MADEx: A System for Detecting Medications, Adverse Drug Events, and Their Relations from Clinical Notes

• Published in: Drug safety 2019-01, Vol.42 (1), p.123-133
• Main Authors: Yang, Xi, Bian, Jiang, Gong, Yan, Hogan, William R., Wu, Yonghui
• Format: Article
• Language: English
• Subjects: Algorithms; Artificial intelligence; Drug Safety and Pharmacovigilance; Electronic health records; Electronic medical records; Information processing; Learning algorithms; Long short-term memory; Machine learning; Medicine; Medicine & Public Health; Natural language processing; Neural networks; NLP Challenge for Detecting Medication and Adverse Drug Events from Electronic Health Records (MADE 1.0); Original Research Article; Pharmacology; Pharmacology/Toxicology; Pharmacovigilance; Product safety; Recurrent neural networks; Researchers; Safety; Strategy; Support vector machines; Surveillance; Unstructured data; World Wide Web
• ISSN: 0114-5916; 1179-1942
• DOI: 10.1007/s40264-018-0761-0

Introduction Early detection of adverse drug events (ADEs) from electronic health records is an important, challenging task to support pharmacovigilance and drug safety surveillance. A well-known challenge to use clinical text for detection of ADEs is that much of the detailed information is documented in a narrative manner. Clinical natural language processing (NLP) is the key technology to extract information from unstructured clinical text. Objective We present a machine learning-based clinical NLP system—MADEx—for detecting medications, ADEs, and their relations from clinical notes. Methods We developed a recurrent neural network (RNN) model using a long short-term memory (LSTM) strategy for clinical name entity recognition (NER) and compared it with baseline conditional random fields (CRFs). We also developed a modified training strategy for the RNN, which outperformed the widely used early stop strategy. For relation extraction, we compared support vector machines (SVMs) and random forests on single-sentence relations and cross-sentence relations. In addition, we developed an integrated pipeline to extract entities and relations together by combining RNNs and SVMs. Results MADEx achieved the top-three best performances (F1 score of 0.8233) for clinical NER in the 2018 Medication and Adverse Drug Events (MADE1.0) challenge. The post-challenge evaluation showed that the relation extraction module and integrated pipeline (identify entity and relation together) of MADEx are comparable with the best systems developed in this challenge. Conclusion This study demonstrated the efficiency of deep learning methods for automatic extraction of medications, ADEs, and their relations from clinical text to support pharmacovigilance and drug safety surveillance.