P.002 Improving the quality of systematic reviews of neurological conditions with more accurate search strategies: a series of validation studies

Abstract

Background: A well-constructed search strategy is an important feature of any systematic review. We aimed to design and validate electronic database (e.g. Pubmed) search strategies (i.e. a hedge or series of words used to identify articles of interest) for six neurological conditions. Methods: We enumerated 10311 consecutive articles in the 21 highest impact factor English-language general neurology journals. We constructed a simple hedge, limited to one keyword, for each condition. We also constructed a complex hedge using a series of MeSH terms and keywords. Two reviewers independently reviewed (confirmed by a third reviewer) all articles and established which condition(s) were the article’s subject. We calculated sensitivity/specificity estimates for the simple and complex hedges, and compared these using McNemar’s test. Results: The results are summarized in the Table. Conclusions: Our complex hedges for most conditions dramatically improve sensitivity without compromising specificity. This study will help improve the accuracy of search strategies in future systematic reviews.