Factors Influencing Foodborne Illness Self-Identification From User-Generated Data – Minnesota, 2024

Abstract

Foodborne illness outbreaks are a serious public health concern; faster identification enables the implementation of control measures to prevent others from becoming ill. User-generated data and artificial intelligence can be used to develop outbreak signals that could be disclosed to the public before they are associated with an actual outbreak of foodborne illness. We employed a vignette-based discrete choice experiment survey to examine when and why individuals would self-identify as ill during a publicized foodborne illness outbreak. Scenarios presented to participants included four attributes describing the event which were publicized in the headlines: the number of people (i.e., 20, 200, 8,500) stated as ill; symptoms (i.e., nausea, vomiting, and diarrhea) included or excluded; a statement that FDA is investigating included or excluded; and a call to action asking readers to report their symptoms included or excluded. The research found that people will self-identify from a publicized foodborne illness event, with a positive association for all attributes. The odds of self-identifying as ill from a publicized foodborne illness event more than doubled when the number of people publicized as ill is 8,500 (OR = 2.42, CI [2.16, 2.71], p < 0.001) or symptoms (OR = 2.21, CI [2.02, 2.42], p < 0.001) are included. This study highlights factors that influence a person to self-identify as ill from a publicized foodborne illness event, regardless of whether an actual outbreak exists, demonstrating a limitation of novel data streams in detecting foodborne illness outbreaks in the absence of public health authority confirmation.