An integrated LSTM-HeteroRGNN model for interpretable opioid overdose risk prediction

Abstract

Opioid overdose (OD) has become a leading cause of accidental death in the United States, and overdose deaths reached a record high during the COVID-19 pandemic. Combating the opioid crisis requires targeting high-need populations by identifying individuals at risk of OD. While deep learning emerges as a powerful method for building predictive models using large scale electronic health records (EHR), it is challenged by the complex intrinsic relationships among EHR data. Further, its utility is limited by the lack of clinically meaningful explainability, which is necessary for making informed clinical or policy decisions using such models. In this paper, we present LIGHTED, an integrated deep learning model combining long short term memory (LSTM) and graph neural networks (GNN) to predict patients' OD risk. The LIGHTED model can incorporate the temporal effects of disease progression and the knowledge learned from interactions among clinical features. We evaluated the model using Cerner's Health Facts database with over 5 million patients. Our experiments demonstrated that the model outperforms traditional machine learning methods and other deep learning models. We also proposed a novel interpretability method by exploiting embeddings provided by GNNs to cluster patients and EHR features respectively, and conducted qualitative feature cluster analysis for clinical interpretations. Our study shows that LIGHTED can take advantage of longitudinal EHR data and the intrinsic graph structure of EHRs among patients to provide effective and interpretable OD risk predictions that may potentially improve clinical decision support.

Keywords: Clinical decision support; Deep learning; Electronic health records; Graph neural network; Long short-term memory; Opioid overdose; Opioid poisoning.

Fig. 1

Flowchart of selecting patients.

Fig. 2

Encounters for prediction and building feature matrix.

Fig. 3

Structure of input feature matrix and model.

Fig. 4

Structure of heterogeneous relational graph.

Fig. 5

An example visual representation of features in a heterogeneous graph. The vector for each node is updated by aggregating node vectors of its neighborhood nodes and its own vector from the last layer. The feature node vector will be initiated by one-hot encoding.

Fig. 6

Structure of integrated LSTM-GNN model.

Fig. 7

The workflow of applying clustering algorithms on graph embeddings.

Fig. 8

(A) ROC curves for different models. (B) Specificity and sensitivity curves of LIGHTED over different probability thresholds. (C) Average time for epoch in training phase for each model.

Fig. 9

Top 50 important features identified by Shapley value. Features in yellow are related to pain/opioid/drug misuse, features in blue are related to mental disorders and features in green are related to respiratory system. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)