Using recurrent neural network models for early detection of heart failure onset

By Edward Choi, Andy Schuetz, Walter F. Stewart, Jimeng Sun. 2016

825 citations recorded by the source corpus.

Objective: We explored whether use of deep learning to model temporal relations among events in electronic health records (EHRs) would improve model performance in predicting initial diagnosis of heart failure (HF) compared to conventional methods that ignore temporality. Materials and Methods: Data were from a health system’s EHR on 3884 incident HF cases and 28 903 controls, identified as primary care patients, between May 16, 2000, and May 23, 2013. Recurrent neural network (RNN) models using gated recurrent units (GRUs) were adapted to detect relations among time-stamped events (eg, disease diagnosis, medication orders, procedure orders, etc.) with a 12- to 18-month observation window of cases and controls. Model performance metrics were compared to regularized logistic regression, neural network, support vector machine, and K-nearest neighbor classifier approaches. Results: Using a 12-month observation window, the area under the curve (AUC) for the RNN model was 0.777, compared to AUCs for logistic regression (0.747), multilayer perceptron (MLP) with 1 hidden layer (0.765), support vector machine (SVM) (0.743), and K-nearest neighbor (KNN) (0.730). When using an 18-month observation window, the AUC for the RNN model increased to 0.883 and was significantly higher than the 0.834 AUC for the best of the baseline methods (MLP). Conclusion: Deep learning models adapted to leverage temporal relations appear to improve performance of models for detection of incident heart failure with a short observation window of 12–18 months.

Subfields: Artificial Intelligence, Cardiology and Cardiovascular Medicine, Health Information Management

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Cardiology and Cardiovascular Medicine

In Health Information Management

Citations in the corpus

Interpretable and Steerable Sequence Learning via Prototypes (2019, methodology): “, electronic health records (EHR) in health care [10, 15], click streams in software applications and vehicle fault logs in automobiles.”
Interpretable Representation Learning for Healthcare via Capturing Disease Progression through Time (2018, methodology): “Many other convolutional neural networks [4, 21] and recurrent neural networks [9, 25] have been applied to modeling EHR data for different prediction tasks, but their focus is slightly outside of the scope of this paper.”
Recurrent Neural Networks for Classifying Relations in Clinical Notes (2017, positive): “[39] applied Gated Recurrent Unit networks (GRUs) for early detection of heart failure onset using time-stamped medical events (diagnosis, medications and procedures).”
GRAM: Graph-based Attention Model for Healthcare Representation Learning (2016, methodology): “…(HF) cohort, which is a subset of Sutter PAMF data for a heart failure onset prediction study with 3.4K HF cases chosen by a set of criteria described in Vijayakrishnan et al. (2014); Gurwitz et al. (2013) and 27K matching controls chosen by a set of criteria described in Choi et al. (2016e).”
Converging blockchain and next-generation artificial intelligence technologies to decentralize and accelerate biomedical research and healthcare (2015, methodology): “Recently, RNNs were used to predict heart failure of patients based on clinical events in their records [44].”
Learning to Detect Sepsis with a Multitask Gaussian Process RNN Classifier (2017, methodology): “…(Lipton et al., 2016) used Long-Short Term Memory (LSTM) RNNs to predict diagnosis codes given physiological time series from the ICU, and (Choi et al., 2016) used Gated Recurrent Unit RNNs to predict onset of heart failure using categorical time series of diagnosis and procedural codes.”
Task-oriented Dialogue System for Automatic Diagnosis (2018, background): “Experimental results prove the effectiveness of the identification of heart failure (Jonnalagadda et al., 2017; Choi et al., 2016), type 2 diabetes (Li et al.”
MetaPred: Meta-Learning for Clinical Risk Prediction with Limited Patient Electronic Health Records (2019, background): “Many computational algorithms have been developed to overcome these challenges, including both conventional approaches [6] and deep learning models [8].”
"The human body is a black box": supporting clinical decision-making with deep learning (2019, methodology): “Individual patient cases were also visualized to show how MGP-RNN often detected sepsis hours before the clinical diagnosis.”
An Improved Multi-Output Gaussian Process RNN with Real-Time Validation for Early Sepsis Detection (2017, methodology): “In particular, (Lipton et al. (2016a)) use Long-Short Term Memory (LSTM) RNNs to predict diagnosis codes given physiological time series from the ICU, and (Choi et al. (2016b)) use Gated Recurrent Unit RNNs to predict onset of heart failure using categorical time series of billing codes.”
Artificial Intelligence and Surgical Decision-Making (2019, methodology): “19,26,41,64 When applied to the same variable set used to calculate SOFA scores, deep models outperform traditional SOFA modeling in predicting in-hospital mortality for ICU patients.”
The History Began from AlexNet: A Comprehensive Survey on Deep Learning Approaches (2018, methodology): “From the application point of view, RNNs can solve different types of problems which need different architectures of RNNs shown in Fig.”
Adaptive Laplace Mechanism: Differential Privacy Preservation in Deep Learning (2017, background): “Furthermore, deep learning has applications in a number of healthcare areas, e.g., phenotype extraction and health risk prediction [3], prediction of the development of various diseases, including schizophrenia, cancers, diabetes, heart failure, etc. [4], and many more.”
Artificial Intelligence in Cardiology. (2018, methodology): “(36) used an RNN to predict heart failure diagnosis from EHRs.”
AI in Health: State of the Art, Challenges, and Future Directions (2019, background): “[34] leveraged RNN to predict the onset risk of Congestive Heart Failure (CHF).”