Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predictive Modeling with Temporal Graphical Representation on Electronic Health Records

Jiayuan Chen, Changchang Yin, Yuanlong Wang, Ping Zhang

TL;DR

The paper tackles predictive modeling from electronic health records by requiring representations that capture both temporal dynamics across patient visits and the graph-structured relationships among medical events. It introduces a temporal heterogeneous graph with time-aware visit and medical-event nodes, and a Temporal Graph Transformer (TRANS) that fuses temporal embedding, spatial encoding, and heterogeneous graph convolutions. Key contributions include the novel graph construction, the TRANS architecture, Time2Vec and functional time encoding for temporal information, meta-path-based spatial encoding, and a graph explainer for interpretability, all validated on MIMIC-III, MIMIC-IV, and CCAE with state-of-the-art results. The work advances healthcare AI by delivering accurate diagnosis predictions along with interpretable insights into COPD-related comorbidities in real-world datasets, with potential for broader clinical applicability.

Abstract

Deep learning-based predictive models, leveraging Electronic Health Records (EHR), are receiving increasing attention in healthcare. An effective representation of a patient's EHR should hierarchically encompass both the temporal relationships between historical visits and medical events, and the inherent structural information within these elements. Existing patient representation methods can be roughly categorized into sequential representation and graphical representation. The sequential representation methods focus only on the temporal relationships among longitudinal visits. On the other hand, the graphical representation approaches, while adept at extracting the graph-structured relationships between various medical events, fall short in effectively integrate temporal information. To capture both types of information, we model a patient's EHR as a novel temporal heterogeneous graph. This graph includes historical visits nodes and medical events nodes. It propagates structured information from medical event nodes to visit nodes and utilizes time-aware visit nodes to capture changes in the patient's health status. Furthermore, we introduce a novel temporal graph transformer (TRANS) that integrates temporal edge features, global positional encoding, and local structural encoding into heterogeneous graph convolution, capturing both temporal and structural information. We validate the effectiveness of TRANS through extensive experiments on three real-world datasets. The results show that our proposed approach achieves state-of-the-art performance.

Predictive Modeling with Temporal Graphical Representation on Electronic Health Records

TL;DR

The paper tackles predictive modeling from electronic health records by requiring representations that capture both temporal dynamics across patient visits and the graph-structured relationships among medical events. It introduces a temporal heterogeneous graph with time-aware visit and medical-event nodes, and a Temporal Graph Transformer (TRANS) that fuses temporal embedding, spatial encoding, and heterogeneous graph convolutions. Key contributions include the novel graph construction, the TRANS architecture, Time2Vec and functional time encoding for temporal information, meta-path-based spatial encoding, and a graph explainer for interpretability, all validated on MIMIC-III, MIMIC-IV, and CCAE with state-of-the-art results. The work advances healthcare AI by delivering accurate diagnosis predictions along with interpretable insights into COPD-related comorbidities in real-world datasets, with potential for broader clinical applicability.

Abstract

Deep learning-based predictive models, leveraging Electronic Health Records (EHR), are receiving increasing attention in healthcare. An effective representation of a patient's EHR should hierarchically encompass both the temporal relationships between historical visits and medical events, and the inherent structural information within these elements. Existing patient representation methods can be roughly categorized into sequential representation and graphical representation. The sequential representation methods focus only on the temporal relationships among longitudinal visits. On the other hand, the graphical representation approaches, while adept at extracting the graph-structured relationships between various medical events, fall short in effectively integrate temporal information. To capture both types of information, we model a patient's EHR as a novel temporal heterogeneous graph. This graph includes historical visits nodes and medical events nodes. It propagates structured information from medical event nodes to visit nodes and utilizes time-aware visit nodes to capture changes in the patient's health status. Furthermore, we introduce a novel temporal graph transformer (TRANS) that integrates temporal edge features, global positional encoding, and local structural encoding into heterogeneous graph convolution, capturing both temporal and structural information. We validate the effectiveness of TRANS through extensive experiments on three real-world datasets. The results show that our proposed approach achieves state-of-the-art performance.
Paper Structure (40 sections, 14 equations, 8 figures, 6 tables)

This paper contains 40 sections, 14 equations, 8 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Preliminary
  3. Medical Codes
  4. EHR Data
  5. Sequential Diagnosis Prediction
  6. Methodology
  7. Graph Construction
  8. Temporal Heterogeneous Message Passing
  9. Temporal Embedding
  10. Heterogeneous Message Passing
  11. Spatial Encoding
  12. Patient Explainer
  13. Experiments
  14. Experimental Setting
  15. Datasets
  16. ...and 25 more sections

Figures (8)

  • Figure 1: Different EHR representation methods. (a) An example of sequential representation. (b) Examples of graphical representation, which include graphs at the visit level and the patient level. (c) The temporal graphical representation proposed in this paper.
  • Figure 2: The overall framework of TRANS. The input EHR data is first constructed into a heterogeneous graph containing visit nodes and medical event nodes. Then the node features are mapped to an embedding space and combined with structural and sequential information. Subsequently, a temporal graph transformer is used to aggregate the node information. Finally, the features of the last visit node are input into a predictor.
  • Figure 3: t-SNE Scatterplots of medical codes Learned by DDHGNN and TRANS on the CCAE dataset.
  • Figure 4: Distribution of visits lengths on the CCAE dataset. 'a-b' represents the visit length interval (a, b].
  • Figure 5: Visualization of Temporal Embeddings.
  • ...and 3 more figures