Cochrane RCT search filters, in order to improve the efficiency and quality of systematic reviews.