Quick Overview

The UK Health Security Agency (UKHSA) is investigating the potential of artificial intelligence (AI) to analyze online restaurant reviews for identifying sources of food poisoning.

Key Insights

• UKHSA researchers evaluated various large language models (LLMs) to determine their effectiveness in detecting symptoms of gastrointestinal (GI) illnesses from numerous online restaurant reviews.
• Symptoms such as abdominal pain, diarrhea, and vomiting were assessed, along with the types of food consumed by patrons, aiming to enhance the investigation of foodborne illness outbreaks.
• This approach could yield additional data on GI illness rates that current systems do not capture, aiding in pinpointing potential sources of food poisoning.

Expert Commentary

Steven Riley, Chief Data Officer at UKHSA, stated, “We are continuously seeking innovative methods to improve our disease surveillance. Utilizing AI in this manner could assist us in identifying the probable sources of foodborne illness outbreaks, complementing traditional epidemiological techniques to reduce illness rates.”

Study Details

• The initiative is part of UKHSA’s broader evaluation of AI applications in public health.
• In this recent study, researchers analyzed a more extensive list of terms and language to better identify sources of illness.
• Over 3,000 reviews were manually annotated by epidemiologists, focusing on those containing specific GI-related keywords.
• General symptoms like headaches and respiratory issues were excluded as they are not specific to GI illnesses.

Challenges Identified

• Access to real-time data remains a significant challenge for researchers.
• While LLMs can provide general information about the food consumed, identifying specific ingredients linked to illnesses is complex.
• Issues such as spelling variations, slang usage, and misattribution of illness to specific meals were noted as obstacles.

Future Directions

Riley emphasized that further research is necessary before integrating these AI methods into routine practices for addressing foodborne illness outbreaks.

Related Developments

In March 2025, NHS England announced plans to implement AI software by Cera to detect symptoms of winter illnesses, including Covid, flu, RSV, and norovirus.

Sources

• UKHSA using AI to identify sources of food poisoning
• Bad restaurant reviews to be used to track food poisoning outbreaks
• AI could help detect and investigate foodborne illness outbreaks
• UKHSA Explores AI Surveillance of Online Reviews for Foodborne Illness Outbreak Investigations
• UKHSA investigates use of AI in outbreak detection