(P02-039-25) AI-Driven Surveillance of Foodborne Diseases via Social Media Reviews: The SAFEGUARD System Approach

Objectives: Foodborne diseases represent a critical challenge to public health, impacting millions globally every year. The objective of this study was to develop and evaluate SAFEGUARD (Social-media Analytics for Foodborne Events, Gpt-Based Understanding Analysis and Rapid Detection), an innovative syndromic surveillance system leveraging Artificial Intelligence (AI) to identify and monitor potential foodborne infection events through social media reviews.

Methods: SAFEGUARD workflow analyzes Google Reviews, which serve as a primary data source for identifying potential foodborne infection events. Data is retrieved using an API-based tool that enables extraction of restaurant reviews. The data is organized in a MongoDB database comprising two collections: 'Restaurants' for metadata and 'Reviews' for individual review content. Reviews are classified using GPT-4.0 Mini, a large language model integrated via the gpteasyr R library, which processes textual data based on structured prompts to identify indicators of foodborne illnesses. The workflow is automated for database synchronization and periodic updates. A web-based dashboard allows visual representations of flagged reviews for enhanced outbreak monitoring.

Results: The SAFEGUARD system successfully analyzed a dataset of Google Reviews, demonstrating a 93.9% accuracy in identifying reviews indicative of foodborne illnesses. Statistical analysis confirmed the system’s robustness in differentiating between general reviews and those specifically indicative of foodborne disease symptoms.

Conclusions: SAFEGUARD presents a novel approach to syndromic surveillance by utilizing AI and social media data to enhance the early detection of foodborne disease outbreaks. Its ability to provide timely and accurate insights could significantly improve public health responses to address foodborne disease challenges.

Funding Sources: This research was supported by Azione IV.5 – GREEN, Asse IV del PON Ricerca e Innovazione 2014-2020 “Istruzione e ricerca per il recupero – FSE REACT-EU”.