ProAID: path-based reasoning for self-attentional disease prediction

摘要

Due to introduction and increased availability of Electronic Health Records (EHRs), disease prediction has recently gained immense research attention and achieved impressive progress. Existing methods are based on RNN-like architectures, which treat every disease equally, and learn the representations from medical knowledge. However, strong structural information among diseases is ignored in these methods. In this paper, we introduce a novel Path-based reasoning model for self-AttentIonal Disease prediction (ProAID), which utilizes medical paths extracted from patient EHR and external medical knowledge bases to augment the latent interaction between diseases and learn highly representative patient embeddings. By explicitly incorporating medical paths, ProAID effectively generates embeddings that capture the hierarchical information of diseases and learn effective representations of a patient based on the historical patient admission sequences in her/his EHRs to allow accurate disease prediction for the next hospital admission. Extensive experiments on public medical datasets show that ProAID achieves better performance than the compared methods, which indicates the effectiveness of the proposed model.