SurvDiff: A Diffusion Model for Generating Synthetic Data in Survival Analysis

Survival analysis is a cornerstone of clinical research by modeling time-to-event outcomes such as metastasis, disease relapse, or patient death. Unlike standard tabular data, survival data often come with incomplete event information due to dropout, or loss to follow-up. This poses unique challenges for synthetic data generation, where it is crucial for clinical research to faithfully reproduce both the event-time distribution and the censoring mechanism. In this paper, we propose SurvDiff, an end-to-end diffusion model specifically designed for generating synthetic data in survival analysis. SurvDiff is tailored to capture the data-generating mechanism by jointly generating mixed-type covariates, event times, and right-censoring, guided by a survival-tailored loss function. The loss encodes the time-to-event structure and directly optimizes for downstream survival tasks, which ensures that SurvDiff (i) reproduces realistic event-time distributions and (ii) preserves the censoring mechanism. Across multiple datasets, we show that SurvDiff outperforms state-of-the-art generative baselines in both distributional fidelity and survival model evaluation metrics across multiple medical datasets. To the best of our knowledge, SurvDiff is the first end-to-end diffusion model explicitly designed for generating synthetic survival data.

Doctors and researchers often study how long it takes until an important medical event happens, such as cancer relapse, disease progression, or death. We wondered whether computers could generate realistic versions of this kind of data, so that survival analysis methods can be tested and developed more reliably. This is hard because survival data is different from ordinary tables. For some patients, the event is observed; for others, the study ends or the patient leaves before the event happens. A good synthetic dataset therefore has to capture not only patient characteristics, but also event times and whether the event was actually observed. We developed SurvDiff, a computer model that learns from real survival datasets and generates new synthetic ones. Unlike general-purpose data generators, SurvDiff is built specifically for survival data. Across several medical datasets, it produced synthetic data that was useful for training and evaluating survival models, while also closely resembling the original data in important ways.