Balancing the privacy-utility tradeoff for synthetic time-to-event data

Generation of synthetic patient records can preserve the structure and statistical properties of the original data without violating privacy, providing access to high-quality data for research and innovation. Few synthetization methods account for the censoring mechanism in time-to-event data, and formal privacy risk evaluations are often lacking. Improvements in synthetic data utility come with increased risks of privacy disclosure, necessitating a careful evaluation to obtai n the proper balance. In this talk, I will demonstrate a method for generating synthetic time-to-event data based on regression models and a flexible parametric survival model in Stata. I show how to evaluate the synthetic data utility and present a method for estimating the privacy loss from publishing a synthetic dataset.