Effective Identification of Similar Patients Through Sequential Matching over ICD Code Embedding

Nguyen, Dang; Luo, Wei; Venkatesh, Svetha; Phung, Dinh

doi:10.1007/s10916-018-0951-4

Cited by 13 publications

(5 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Then, we use t-SNE [40] to project X into two dimensions. In the next step, we select 15 types of heart failure 3 , 3 types of essential hypertension 4 , and 7 types of acute rheumatic fever 5 that appear in MIMIC-III and plot them in Fig. 5(a).…”

Section: Case Studies For Model Interpretationmentioning

confidence: 99%

“…A variety of predictive models using deep learning technology have been proposed for predicting temporal events, such as diagnosis prediction [1]- [4], mortality prediction [5]- [7], risk prediction [8]- [10], and medicine recommendation [11], [12]. A common supervised training approach to utilize EHR data for temporal event prediction is to use previous records as features and the records of next admissions as labels.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Self-Supervised Graph Learning with Hyperbolic Embedding for Temporal Health Event Prediction

Lu,

Reddy,

Ning

2021

Preprint

View full text Add to dashboard Cite

Electronic Health Records (EHR) have been heavily used in modern healthcare systems for recording patients' admission information to hospitals. Many data-driven approaches employ temporal features in EHR for predicting specific diseases, readmission times, or diagnoses of patients. However, most existing predictive models cannot fully utilize EHR data, due to an inherent lack of labels in supervised training for some temporal events. Moreover, it is hard for existing works to simultaneously provide generic and personalized interpretability. To address these challenges, we first propose a hyperbolic embedding method with information flow to pre-train medical code representations in a hierarchical structure. We incorporate these pre-trained representations into a graph neural network to detect disease complications, and design a multi-level attention method to compute the contributions of particular diseases and admissions, thus enhancing personalized interpretability. We present a new hierarchy-enhanced historical prediction proxy task in our selfsupervised learning framework to fully utilize EHR data and exploit medical domain knowledge. We conduct a comprehensive set of experiments and case studies on widely used publicly available EHR datasets to verify the effectiveness of our model. The results demonstrate our model's strengths in both predictive tasks and interpretable abilities.

show abstract

Section: Case Studies For Model Interpretationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Self-Supervised Graph Learning with Hyperbolic Embedding for Temporal Health Event Prediction

Lu,

Reddy,

Ning

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Reference [28] proposed a locally supervised metric learning which is used for measuring similarities between patients represented by multi-dimensional time series. Nguyen et al [29] proposed the sequential matching procedure to calculate the distance between two patients, which can utilize the sequential order of medical concepts. In addition, there are also a number of patient similarity measure methods taking into account the temporal information in EHRs.…”

Section: A Patient Similaritymentioning

confidence: 99%

Patient Similarity via Joint Embeddings of Medical Knowledge Graph and Medical Entity Descriptions

2020

View full text Add to dashboard Cite

With the prevalence and growing volume of Electronic Health Records (EHRs), there has been increasing interest in mining EHRs for improving clinical decision support. The accurate identification of patients with similar conditions based on EHRs is a key step in personalized healthcare. Existing studies model EHRs by medical knowledge graph embedding to learn the latent embeddings of medical entities (e.g., patients, medications, diagnoses and procedures). However, such precisely structured data is usually limited in quantity and in scope. Therefore, to enhance the quality of the embeddings it is important to consider more widely available medical information such as medical entity descriptions. In this paper we propose a novel framework, called Deep Patient Similarity (DeepPS). Specifically, DeepPS incorporates medical entity descriptions by augmenting the embeddings of medical entities and relations with the embeddings of words, which leverages both information from medical knowledge graph structures and the contexts of medical entity descriptions. Furthermore, DeepPS employs the embeddings to patient similarity learning by leveraging Siamese Convolutional Neural Network (CNN) with Spatial Pyramid Pooling (SPP). Extensive experiments on real datasets are conducted to show superior performance of our proposed framework.INDEX TERMS Patient similarity, medical knowledge graph embedding, medical entity descriptions, Siamese CNN with SPP.

show abstract

“…A variety of predictive models using deep learning technology has been proposed for predicting temporal events, such as diagnosis prediction [1]- [5]; mortality prediction [6]- [9]; risk prediction [10]- [13]; and medication recommendation [14], [15]. A common supervised training approach to utilize EHR data for temporal event prediction is to use previous records as features and the records of next admissions as labels.…”

Section: Introductionmentioning

confidence: 99%

Self-Supervised Graph Learning With Hyperbolic Embedding for Temporal Health Event Prediction

Lü

Reddy

Ning

2023

IEEE Trans. Cybern.

View full text Add to dashboard Cite

Electronic health records (EHRs) have been heavily used in modern healthcare systems for recording patients' admission information to health facilities. Many data-driven approaches employ temporal features in EHR for predicting specific diseases, readmission times, and diagnoses of patients. However, most existing predictive models cannot fully utilize EHR data, due to an inherent lack of labels in supervised training for some temporal events. Moreover, it is hard for the existing methods to simultaneously provide generic and personalized interpretability. To address these challenges, we propose Sherbet, a self-supervised graph learning framework with hyperbolic embeddings for temporal health event prediction. We first propose a hyperbolic embedding method with information flow to pretrain medical code representations in a hierarchical structure. We incorporate these pretrained representations into a graph neural network (GNN) to detect disease complications and design a multilevel attention method to compute the contributions of particular diseases and admissions, thus enhancing personalized interpretability. We present a new hierarchy-enhanced historical prediction proxy task in our self-supervised learning framework to fully utilize EHR data and exploit medical domain knowledge. We conduct a comprehensive set of experiments on widely used publicly available EHR datasets to verify the effectiveness of our model. Our results demonstrate the proposed model's strengths in both predictive tasks and interpretable abilities.

show abstract

Effective Identification of Similar Patients Through Sequential Matching over ICD Code Embedding

Cited by 13 publications

References 17 publications

Self-Supervised Graph Learning with Hyperbolic Embedding for Temporal Health Event Prediction

Self-Supervised Graph Learning with Hyperbolic Embedding for Temporal Health Event Prediction

Patient Similarity via Joint Embeddings of Medical Knowledge Graph and Medical Entity Descriptions

Self-Supervised Graph Learning With Hyperbolic Embedding for Temporal Health Event Prediction

Contact Info

Product

Resources

About