Research Fellow, Center for Academic and Health Policies, Tehran University of Medical Sciences, Tehran, Iran Researcher, Systematic Review Study Group, Students' Scientific Research Center, Tehran University of Medical Sciences, Tehran, Iran and Resident of Otorhinolaryngology: Mashhad University of Medical Sciences, Mashhad, Iran Researcher, Systematic Review Study Group, Students' Scientific Research Center, Tehran University of Medical Sciences, Tehran, Iran.
J Eval Clin Pract. 2013 Oct;19(5):723-6. doi: 10.1111/j.1365-2753.2010.01554.x. Epub 2010 Sep 16.
The aim of this study was to compare the performance of three search methods in the retrieval of relevant clinical trials from PubMed to answer specific clinical questions.
Included studies of a sample of 100 Cochrane reviews which recorded in PubMed were considered as the reference standard. The search queries were formulated based on the systematic review titles. Precision, recall and number of retrieved records for limiting the results to clinical trial publication type, and using sensitive and specific clinical queries filters were compared. The number of keywords, presence of specific names of intervention and syndrome in the search keywords were used in a model to predict the recalls and precisions.
The Clinical queries-sensitive search strategy retrieved the largest number of records (33) and had the highest recall (41.6%) and lowest precision (4.8%). The presence of specific intervention name was the only significant predictor of all recalls and precisions (P = 0.016).
The recall and precision of combination of simple clinical search queries and methodological search filters to find clinical trials in various subjects were considerably low. The limit field strategy yielded in higher precision and fewer retrieved records and approximately similar recall, compared with the clinical queries-sensitive strategy. Presence of specific intervention name in the search keywords increased both recall and precision.
本研究旨在比较三种检索方法在从 PubMed 中检索相关临床试验以回答特定临床问题方面的性能。
纳入的研究为在 PubMed 中记录的 100 项 Cochrane 综述的样本，作为参考标准。检索查询是基于系统评价标题制定的。比较了限制结果为临床试验出版类型、使用敏感和特异性临床查询过滤器时的精度、召回率和检索记录数。在一个模型中使用关键字的数量、干预和综合征的特定名称在搜索关键字中的存在来预测召回率和精度。
临床查询敏感搜索策略检索到的记录最多（33 条），召回率最高（41.6%），精度最低（4.8%）。特定干预名称的存在是所有召回率和精度的唯一显著预测因素（P = 0.016）。
组合简单临床搜索查询和方法学搜索过滤器以在各种主题中查找临床试验的召回率和精度相当低。与临床查询敏感策略相比，字段限制策略的精度更高，检索记录更少，召回率大致相似。搜索关键字中存在特定的干预名称会提高召回率和精度。
