What Should We Panic About?

In a pun-laden op-ed in the Wall Street Journal (examples: “Lettuce Try Not to Panic”, “Will a tragic overreaction topple Caesar and lead to the decline of the romaine empire?”) Jim Prevor argues that the CDC overreacted to the E. coli outbreak of a few weeks ago:

The Centers for Disease Control and Prevention urged before Thanksgiving that “U.S. consumers not eat any romaine lettuce, and retailers and restaurants not serve or sell any” until the current E. coli outbreak is resolved. This effectively closed down the romaine industry, producing tens of millions of dollars in losses of the highly perishable crop. The advisory remains in effect for romaine from the Central Coastal growing regions of Northern and Central California.

The waste is worth it, right? It seems straightforward that no one should eat romaine when the lettuce is making people sick. But it isn’t so clear when you look at the numbers.

The way the CDC identifies a food-safety outbreak is by interviewing sick people and healthy people. If there is a big difference in their answers, the CDC zeroes in on a cause. There are 43 people known to be infected with the outbreak strain of E. coli 0157:H7. The CDC interviewed 25 of them. Eighty-eight percent of those 25 people, as opposed to 47% of the general population, said they ate romaine lettuce in the week before they got sick.

So it probably was romaine that got those people sick—16 severely enough to be hospitalized. But what rational people should do with this information is much less certain.

The U.S. population is about 326 million. If 47% of the population eats romaine each week, that’s about 153 million people. We know of 43 people who have been infected with E. coli from romaine lettuce. According to the CDC, illness start dates range from Oct. 8 to Oct. 31—a period of about three weeks. If we assume, conservatively, that each of those 153 million people eats one serving of romaine each week, then we can figure there were 459 million servings consumed during the three weeks the infection was being transmitted.

This means the odds that eating a serving of romaine will make you sick are about 1 in 11 million, and the odds it will put you in the hospital are less than 1 in 28 million. To put this in perspective, the probability of getting a royal flush in poker is dozens of times as great, at 1 in 649,740, and the probability of an amateur hitting a hole-in-one in golf is hundreds of times as great, at 1 in 12,000. If you are that risk-averse, you should stay away from dogs—the lifetime odds of getting killed by a dog attack are about 1 in 112,000. Even the odds of getting struck by lightning in a particular year are higher than 1 in a million.

To put it another way: If this outbreak were active every day, and you ate one salad a day, on average you would be hospitalized for E. coli once every 77,000 years.

I suppose I shouldn’t blame Mr. Prevor for the puns. Those are probably editorial.

I think his argument is specious. The cases weren’t distributed uniformly throughout the country—they occurred in clusters. Since we don’t keep records at a fine enough granularity (and shouldn’t) we can’t identify precisely what field produced the lettuce that carried the bacteria. We don’t know the actual incidence (it could have been hundreds of times higher but unreported). We don’t even know the actual incidence in the areas in which the cases were reported.

E. coli isn’t endemic in romaine lettuce. It was introduced from outside. Possibilities include animals, the water used in the field, or poor field sanitation conditions. One worker eliminating in a field can sicken hundreds of people. What incentives should be put in place to ensure that producers take sufficient care?

The argument Mr. Prevor is making can be levelled against most laws and most preventive measures. Every restaurant (at least every restaurant I’ve ever worked in) has prominent signs urging workers to wash their hands. Should those be removed? They were made the law due to incidents of disease being spread in restaurants a century ago. Has it all been a waste? Should we abandon all public health measures?

That’s my question for Mr. Prevor. What prevalence is enough for us to take steps in cases of preventable death?

Update

It isn’t just about reducing risk. It’s also about confidence. Do warnings like the CDC’s increase or decrease confidence? Our present system of food distribution depends on trust. What would happen to California produce growers if there were a catastrophic decrease in trust in the food supply?