Attention-Based Hierarchical Recurrent Neural Network for Phenotype Classification

This paper focuses on labeling phenotypes of patients in Intensive Care Unit given their records from admission to discharge. Recent works mainly rely on recurrent neural networks to process such temporal data. However, such prevalent practice, which leverages the last hidden state in the network for sequence representation, falls short when dealing with long sequences. Moreover, the memorizing strategy inside the recurrent units does not necessarily identify the key health records for each specific class. In this paper, we propose an attention-based hierarchical recurrent neural network (AHRNN) for phenotype classification. Our intuition is to remember all the past records by a hierarchical structure and make predictions based on crucial information in the label’s perspective. To the best of our knowledge, it is the first work of applying attention-based hierarchical neural networks to clinical time series prediction. Experimental results show that our model outperforms the state-of-the-arts in accuracy, time efficiency and model interpretability.