Testing filter term performance in PsycINFO to identify evidence syntheses in crime reduction, using the relative recall method

Abstract

Objectives: To test filter term performance against an original search strategy to identify evidence syntheses with a crime reduction outcome in the PsycINFO database, with a view to maximising efficiency and/or effectiveness in the search phase of a systematic review. Methods: A search strategy was developed to identify evidence syntheses with crime reduction outcomes. A ‘quasi-gold standard’ set of 255 relevant studies that were indexed in the PsycINFO database was derived from this initial work and was used to test various filter terms available in the database using the relative recall method. Precision and sensitivity statistics were generated for each search strategy. Results: Seven search strategies were tested using three clusters of index terms, on (1) method filter terms, (2) topic filter terms and (3) method and topic filter terms. These were applied as filters for the original search strategy and, to facilitate comparison, against all records in PsycINFO. The most sensitive filter scored 74.1%, the most precise scored 44.1% and the best compromise between sensitivity and precision scored 53.7% sensitivity and 16.3% precision. Conclusions: Filter term performance in PsycINFO can be used to inform search strategies used within criminology and allied fields for systematic reviews. The variety of filter terms tested here, in the absence and presence of a keyword search, caters for researchers with different information requirements. Using an evidence-based approach to systematic searching can yield considerable resource savings in conducting a systematic review.