Conditioning on Time is All You Need for Synthetic Survival Data Generation

2024-05-27

arXiv.org

AI Summary

To address distributional distortion of event times due to right-censoring and poor statistical fidelity in synthetic survival data generation, this paper proposes a novel paradigm that conditions covariate synthesis on both the event time $t$ and censoring indicator $delta$, systematically introducing conditional generative modeling to survival data synthesis for the first time. Unlike conventional approaches, our method avoids explicit modeling of the unknown censoring mechanism and leverages existing conditional generative models (e.g., cGANs, cVAEs), achieving a favorable trade-off between statistical accuracy and computational efficiency. Extensive evaluation across multiple real-world survival datasets demonstrates that downstream survival models (Cox, DeepSurv) trained on synthetic data achieve an average 3.2% improvement in C-index on real test sets—significantly outperforming state-of-the-art baselines. Our core contribution is the establishment of an “event-structure-conditioned” generative framework, offering a principled, high-fidelity approach to synthetic survival data generation.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

Synthetic data generation holds considerable promise, offering avenues to enhance privacy, fairness, and data accessibility. Despite the availability of various methods for generating synthetic tabular data, challenges persist, particularly in specialized applications such as survival analysis. One significant obstacle in survival data generation is censoring, which manifests as not knowing the precise timing of observed (target) events for certain instances. Existing methods face difficulties in accurately reproducing the real distribution of event times for both observed (uncensored) events and censored events, i.e., the generated event-time distributions do not accurately match the underlying distributions of the real data. So motivated, we propose a simple paradigm to produce synthetic survival data by generating covariates conditioned on event times (and censoring indicators), thus allowing one to reuse existing conditional generative models for tabular data without significant computational overhead, and without making assumptions about the (usually unknown) generation mechanism underlying censoring. We evaluate this method via extensive experiments on real-world datasets. Our methodology outperforms multiple competitive baselines at generating survival data, while improving the performance of downstream survival models trained on it and tested on real data.

Problem

• Generating synthetic survival data with accurate event time distributions
• Addressing censoring challenges in survival analysis data generation
• Improving downstream survival model performance via synthetic data

Innovation

• Generates synthetic data conditioned on outcomes
• Leverages existing tabular data generation models
• Improves downstream survival model performance