TORONTO – Using crowdsourced data for foodborne outbreak investigations can work but poses a number of challenges, according to a session at the International Association for Food Protection (IAFP) conference.
Jennifer Beal, from the U.S. Food and Drug Administration and Laura Gieraltowski, from the U.S. Centers for Disease Control and Prevention gave the U.S. perspective while Anna Manore, of the Public Health Agency of Canada spoke about the practice in Canada.
Social networking websites like Twitter, Facebook, and Reddit focused on foodborne illness reporting such as iwaspoisoned.com. Some forums including iwaspoisoned.com and Yelp, allow people to share details of symptoms, illnesses, or experiences with food products, companies, or brands and retail or dining locations.
However, because these platforms are open, posts vary widely in format and content. These inconsistencies can make it difficult to determine the accuracy of information and to interpret the data. Anonymous posts can also make further investigation with consumers challenging.
Crowdsourced data use in outbreak investigations
Public health professionals and food regulators are evaluating these tools to determine their usefulness in outbreak investigations.
Setting the scene, Ben Chapman, from North Carolina State University, said crowdsourcing could be used as part of early warning systems and contribute to real-time monitoring of food safety hazards.
Benefits include increased or enhanced data collection, community engagement, a deeper view on certain situations, and a cost-effective approach to data collection.
Chapman said ProMED-mail was likely the first example of crowdsourcing in the field of public health while posts on Twitter and iwaspoisoned.com have played roles in identifying outbreaks.
However, he warned about “armchair epidemiology” and separating out “noise” from useable information. Other negatives include data quality, privacy concerns, participant biases, bots on social media and misinformation skewing data.
Chapman added it was important to look at as many data sources as possible, depending on the question that needs to be answered. Crowdsourced data is a tool to signal things that might not have been caught before and then resources can be allocated to see if it is a real problem.
General Mills cereal example
Beal and Gieraltowski talked through the challenges via four stages including the case definition and identification of a cluster, confirming an outbreak vehicle, implementing control measures and communicating unknowns, using an example of people reporting illness on iwaspoisoned.com after eating Lucky Charms cereal in 2022.
Beal said crowdsourced data turns the normal process upside down as CDC traditionally leads epidemiological investigations but it is the FDA that heads-up the investigation for crowdsourced or non-traditional data types.
One issue is epidemiological data is not standardized as it is provided by complainants, there are also questions around identifying the vehicle, the agent involved and contact with the company.
Beal said FDA was unsure what to tell General Mills, which hampered the firm’s ability to find out what was happening, as there was limited lot code information and variability in symptoms made it hard to know what testing to do.
She added the public scrutiny that often accompanies such incidents adds urgency but people tend to think it was the last thing they ate that made them sick. Also, if they saw others saying Lucky Charms made them sick, they could assume it was the cereal that also made them ill.
Manore presented the use of online surveys in two outbreak investigations.
The first was Salmonella Newport in 2018 and confirmed what Foodbook had told health officials, that the incident was linked to red onions. Foodbook is a survey that was done to describe what foods Canadians eat over a seven-day period to inform outbreak investigation and response.
The second was an outbreak of Salmonella Enteritidis in 2019. Two foods were of interest: frozen fish and mixed fruit cups but they were not asked about in Foodbook. However, it was later found the epidemic was caused by contaminated imported frozen profiteroles.
She said this shows online surveys require careful consideration and their use must be considered alongside other available evidence from epidemiological, food safety and traceback and lab investigations.
(To sign up for a free subscription to Food Safety News, click here)
