Exploiting complex medical data with interpretable deep learning for adverse drug event prediction
作者：
Highlights：
• Prediction of ADEs from heterogeneous medical records using RNNs.
• A new RNN architecture, augmenting RETAIN to permit ADE risk factors.
• Further augmentation RETAIN by including clinical text features.
• Empirical evaluation of various architecture configurations.
• Demonstration of the importance of interpretability supported by attention mechanisms.
摘要
•Prediction of ADEs from heterogeneous medical records using RNNs.•A new RNN architecture, augmenting RETAIN to permit ADE risk factors.•Further augmentation RETAIN by including clinical text features.•Empirical evaluation of various architecture configurations.•Demonstration of the importance of interpretability supported by attention mechanisms.
论文关键词：Deep learning,Text mining,Explainable AI,Adverse drug events,Medical records
论文评审过程：Received 18 November 2019, Revised 18 June 2020, Accepted 10 August 2020, Available online 15 September 2020, Version of Record 25 September 2020.
论文官网地址：https://doi.org/10.1016/j.artmed.2020.101942
