Department of Epidemiology and Biostatistics, College of Public Health, University of Georgia, Athens, GA, USA.
HRB Centre for Primary Care Research, Royal College of Surgeons in Ireland, Dublin, Ireland.
J Clin Epidemiol. 2020 Sep;125:26-29. doi: 10.1016/j.jclinepi.2020.05.013. Epub 2020 May 13.
The aim of the study was to develop an improved search strategy for clinical prediction rules.
We first refined a list of 30 primary care-relevant journals and improved the efficiency of the Haynes Narrow Filter/Teljour/Murphy Inclusion Filter with 26 items by removing one term (Modified Haynes 26 filter). We then developed the "Royal College of Surgeons in Ireland (RCSI) filter" and compared it with the modified HNF/TMIF26 for its ability to detect prediction rules in the primary care literature. All abstracts and, if necessary, full text were reviewed independently in parallel by primary care physicians. The key outcomes were the percentage of prediction rules identified out of the total identified by both search strategies (sensitivity) and the number of articles that had to be reviewed to identify them (efficiency).
The Modified Haynes 26 filter returned 1,701 abstracts vs. 1,062 for the RCSI filter. The RCSI filter identified 105 of 111 of all prediction rules identified by either filter, compared with 107 of 111 by the Modified Haynes 26 filter (94.6% vs. 96.4%; P = 0.52). In addition, 9.9% of abstracts found using the RCSI filter were prediction rules, compared with only 6.3% using the Modified Haynes 25 filter (P = 0.001).
We have developed a novel "RCSI filter" that more efficiently identifies prediction rules in the medical literature.
本研究旨在开发一种改进的临床预测规则搜索策略。
我们首先改进了一份包含 30 种初级保健相关期刊的清单，并通过删除一个术语（改良 Haynes 26 过滤器）提高了 Haynes 窄过滤器/Teljour/Murphy 纳入过滤器的效率，得到 26 个项目。然后，我们开发了“爱尔兰皇家外科医学院（RCSI）过滤器”，并比较了改良 HNF/TMIF26 过滤器，以评估其在初级保健文献中检测预测规则的能力。所有摘要，如有必要，全文均由初级保健医生独立平行审查。主要结果是两种搜索策略识别出的预测规则总数的百分比（敏感性）和识别出这些规则所需的文章数量（效率）。
改良 Haynes 26 过滤器返回 1701 篇摘要，而 RCSI 过滤器返回 1062 篇。RCSI 过滤器识别出两种过滤器中所有预测规则的 105 个，而改良 Haynes 26 过滤器识别出 111 个中的 107 个（94.6%对 96.4%；P=0.52）。此外，使用 RCSI 过滤器发现的 9.9%的摘要为预测规则，而使用改良 Haynes 25 过滤器仅发现 6.3%的预测规则（P=0.001）。
我们开发了一种新的“RCSI 过滤器”，可更有效地识别医学文献中的预测规则。
