A PubMed search filter for efficiently retrieving exercise training studies.

Abstract

Background: A barrier to evidence-informed exercise programming is locating studies of exercise training programs. The purpose of this study was to create a search filter for studies of exercise training programs for the PubMed electronic bibliographic database.

Methods: Candidate search terms were identified from three sources: exercise-relevant MeSH terms and their corresponding Entry terms, word frequency analysis of articles in a gold-standard reference set curated from systematic reviews focused on exercise training, and retrospective searching of articles retrieved in the search filter development and testing steps. These terms were assembled into an exercise training search filter, and its performance was assessed against a basic search string applied to six case studies. Search string performance was measured as sensitivity (relative recall), precision, and number needed to read (NNR). We aimed to achieve relative recall ≥ 85%, and a NNR ≥ 2.

Results: The reference set consisted of 71 articles drawn from six systematic reviews. Sixty-one candidate search terms were evaluated for inclusion, 21 of which were included in the finalized exercise-training search filter. The relative recall of the search filter was 96% for the reference set and the precision mean ± SD was 54 ± 16% across the case studies, with the corresponding NNR = ~ 2. The exercise training search filter consistently outperformed the basic search string.

Conclusion: The exercise training search filter fosters more efficient searches for studies of exercise training programs in the PubMed electronic bibliographic database. This search string may therefore support evidence-informed practice in exercise programming.