Quantum Synthetic Data Generation for Structured Health Data

Thomas M. Menino Convention & Exhibition Center 

Synthetic data generation is a valuable tool for the analysis of health data with varied use cases, such as privacy protection and data supplementation for rare events. We present a pilot study evaluating the use of quantum-enhanced generative adversarial networks (GANs) to generate synthetic records based on structured fields in the FDA Adverse Event Reporting System (FAERS) to evaluate if these enhancements can provide deeper or equivalent insights for health data using smaller samples. The FAERS tracks adverse drug reactions, including rare reactions and reactions associated with rarely prescribed drugs. The model is implemented in a simulated quantum environment run on a classical computer using open-source tools. The presentation will conclude with an evaluation based on data fidelity and data utility in comparison to classical GAN models. Further, limitations of quantum simulations will be discussed.

pharmacovigilance

generative adversarial networks

FAERS