Skip to main content

Validating prognosis search filters using relative recall based on prognosis systematic reviews

Date and Location




Sunday 22 September 2013 - 10:30 - 12:00


Presenting author and contact person

Presenting author

Robin Parker

Contact person

Robin Parker
Abstract text
Background: Methods for conducting systematic reviews of prognosis research are currently being developed and refined. The comprehensive identification of primary studies is a crucial component of systematic reviews; search filters have played a significant role in the effective retrieval of relevant studies. Previous research has demonstrated that the inconsistent use of prognosis-related language in the citations of prognosis studies makes systematic searching difficult, and presents challenges to the development and use of methodological filters. Furthermore, few prognosis reviews employ search filters to assist in citation retrieval. Validating filters based on relative recall of included studies from systematic reviews has been effectively used in the past to create test sets but this approach has not yet been widely applied in prognosis research. Objectives: We will use the included studies from prognosis reviews identified from the Prognosis Systematic Review Database (PSRD) to validate various PubMed prognosis search filters using relative recall. Methods: From reviews in the PSRD, we will identify those that used an explicit prognosis search strategy; we will include systematic reviews that meet methodological standards. Data will be extracted about the topic, search strategy, methodological search filter, and citation information for all included studies. Included studies retrieved through PubMed will serve as a test set for validation of the modified prognosis filters. Results: For the PSRD, we used a sensitive strategy to search 5 high impact journals (all years) and select prognosis systematic reviews. Relative recall using modified prognosis filters will be compared to published precision of the Clinical Queries prognosis filter. We will explore differences in effectiveness of the filters for different research topics and for different types of prognosis systematic reviews. Conclusions: Using a relative recall approach, we will test the recall and precision of published and modified prognosis filters to inform best practices for prognosis systematic reviews.