Exploring the Relationship Between Foodborne Outbreaks and Climate Change Using Social Media Data in the Years 2016-2021.

Evidence suggests that weather affects the incidence of enteric infections, but most foodborne-illness research focuses on short-lived extreme events (e.g., heatwaves and floods) rather than gradual climate variability. A second, related limitation is surveillance under-ascertainment: many mild-to-moderate cases are rarely reported, which constrains our ability to quantify climate-sensitive changes in disease burden. This study addresses these gaps by testing whether symptom-related social media messages can serve as a complementary signal for foodborne illness and, when combined with climate anomalies, improve the characterization of climate-disease relationships. We analyzed 2016-2021 Twitter (X) messages from Australia and New Zealand. Messages were retrieved using symptom and food-exposure keywords, cleaned with Natural Language Processing, and screened for foodborne relevance using symptom-ingestion term co-occurrence. We derived spatial clusters of messages using Gaussian Mixture Models and linked monthly cluster-level message frequencies to state/region temperature and rainfall anomalies and to notified salmonellosis and campylobacteriosis counts (Australia, 2020-2021) using generalized additive models. Climate anomalies alone explained little variation in notified cases (R2 = 0.01-0.12), while message frequency alone provided modest explanatory power (R2 = 0.22-0.48). Models combining message frequency with climate anomalies performed substantially better (R2 = 0.87-0.97), indicating that social media-derived indicators can strengthen climate-sensitive foodborne illness monitoring. Limitations include platform and user representation biases, and a modeling evaluation restricted to specific pathogens, places and years. Therefore, generalizability should be tested using additional platforms and independent datasets.