2024-7-20
I wanted to make poke bowls at home and took a random walk down the Internet Street. Before long, I dropped by this post (2023) which mentioned weirdly round numbers about how like do eggs carry Salmonella:
A study testing 100000 eggs from Japan found a rate of salmonella presence at 0.003%. [...] in the US it is 1 out of 20,000 compared to 1 out of 33,333 for Japan.
and of course in the comments that followed, this propagated to
If you ate one raw egg every single day, it would take 55 years to get salmonella in America and 91 years in Japan.
Since I am not in Japan, I am only interested in verifying the US's 1/20,000 ratio to find out if I want to put egg yolks in to my marinates.
Googling USDA fresh egg salmonella rates,
google's featured snippets will decide to put
USDA's "Raw Liquid Eggs Survey" (2012-2013)
at the top and to highlight the following sentence from the report:
The estimated prevalence of Salmonella in egg white is 6.99% with a 95% confidence interval between 2.44% and 11.54%.
which establishes a 1400 fold difference from the reddit post's claim.
It is right there in the title though, that this report is about
the processed product,
not the fresh eggs. (One has to use quotes, i.e. "fresh egg" in the query,
or ad blocker rules to avoid seeing this uselessly promoted search result.)
USDA does not seem to have a solid, or even a wobbly number, for the Salmonella risks in fresh eggs. "Shell Eggs from Farm to Table" simply advices against consuming raw egg white or yolk altogether. Although interestingly, it also mentions that
shell eggs can be pasteurized in order to destroy Salmonella,
if raw shell eggs are not treated to destroy Salmonella, the packages must carry a certain safe handling statement.
The promisingly-titled "Salmonella By the Numbers" is about inspection programs.
On the other hand, the 1/20,000 can be found all over the place
on the internet. For example, googling "1 out of 20,000 egg salmonella"
leads one to this page by the University of Minnesota Extension who vouched the following,
without giving a reference,
The Centers for Disease Control estimates that 1 in every 20,000 eggs are contaminated with Salmonella.
SELF Magazine (self.com) too mentioned without a reference that,
In 2000, researchers from the USDA estimated that one in every 20,000 eggs may contain salmonella.
Both were actually referring to one same paper, although the University of Minnesota Extension's post probably got the authors' affiliation wrong. The research paper is titled "Estimating the annual fraction of eggs contaminated with Salmonella enteritidis in the United States", by Eric Ebel and Wayne Schlosser, and was published in International Journal of Food Microbiology, Volume 61, Issue 1, 1 October 2000, Pages 51-62. The numbers given were:
The expected value of this distribution is approximately one SE-affected egg in every 20,000 eggs annually produced, and the 90% certainty interval is between one SE-contaminated egg in 30,000 eggs, and one SE-contaminated egg in 12,000 eggs.
The paper's affiliation section only has "USDA, Food Safety and Inspection Service, Washington, DC, USA" in it. The journal is under Elsevier and the paper is closed access to this day, which not only means one can not inspect the full study design, but also unable to know who funded this project. Fortunately there is a sci-hub backup. There appears to be no Funding Sources or Conflict of Interests sections, but the first two paragraphs of the Introduction strongly imply that the results can be read as coming from the USDA research:
The US Department of Agriculture's (USDA) Food Safety and Inspection Service (FSIS) has developed a farm to table risk assessment model of Salmonella enteritidis (SE) in eggs (USDA-FSIS, 1998).
This report describes development and results of a version of the production segments of the FSIS risk assessment model for SE.
Ebel and Schlosser opened by stating that actually surveying a large amount of eggs to count how many of them are contaminated and under what pre-conditions would be nice, but this would be "extremely expensive to conduct" and "To date [2000], no such survey has been completed". This study is conducted on the USDA-FSIS's risk assessment model. More importantly, the authors made it clear that
In this version, the model estimates the fraction of all eggs produced in the US (during a 1-year period) that are SE-contaminated at lay.
, and,
Certainly, a full assessment of the exposure risk to consumers from SE-contaminated eggs must also account for microbial growth dynamics as eggs are marketed and stored, as well as egg preparation practices (e.g. pooling of eggs in meals, cooking practices) and consumption patterns among the consumer population.
The evidents provided in the paper could well suffice to support its claims (and I am uneducated to judge any of the design & interpretation), but bottom line is that the estimated probability of encountering a contaminated egg does not directly translate to one that can be relied upon in the home kitchen.
Can we do better, within reasonable web search effort? I am not sure. Cardoso et al. (2021) has a review article. Cardoso et al. cited Ebel and Schlosser (2000, the paper shown above), and followed by citing Europe's EFSA and ECDC from 2019 that:
roughly 0.37% of the tested table eggs (n = 6252) were Salmonella positive (n = 23).
... which is 111x the lower bound of 90% confidence interval from Ebel & Schlosser. Transportation and how eggs were handled prior to transport/sell probably matter quite much. From Ebel & Schlosser's numbers I would need to take a leap of faith to estimate my real risks. Now, allow me to move on to stuff about raw Salmon fillets until I forget that I wanted to have poke bowls...