An evaluation of the relationship between public communications on food recalls and online self-reporting of foodborne illness

The US Food and Drug Administration routinely announces product recalls to inform the public about mislabeled or potentially hazardous products. For example, food recalls may occur when a product is contaminated with a foodborne pathogen (e.g., Listeria monocytogenes) or contains an undeclared allergen (e.g., nuts). In addition to governmental systems that capture consumer or industry self-reported problems with food at the federal, state and county level, there are platforms such as https://iwaspoisoned.com (IWP) through which the public can self-report when they are sick, share their symptoms, and note what they think made them sick. Here we use the Granger Causality (GC) test to investigate two information streams - IWP and FDA recall announcements - to determine whether there is a statistically supported relationship between them and if so the extent of that relationship (i.e., to understand if self-reporting through IWP precedes recalls or vice versa). Among 48 foods selected due to their high number of occurrences in both IWP and FDA recalls, whether IWP was predictive of bacterial recalls had the highest number of positive GC tests with 15 foods (31%). There were 12 foods that had positive GC tests for both bacterial recalls being predictive of increased IWP reports and allergen recalls being predictive of increased IWP reports. The lack of a stronger relationship between the information streams does not detract from the importance and utility of self-reporting platforms and communications by public health agencies. Rather they illustrate that based on the method employed here that the influence on public behavior and the degree to which self-reporting can forecast future recalls is difficult to detect. However, with further developments in the modeling of the relationship, data capture approaches, and public communication strategies the relationship will likely be improved.