TarDiff: Target-Oriented Diffusion Guidance for Synthetic Electronic Health Record Time Series Generation

2025-04-24

AI Summary

Existing EHR time-series synthesis methods primarily focus on statistical distribution matching, yielding limited improvements in downstream clinical model performance—particularly for critical scenarios such as rare diseases. This paper proposes a task-oriented diffusion-based generative framework that innovatively incorporates influence functions into the diffusion reverse process, using task loss reduction as the optimization signal—enabling utility-driven (rather than distribution-driven) synthesis. The method integrates EHR-specific temporal modeling, influence function-based gradient estimation, and task-loss-sensitive guidance. Evaluated on six public EHR datasets, it improves downstream models’ AUPRC by up to 20.4% and AUROC by up to 18.4%, significantly alleviating data scarcity and severe class imbalance. The synthesized data demonstrate high clinical utility for predictive modeling.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Time-Series/Data Streams
• Humans and AI: Other Foundations of Human Computation & AI

Application Category

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

Abstract

Synthetic Electronic Health Record (EHR) time-series generation is crucial for advancing clinical machine learning models, as it helps address data scarcity by providing more training data. However, most existing approaches focus primarily on replicating statistical distributions and temporal dependencies of real-world data. We argue that fidelity to observed data alone does not guarantee better model performance, as common patterns may dominate, limiting the representation of rare but important conditions. This highlights the need for generate synthetic samples to improve performance of specific clinical models to fulfill their target outcomes. To address this, we propose TarDiff, a novel target-oriented diffusion framework that integrates task-specific influence guidance into the synthetic data generation process. Unlike conventional approaches that mimic training data distributions, TarDiff optimizes synthetic samples by quantifying their expected contribution to improving downstream model performance through influence functions. Specifically, we measure the reduction in task-specific loss induced by synthetic samples and embed this influence gradient into the reverse diffusion process, thereby steering the generation towards utility-optimized data. Evaluated on six publicly available EHR datasets, TarDiff achieves state-of-the-art performance, outperforming existing methods by up to 20.4% in AUPRC and 18.4% in AUROC. Our results demonstrate that TarDiff not only preserves temporal fidelity but also enhances downstream model performance, offering a robust solution to data scarcity and class imbalance in healthcare analytics.

Problem

• Generates synthetic EHR time-series to address data scarcity
• Enhances model performance by targeting rare clinical conditions
• Optimizes synthetic data utility via task-specific influence guidance

Innovation

• Target-oriented diffusion framework for EHR
• Influence functions guide synthetic data generation
• Optimizes utility via task-specific loss reduction