This paper surveys deep learning applications in EHR analysis.
problem Leveraging EHR data for clinical informatics tasks.
method Reviews deep learning architectures and techniques applied to EHR data.
result Identifies current limitations and future research directions.
ME2Vec learns medical entity vectors from EHR data.
problem Learning structured information from EHR data.
method Graph-based medical entity embedding using diverse graph embedding techniques.
result ME2Vec outperforms baselines in disease diagnosis prediction.
Bayesian model predicts patient survival from sparse EHR data.
problem Analyzing EHR data with few samples and diverse information.
method Nonparametric probabilistic model using Bayesian trees.
result Improved survival trajectory predictions on patient data.
EKG-based models show better stability across patient populations than EHR-based models.
problem Model generalization issues in EHR and EKG-based predictive models.
method Two tests to measure model generalization, comparing EHR and EKG data.
result EKG-based models are more stable across different patient populations.
mixEHR model reveals latent disease topics from EHR data.
problem Mining heterogeneous, non-randomly missing EHR data.
method Generative model integrating collaborative filtering and latent topic models.
result mixEHR reveals meaningful multi-disease insights and outperforms previous methods.
Transformer learns hidden structure of EHR data for better prediction.
problem Lack of complete structure information in EHR data.
method Graph Convolutional Transformer using data statistics to learn structure.
result Consistently outperforms previous approaches on various prediction tasks.
Neural TPPs improve EHR modelling efficiency.
problem Challenges in modelling irregularly timed, noisy EHRs.
method Proposed neural network parameterizations of Temporal Point Processes (TPPs) for EHRs.
result Neural TPPs outperform non-TPP models on EHRs.
New method generates diverse EHR data types while maintaining privacy.
problem Privacy risks in sharing EHRs across large scales.
method Refined GAN model, feature constraints, and utility measures.
result New model preserves EHR statistics, correlations, and constraints.
Proposes a new model for EHR data using time-dependent Gaussian processes.
problem Joint modeling of multiple clinical variables over time.
method Multivariate nonstationary Gaussian processes with time-varying parameters and posterior inference via HMC.
result The proposed model outperforms stationary models and reveals latent correlations predictive of patient risk.
TAPER learns unified patient EHR representations for healthcare tasks.
problem Irregular and multimodal data in electronic health records.
method Transformer networks and BERT for embedding structured and unstructured data.
result TAPER model outperforms on mortality, readmission, and length of stay tasks.
This review explores methods for detecting Adverse Drug Events in EHRs.
problem Detecting Adverse Drug Events (ADEs) in EHRs using temporal data.
method Examines 5 main approaches: temporal abstraction, graph-based, learning weights, and time series data.
result Challenges remain in exploiting heterogeneous temporal data for ADE detection.
Generative adversarial networks enhance risk prediction in EHRs.
problem Limited labeled data in EHRs for deep learning risk prediction.
method ehrGAN, a modified GAN, generates realistic EHR data for semi-supervised learning.
result Significant improvements in classification tasks with generated data over baselines.
Learn embeddings from EHRs to predict ICD codes.
problem Predicting ICD codes from patient visits in EHRs.
method Deep neural network trained to predict ICD codes, capturing clinical information.
result Embeddings capture relevant clinical information and can be used in machine learning models.
Generates synthetic EHR chief complaints from structured data.
problem Generating unstructured text in EHRs.
method Encoder-decoder deep learning model trained on EHR data.
result Synthetic chief complaints preserve epidemiological information.
Paper proposes a framework for disease prediction from EHRs with missing data.
problem Missing data in EHRs for disease prediction.
method Two-stage framework including missing data imputation and disease prediction using GANs and stacked autoencoders.
result Significantly improved disease prediction accuracy with AC-GAN and stacked autoencoder.
AdaptHetero uses MLI to tailor EHR models for subgroup-specific predictions.
problem Lack of subgroup-specific, operationalizable modeling strategies in EHRs.
method Integrates MLI with unsupervised clustering to identify subgroup-specific characteristics.
result Improves predictive performance by up to 174.39 percent across many subpopulations.
Patient2Vec learns personalized deep EHR representations for better health outcomes prediction.
problem Complexities in EHR data hinder machine learning applications.
method Patient2Vec: A personalized deep learning framework for longitudinal EHR data.
result Patient2Vec achieves AUC of 0.799 in predicting future hospitalizations, outperforming baseline methods.
MiME learns EHR data structure for predictive healthcare tasks.
problem Data insufficiency in EHR for predictive healthcare tasks.
method Leverages multilevel structure of EHR data and learns multilevel embedding.
result MiME outperforms baseline methods in diverse evaluation settings.
Language models improve clinical prediction models using EHR data.
problem Limited patient data for training clinical prediction models.
method Using patient representation schemes from natural language processing.
result 3.5% mean improvement in AUROC on five prediction tasks.
Develops algorithm to find subgroups with different treatment effects in HIV patients.
problem Estimating treatment effects in EHR data with challenges like time-varying confounding.
method SDLD algorithm combining generalized interaction tree and longitudinal targeted maximum likelihood estimation.
result Identifies subgroups of HIV patients at higher risk of weight gain with dolutegravir-containing ARTs.
Representation learning improves EHR data for healthcare tasks.
problem Transforming EHR data into useful representations for machine learning.
method Deep learning and disentangling underlying factors from EHR data.
result Better representations improve machine learning performance in healthcare.
Paper proposes multimodal contrastive learning for EHR data.
problem Separate treatment of structured and unstructured EHR data.
method Proposes a multimodal feature embedding generative model and a multimodal contrastive loss.
result Multimodal learning yields better feature representation than single-modality learning.
Framework harmonizes EHR data across institutions for better analysis.
problem Heterogeneity of medical codes and terminologies hinder EHR data analysis.
method MASH (Multi-source Automated Structured Hierarchy) uses neural optimal transport and learned hyperbolic embeddings to align and structure EHR data.
result MASH generates interpretable hierarchical graphs for unstructured local laboratory codes.
AI system predicts acute critical illness from EHRs with explainability.
problem Lack of clinical interpretability in AI predictions for acute critical illness.
method Developed an explainable AI early warning score (xAI-EWS) system.
result System provides clinicians with insights into EHR data explaining predictions.
PEHRT harmonizes EHR data for translational research.
problem Barriers in using EHR data for translational research.
method Common pipeline including open-source code, visualization tools, and detailed documentation.
result PEHRT harmonizes EHR data to standardized ontologies and generates robust embeddings.
AI predicts medical specialty diagnostic choices from EHR records.
problem Predicting timely medical specialty diagnostic workups for patients.
method Ensemble of feed-forward neural networks trained on EHR data.
result Significantly higher accuracy compared to traditional checklists.
VBphenoR uses variational Bayes for EHR-based patient phenotyping.
problem Phenotyping patients from EHR data for targeted treatments.
method Variational Bayes Gaussian Mixture Model (GMM) and logistic regression.
result Closed-form inference for efficient patient phenotype determination.
This study evaluates a health knowledge graph for robustness in EHRs.
problem Evaluate robustness of a health knowledge graph in EHRs.
method Analyze sample size, unmeasured confounders, and non-linear functions.
result Identify sample size and unmeasured confounders as major sources of error.
Study proposes a model to improve patient subtyping from EHR data.
problem Challenges in subtyping temporal EHR datasets.
method Self-supervised Mamba-based model for learning EHR representations.
result Model outperforms baseline models in EHR data subtyping.
Proposes a method to derive knowledge graphs from EHR data.
problem Challenges in deriving generalizable knowledge from EHR data.
method Infer conditional dependency structure via a latent graphical block model (LGBM).
result Perfect recovery of block structure demonstrated.
Predicts EHR features that improve prediction, learning coherent topics.
problem Balancing prediction quality and coherence in EHR topic models.
method Prediction-focused topic model that uses supervisory signal to retain relevant features.
result Prediction-focused topic models learn more coherent topics while maintaining competitive predictions.
Medical deconfounder uses EHRs to estimate treatment effects without confounders.
problem Bias in assessing treatment effects from EHRs due to unobserved confounders.
method Develops a machine learning algorithm (medical deconfounder) to adjust for confounders.
result Medical deconfounder produces more accurate treatment effect estimates and identifies effective medications.
Machine learning faces challenges in healthcare data, but offers opportunities.
problem Poorly labeled data, multiple endotypes, and underrepresented healthy individuals.
method Review of existing machine learning methods and challenges in healthcare.
result Opportunities for machine learning in healthcare identified.
Novel imputation method for EHRs with structured and sporadic missingness.
problem Missing data in integrated EHR datasets for clinical applications.
method Macomss, a novel imputation framework for structurally and heterogeneously missing data.
result Macomss outperforms existing methods in imputation and downstream prediction accuracy.
This work generates synthetic EHRs with privacy guarantees for machine learning tasks.
problem Privacy concerns and heterogeneity in EHR data limit their use in machine learning.
method Generative Adversarial Networks (GANs) with differential privacy (DP) for synthetic data generation.
result Synthetic EHRs maintain performance close to real data, even with DP applied.
ODVICE augments EHR cohorts using ontology to improve analysis robustness.
problem Limited records in cohorts for rare diseases hamper robust analysis.
method Ontology-driven Monte-Carlo graph spanning algorithm for data augmentation.
result ODVICE augmented cohorts show ~30% improvement in AUC over non-augmented datasets.
Proposes statistical inference for dependency knowledge graphs from EHR data.
problem Statistical uncertainty in linking entities in EHR data.
method Dynamic log-linear topic model with singular value decomposition.
result Established asymptotic normality for sparse graph edge recovery.
Paper proposes a graph network for EHR data that learns robust representations.
problem Learning robust representations for EHR data with implicit connections.
method Variationally regularized encoder-decoder graph network.
result Model outperforms existing methods in various EHR predictive tasks.
Paper proposes GANs for predicting drug-induced lab test trajectories in EHRs.
problem Predicting drug-induced lab test trajectories in EHRs.
method Proposes a framework for GANs in healthcare, measures predictive power of synthetic data, and incorporates representation learning.
result Incorporating representation learning improves GANs' predictive power for drug-induced lab test trajectories.
Multitask learning improves phenotyping in EHR data, but its benefits vary by phenotype complexity.
problem Improving phenotyping accuracy in EHR data using multitask learning.
method Investigated multitask learning for phenotyping rare and common phenotypes in EHR data using neural nets and logistic regression.
result Multitask learning with neural nets consistently outperforms single-task neural nets for rare phenotypes but underperforms for common phenotypes.
The study uses demographical data to predict health conditions.
problem Predicting health conditions based on patient demographics and symptoms.
method Analyzing Electronic Health Records (EHR) from Brazil to identify age-related clusters of health conditions.
result Age of patients significantly influences the likelihood of certain health conditions.
Bayesian deep learning predicts uncertainty in EHRs for better healthcare decisions.
problem Lack of interpretability and trustworthiness in deep learning models for EHRs.
method Proposes a Bayesian Neural Network (BNN) model to predict uncertainty in EHRs.
result High uncertainty instances harm model performance; distributions reveal patients for timely intervention.
EHR-MPC optimizes sepsis treatment using digital twins and inference-time control.
problem Optimal sepsis treatment policies are contested and difficult to adapt during inference.
method EHR-MPC decouples learning patient dynamics from treatment optimization, enabling inference-time control over learned digital twins.
result EHR-MPC achieves comparable off-policy performance and improved simulation performance compared to RL baselines.
Neural network predicts cardiovascular events from EHRs with high accuracy.
problem Predicting onset of cardiovascular diseases from electronic health records.
method Multi-task gated recurrent units with attention mechanism.
result Model outperforms clinical risk scores in predicting stroke and myocardial infarction.
Model learns hierarchical EHR representation for clinical outcome prediction.
problem Capturing temporal patterns in irregular clinical event sequences.
method Proposes differentiated mechanisms to model events at different time scales, learning hierarchical representations.
result Significantly improves clinical outcome prediction, achieving AUC scores of 0.94 and 0.90 for death and ICU admission respectively.
Paper improves communication in decentralized federated learning for EHRs.
problem Efficiently learn from large, decentralized EHR databases.
method Fully decentralized federated learning with iterative local updates and reduced communication rounds.
result Significant reduction in communication rounds without compromising solution optimality.
Convolutional residual model predicts diagnoses from EHR notes.
problem Predicting multiple diagnoses from medical text data.
method Convolutional Neural Network (CNN) + Deep Residual Network.
result Superior performance compared to baseline models.
Deep learning model creates patient representations for scalable EHR-based stratification.
problem Challenges in summarizing and representing patient data from EHRs prevent scalable stratification analysis.
method Unsupervised framework based on deep learning (ConvAE) using word embeddings, CNNs, and autoencoders.
result ConvAE significantly outperformed baselines in clustering diverse patient cohorts, identifying clinically relevant subtypes.