Westat Presents Cutting-Edge Research at JMM 2026

Westat researchers will share their work at the 2026 Joint Mathematics Meetings (JMM), held in Washington, DC, from January 4 to 7. Now in its 4th year under a reimagined format, JMM convenes more than 17 mathematics organizations for the world’s largest annual mathematics gathering—a vibrant hub for collaboration, discovery, and innovation. Register now for this event.

Westat will contribute to the U.S. Food and Drug Administration’s (FDA) special session, “Mathematical Applications in Medical Product and Food Research, Development, and Regulation for the Advancement of Public Health,” on Tuesday, January 6. Westat’s Fatih Selekler, a Vice President of Technology and Digital Services, is a co-organizer of this session.

The discussion will showcase how applied mathematics fuels innovation and strengthens public health, connecting research and development in medical products and food with the principles of regulatory science. Presenting authors (indicated with ^*) will discuss the use of quantitative approaches to enhance product development, improve regulatory review, and guide evidence-based decisions that safeguard public health.

Tuesday, January 6

3:00 pm

AI-Driven Synthetic Data Generation With Complex Longitudinal Health Data

Minsun Riddles^*

This talk explores how synthetic data generated with artificial intelligence (AI) can expand access to complex health datasets while protecting privacy. Using generative pre-trained transformers (GPT), realistic patient trajectories can be simulated to facilitate collaboration and innovation without exposing sensitive information.

3:30 pm

Fine-Tuning Large Language Models for Adverse Drug Event Detection

Sean Chickery,^* John Riddles, Rashi Saluja, Jeremy Corry, Julianna Lee, Marcelo Simas, Kevin Wilson, Gizem Korkmaz

Adverse drug events threaten patient safety, requiring enhanced surveillance methods. Using the FDA’s Adverse Event Reporting System (FAERS) Q2 2024 data, Westat tested whether embeddings from fine-tuned large language models (LLMs) could improve the classification of serious outcomes. After combining LLM embeddings with structured features, Westat found that classifiers achieved an accuracy of up to 93% and an area under the curve of 0.98, demonstrating strong potential for regulatory case prioritization.

4:00 pm

Quantum Enhancements to Synthetic Data Generation for Structured Health Data

John Riddles,^* Matthew Ring, Sean Chickery, Eric Pan, Kevin Wilson, Gizem Korkmaz, Fatih Selekler

This study tests whether quantum computing can improve the generation of synthetic health data. Using FAERS as a case study, Westat will compare a GPT-based model with a simulated quantum-enhanced version, evaluating statistical accuracy and potential computational gains.