So something happened at the end of 2016 that isn’t discussed enough in QA circles. While many people in industry hem and haw about the data provided through reports at Iwaspoisoned.com, similar data has been available to consumers since December 2016. That month, FDA made all of the data from the CFSAN Adverse Event Reporting System (CAERS) (those reports logged whenever anyone calls FDA to say that a food made them sick) available for anyone to look at online. And they’re there forever.
Here’s the thing, I do not see this data used enough, and, the most recent download (data through Sept. 2017) includes more than 92,000 reports.
So let’s break down some interesting stats!
Note: per CFSAN, this data does not go through any independent validation by FDA (other than to remove duplicates), and in many cases had fields missing or obviously typographic errors. I did my best to clean the data so that it could be usefully sorted while not skewing the results based on bogus outliers.
Reports received by year (all data through Sept. 2017)
These reports are all voluntarily reported by consumers, public health authorities, or physicians. While foodborne illness has been steady or in decline for the last few years, consumers have also remained more vigilant than ever. Note that in the above graph, there are a ton of dietary supplement reports compared to other categories. While this is in part due to a boom in that industry, the general excess of data comes from the fact that manufacturers of dietary supplements are required to file reports when they receive information from customers regarding adverse events, while makers of foods and cosmetics are not required to do so. Let’s keep going, but I’m going to analyze only those adverse events reported for food products (minus dietary supplements) to limit our scope below.
What are the characteristics of the people affected in these reports?
Significantly more women are represented in adverse event reports than men. This has been a consistent trend year-over-year for all available years of data.
There’s a huge spike for events reported for infants and toddlers, presumably because these particular events resulted in doctor or hospital visits where physicians would file the reports. Additionally, these would be a much more susceptible population with additional attention paid by consumers, given that it’s someone’s child. Reports then slow down throughout the early 20’s, persons least likely to suffer extreme food-borne illness, then slightly increase as we get older before tapering off for those of advanced age. This reaffirms what we know, that sensitive populations are bi-modal, and include both the very young and the very old.
What types of foods are they reporting?
This is a particularly interesting piece of data, and one that deserves to be investigated further as we try to determine what sorts of foods consumers trust vs. what foods are making us sick. It’s interesting that nuts and seeds leads the pack at about 11% of all reports. This could correlate with outbreak data from the many multi-state outbreaks associated with nut butters over the years. But the thing is that these reports are not vetted by the FDA. The reports are purely a representation of what people call in and report what they think made them sick. This means that outbreak or recall associated foods may be over-represented.
Here’s where this data gets really fun.
There was a very public and industry-changing listeria outbreak in 2015 regarding a certain national ice cream brand. Listeria was found in the plant and product and WGS traced it to 10 illnesses that had occurred over the previous 5 years. So I was curious…what effect did this have on people reporting adverse events from this particular ice cream?
Well, that’s somewhat damming. While confirmed illnesses were actually happening, adverse events were coming into CFSAN at approximately the same rate. Not that they necessarily should have, there would presumably have been many more illnesses that were either not cultured for WGS or didn’t result in a doctor’s visit. What is extremely interesting however, is that even after the outbreak was discovered and recalls were issued, no additional cases were ever confirmed by FDA/CDC.
However, while the illnesses seemed to stop, the number of people who reported the affected foods made them sick increased by nearly 7-fold. We can see a nice taper from the time the outbreak and recalls were announced back down to the baseline number of reports submitted before the outbreak. This particular case also tells us something, as this particular supplier had a second set of recalls in September 2016, however it didn’t cause the same huge uptick that the original did. This suggests that the extensive media coverage from a confirmed outbreak had much more of an impact on consumers than a run-of-the-mill recall did.
The implications here are clear, if your business is associated with an outbreak, people are going to remember it and report your product for their illnesses, regardless of whether they’re ever actually connected. This would be compounded if you had a series of events. On the other hand, this is good news for recalls, as it demonstrates that consumers aren’t necessarily jumping the gun because of a recall like they would following an outbreak.
I’m interested in digging into this data further to find other consumer trends, and see if the above observation holds true for other product categories. What sort of questions do you think we should ask of this data, or what other data-sets should we be comparing it to?