413P AI-driven privacy-preserving synthetic data generation for mortality prediction
Talha Qaiser, Md Atiqur Rahman, Stanislav O. Zakharkin · ESMO Real World Data and Digital Oncology · 2025
Data privacy regulations restrict access to clinical trials and Real-World data, limiting their use in healthcare applications. Synthetic data replicating statistical properties of real datasets enables privacy-preserving analyses. We used the public MIMIC-IV ICU dataset to generate synthetic data with statistical and AI methods, assess the risk of patient re-identification, and evaluate utility of machine learning algorithms for mortality prediction.