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

TL;DR

This paper introduces a continuous-time diffusion model for generating realistic, mixed-type electronic health record time-series data, improving over discrete models by reducing approximation errors and enabling efficient, conditional data synthesis.

Contribution

The authors propose a novel continuous-time diffusion framework with a bidirectional RNN backbone, unified Gaussian diffusion for categorical variables, and a learnable noise schedule, advancing EHR synthetic data generation.

Findings

Outperforms existing methods in downstream tasks

Achieves higher distribution fidelity and discriminability

Requires only 50 sampling steps versus 1,000 for baselines

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.