Lies, Damned Lies And Statistics
Jim Prevor’s Perishable Pundit, September 29, 2006
Our piece exploring the subject Is Salinas Getting a Bum Rap On Food Safety raised the question of whether the math and methodology used to study foodborne illness outbreaks biases the results against larger producers — such as the Salinas Valley. You can read the piece right here.
We got a lot of feedback on that, including several requests that we lay out the math more exactly. So here it goes:
The key is the FDA’s explanation in its letter to California Lettuce Growers:
FDA is aware of 18 outbreaks of foodborne illness since 1995 caused by Escherichia coli O157:H7 for which fresh or fresh-cut lettuce was implicated as the outbreak vehicle. In one additional case, fresh-cut spinach was implicated. These 19 outbreaks account for approximately 409 reported cases of illness and two deaths. Although tracebacks to growers were not completed in all 19 outbreak investigations, completed traceback investigations of eight of the outbreaks associated with lettuce and spinach, including the most recent lettuce outbreak in Minnesota, were traced back to Salinas, California.
Putting the current outbreak aside, the FDA identified 19 total outbreaks comprising 409 illnesses and two deaths — a total of 411 health incidents. The 411 health incidents divided by the 19 outbreaks means an average of 21.64 people were reported ill or died on each outbreak.
Now, Susan Pollack, an economist at the USDA Economic Research Service in Washington, D.C., shared with us some production statistics for 2005. The gist of the numbers is that California accounts for about 75% of all production of lettuce and spinach in the United States.
It is probably a bit lower than that as many producing states, such as New York, produce such a small quantity of lettuce that USDA doesn’t include their production in the data.
So, if the average California-based outbreak produces illness that makes the FDA aware of only 21.64 health incidents, we can say that a producer elsewhere who might equal 5% of California production will result in only a little more than 1 person being known as getting sick or dying. (With the exception of Arizona, there is no state that equals even 5% of California production on these items.)
These levels of illness are untraceable because you can’t survey people to find out what they ate and find out any statistically valid information on a survey of one or two people.
So the FDA is blocking the sale of spinach from three counties and sending letters to California growers based on a statistical chimera. There is no legitimate reason to believe these fields are currently less safe than any other fields in America. In effect, the FDA is punishing producers in a growing area for being big, not for being bad.
What is really going on here is a triumph of modern civilization. In the old days, people all over America got killed by their horses every single day. Either they were thrown or they were kicked. It was accepted as part of the common struggles and misfortunes of life. Most deaths never even made it into the newspapers.
Flying on a jumbo jet is infinitely safer than riding a horse, but when one jet crashes it kills 400 people at a shot and appears on Internet news reports instantly and prominently.
Our age is one of low incidence of tragedy but high intensity of tragedy because ours is an age of scale.
So people always got food poisoning and always died. In fact, they both got sick and died far more frequently than they do today. But we are now aware of more outbreaks partly as a result of improved detection but also as a result of consolidation of production allowing more effective tracking.
Here is the irony: It is large-scale production and processing that provides both the technological and capital base needed to reduce foodborne illness and also provides the scale that makes detection of foodborne illness more likely.
So as the food supply gets safer, it appears more dangerous. The FDA, which could serve a useful purpose in communicating this phenomenon, has given no indication that it realizes the phenomenon even exists.