Modeling asynchronous event sequences with RNNs

• Published in: Journal of biomedical informatics Vol. 83; pp. 167 - 177
• Main Authors: Wu, Stephen, Liu, Sijia, Sohn, Sunghwan, Moon, Sungrim, Wi, Chung-il, Juhn, Young, Liu, Hongfang
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.07.2018
• Subjects: Asthma; Asthma - classification; Child; Child, Preschool; Computer Simulation; Deep learning; Electronic Health Records; Humans; Infant; Intensive Care Units - statistics & numerical data; Neural Networks, Computer; Recurrence; Temporal data; Deep learning; Temporal data; Electronic health records; Asthma
• ISSN: 1532-0464; 1532-0480
• DOI: 10.1016/j.jbi.2018.05.016

[Display omitted] •Asynchronous clinical data can be modeled as event sequences.•Event sequences are not necessarily synchronous, even, or co-cardinal.•Recurrent Neural Networks (RNNs) can support timing alongside sequence values.•Chronic diseases (asthma) and critical care (ICU mortality) differ in event density.•Chronic diseases and critical care have different optimal RNN architectures. Sequences of events have often been modeled with computational techniques, but typical preprocessing steps and problem settings do not explicitly address the ramifications of timestamped events. Clinical data, such as is found in electronic health records (EHRs), typically comes with timestamp information. In this work, we define event sequences and their properties: synchronicity, evenness, and co-cardinality; we then show how asynchronous, uneven, and multi-cardinal problem settings can support explicit accountings of relative time. Our evaluation uses the temporally sensitive clinical use case of pediatric asthma, which is a chronic disease with symptoms (and lack thereof) evolving over time. We show several approaches to explicitly incorporating relative time into a recurrent neural network (RNN) model that improve the overall classification of patients into those with no asthma, those with persistent asthma, those in long-term remission, and those who have experienced relapse. We also compare and contrast these results with those in an inpatient intensive care setting.