2024 Health Meeting - Session 10A: Creating Synthetic Healthcare Data with Generative AI Methods 6.26.24 Recording

Generative AI has become a hot topic since ChatGPT became one of the most well-known Machine Learning models ever. There are many different generative models out there, and in this presentation, we explore the possibilities, benefits, reservations, and alternative use-cases of using Generative Adversarial Networks to generate health insurance claims data. We offer a proof of concept, and discuss future growth possibilities when leveraging generated data over strictly de-identified data. In using machine learning approaches to modeling the real world, the more data on hand, one would typically get a better understanding of the underlying mechanisms. However, due to high costs of collecting data in some cases or sensitive or personal information in some others, de-identified synthetic data that closely resemble real data may offer valuable insights to various topics in focus. In this session we are going to share our findings and some results exploring various trending generative AI tools on some of the available health research datasets.