Machine learning for identifying Randomized Controlled Trials: An evaluation and practitioner's guide

• Published in: Research synthesis methods Vol. 9; no. 4; pp. 602 - 614
• Main Authors: Marshall, Iain J., Noel‐Storr, Anna, Kuiper, Joël, Thomas, James, Wallace, Byron C.
• Format: Journal Article
• Language: English
• Published: England Wiley-Blackwell 01.12.2018; Wiley Subscription Services, Inc; John Wiley and Sons Inc
• Subjects: Accuracy; Algorithms; Artificial intelligence; Artificial neural networks; Classification; Clinical trials; Computer Software; Databases, Bibliographic; Documentation; Evidence-Based Medicine; Guidance; Humans; Information Retrieval; Information Storage and Retrieval - methods; Information Storage and Retrieval - standards; Machine Learning; Medical Research; Networks; Neural networks; Open Source Technology; Outcomes of Treatment; Randomization; Randomized Controlled Trials; Randomized Controlled Trials as Topic; Registries; Reproducibility of Results; Review Literature as Topic; ROC Curve; Role; Search Engine; Search Strategies; Sensitivity; Sensitivity and Specificity; Special Issue Paper; Special Issue Papers; Subject Headings; Support Vector Machine; Support vector machines; Workflow
• ISSN: 1759-2879; 1759-2887
• DOI: 10.1002/jrsm.1287

Machine learning (ML) algorithms have proven highly accurate for identifying Randomized Controlled Trials (RCTs) but are not used much in practice, in part because the best way to make use of the technology in a typical workflow is unclear. In this work, we evaluate ML models for RCT classification (support vector machines, convolutional neural networks, and ensemble approaches). We trained and optimized support vector machine and convolutional neural network models on the titles and s of the Cochrane Crowd RCT set. We evaluated the models on an external dataset (Clinical Hedges), allowing direct comparison with traditional database search filters. We estimated area under receiver operating characteristics (AUROC) using the Clinical Hedges dataset. We demonstrate that ML approaches better discriminate between RCTs and non‐RCTs than widely used traditional database search filters at all sensitivity levels; our best‐performing model also achieved the best results to date for ML in this task (AUROC 0.987, 95% CI, 0.984‐0.989). We provide practical guidance on the role of ML in (1) systematic reviews (high‐sensitivity strategies) and (2) rapid reviews and clinical question answering (high‐precision strategies) together with recommended probability cutoffs for each use case. Finally, we provide open‐source software to enable these approaches to be used in practice.