ACTRI Highlights Innovation at the 2024 Fall CTSA Program Annual Meeting
The UC San Diego Clinical and Translational Research Institute (ACTRI) proudly participated in the 2024 Fall CTSA Program Annual Meeting, held from November 13-15, 2024. This year’s theme, “Building CTSA Program Impact through Innovation, Collaboration, and Equity,” highlighted the transformative potential of cutting-edge research and partnerships in the translational science community.
The annual meeting served as a dynamic forum for CTSA Program Hubs to network, exchange knowledge, and share best practices. Attendees engaged in robust discussions on diverse topics, including Primary Care Practice-Based Research Networks (PBRNs), PCORnet, The Women’s Health Initiative and DEI, Translational Research to Translational Science, and Generative AI and Large Language Models in Translational Science.
Poster Presentation Spotlight: ACTRI’s Innovative AI Tool “Waldo”
Among the event's highlights was the Poster Presentation session, where hub-selected researchers showcased their innovative projects aligned with the meeting’s overarching theme. ACTRI was represented by Dr. John Ayers, Ph.D., who presented a poster on his groundbreaking project, Waldo: Automated Adverse Event Discovery from Text Data.
Co-authored by ACTRI’s own Dr. Mike Hogarth, M.D., and Dr. Davey Smith, M.D., the project demonstrates ACTRI’s commitment to leveraging cutting-edge technology for public health impact. The research explores the use of neural language processing to develop Waldo, a robust tool designed to detect adverse events (AEs) from unstructured text data.
About Waldo: Transforming Adverse Event Detection with AI
Objective
Waldo aims to streamline adverse event detection using artificial intelligence by analyzing vast amounts of unstructured data, such as user narratives and medical records.
Methods
Three models were tested:
-
N-gram model
-
BERT (Bidirectional Encoder Representations from Transformers)
-
RoBERTa (Robustly Optimized BERT)
After rigorous benchmarking, RoBERTa emerged as the most effective model, significantly outperforming other classifiers and even popular AI chatbots in accuracy and precision.
Results
Waldo’s capabilities were demonstrated by analyzing over 437,000 posts from 20 Reddit subreddits, uncovering 28,832 potential adverse events. For instance, users reported severe panic attacks after cannabis use and tinnitus linked to CBD products. These findings underscore Waldo’s ability to uncover critical safety signals, contributing to real-time pharmacovigilance.
Discussion and Impact
Waldo offers unprecedented efficiency and accuracy in AE detection, making it a vital tool for researchers and clinicians. The tool is open-sourced, inviting the broader health community to adopt it for enhanced safety monitoring.
ACTRI’s Ongoing Commitment to Innovation and Equity
Dr. Ayers’ presentation exemplifies ACTRI’s leadership in advancing translational science through innovation and collaboration. By developing tools like Waldo, ACTRI continues to address critical public health challenges while emphasizing equity in research and practice.
The 2024 Fall CTSA Program Annual Meeting reaffirmed the importance of collective efforts in transforming healthcare. ACTRI is proud to be at the forefront, driving impactful research and contributing to a more equitable and innovative future in translational science.
