Trialstreamer: a living, automatically updated database of clinical trial reports

摘要: Objective Randomized controlled trials (RCTs) are the gold standard method for evaluating whether a treatment works in healthcare, but can be difficult to find and make use of. We describe development evaluation of system automatically categorize all new RCT reports. Materials Methods Trialstreamer, continuously monitors PubMed WHO International Clinical Trials Registry Platform (ICTRP), looking RCTs humans using validated classifier. combine machine learning rule-based methods extract information from abstracts, including free-text descriptions trial populations, interventions outcomes (the 9PICO9) map these snippets normalised MeSH vocabulary terms. additionally identify sample sizes, predict risk bias, text conveying key findings. store extracted data database which we freely available download, via search portal, allows users enter structured clinical queries. Results ranked prioritize larger higher-quality studies. Results As May 2020, have indexed 669,895 publications RCTs, 18,485 were published first four months 2020 (144/day). include 303,319 registrations ICTRP. The median size was 66. Conclusions We present an automated finding categorising RCTs. This yields novel resource: A humans. daily updates this on our website (trialstreamer.robotreviewer.net).