CDMT-EHR: A Continuous-Time Diffusion Framework for Generating Mixed-Type Time-Series Electronic Health Records

2026-03-24

AI Summary

This work addresses the challenges of sharing electronic health records (EHRs) under strict privacy constraints and the difficulty of synthesizing their mixed-type (numerical and categorical) time-series data. The authors propose a continuous-time diffusion framework that unifies categorical variables into a Gaussian diffusion process via learnable continuous embeddings, models temporal dependencies using bidirectional gated recurrent units, and introduces an adaptive noise scheduling strategy based on a feature–timestep factorization. Notably, the method enables efficient classifier-free guided conditional generation with only 50 sampling steps—significantly fewer than the 1,000 steps required by baseline approaches. Evaluated on two intensive care datasets, the proposed model substantially outperforms existing methods in downstream task performance, distributional fidelity, and discriminability, while also demonstrating robustness in class-imbalanced scenarios.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Electronic health records (EHRs) are invaluable for clinical research, yet privacy concerns severely restrict data sharing. Synthetic data generation offers a promising solution, but EHRs present unique challenges: they contain both numerical and categorical features that evolve over time. While diffusion models have demonstrated strong performance in EHR synthesis, existing approaches predominantly rely on discrete-time formulations, which suffer from finite-step approximation errors and coupled training-sampling step counts. We propose a continuous-time diffusion framework for generating mixed-type time-series EHRs with three contributions: (1) continuous-time diffusion with a bidirectional gated recurrent unit backbone for capturing temporal dependencies, (2) unified Gaussian diffusion via learnable continuous embeddings for categorical variables, enabling joint cross-feature modeling, and (3) a factorized learnable noise schedule that adapts to per-feature-per-timestep learning difficulties. Experiments on two large-scale intensive care unit datasets demonstrate that our method outperforms existing approaches in downstream task performance, distribution fidelity, and discriminability, while requiring only 50 sampling steps compared to 1,000 for baseline methods. Classifier-free guidance further enables effective conditional generation for class-imbalanced clinical scenarios.

Problem

• Electronic Health Records
• Synthetic Data Generation
• Mixed-Type Time-Series
• Privacy-Preserving
• Diffusion Models

Innovation

• continuous-time diffusion
• mixed-type time-series
• electronic health records
• learnable noise schedule
• unified Gaussian diffusion