CDC reports people with Covid were twice as likely to have eaten at a restaurant before getting sick

Originally published at:




I wish the study had the data to distinguish between indoor and outdoor restaurant settings. In my town indoor restaurant dining is closed and many have set up outdoor areas with tables spaced out. It’s likely less risky than an indoor setting and probably more risky than eating at home but it would be nice to be able to quantify the relative risks.


This is a classic example of correlation not implying causation. It could, for example, simply be the case that people who do not eat in restaurants are also antisocial in other ways, and are therefore less exposed to contagious diseases. You cannot infer from this statistic that people caught the disease in restaurants.

So, not precisely fake news, but without caveats in the article, designed to mislead people into being against reopening restaurants.


Welcome, new comrade.


Why is 25% always the solution? 25% capacity is way too much for indoor restaurants. We’re gonna be dealing with this unless we get an effective vaccine.



I’d say it is a classic example of correlation not entailing causation. However, just as you can’t assume that eating at restaurants is the cause, neither can you dismiss it.


They also pretty clearly state the other limitations of their method: they can’t rule out other factors (a measured type of risk behavior may well be correlated with unmeasured risk behavior) and the survey participants were recruited after they received their test results. Someone with a positive test result in hand might be just more likely to report this behavior. It also took place in the NYC area during july which might not represent other areas, particularly since NYC had a very low incidence rate at the time. All in all, this is not something you want to draw extremely strong quantitative conclusions about risk, it is one piece of information in a rapidly evolving field.

1 Like

I think you’re probably right, especially considering that numerous studies have show next to no transmission from groceries and food service. It’s the indoors part that matters and that also applies to churches, bars, theaters and a whole lot of other higher risk situations. It would have been nice if they asked other questions such as belief in mask wearing & social distancing, whether they have attended any large gatherings, whether they still physically greet people (hugs, handshakes, etc).

Also, welcome to BBS!

Because no one at any level of leadership has a clue how to approach this so they apply arbitrary regulations in hopes that they will be a silver bullet. Meanwhile, established effective measures like basic mask regulations go completely unenforced.


I think it is also saying that this doesn’t just include restaurants, but also bars? Am I reading that wrong?

“Restricting the analysis to participants without known close contact with a person with confirmed COVID-19, case-patients were more likely to report dining at a restaurant (aOR = 2.8, 95% CI = 1.9–4.3) or going to a bar/coffee shop (aOR = 3.9, 95% CI = 1.5–10.1) than were control-participants.”

Any correlation indicating a RR of 2.0 (or so) is worth noting. Determining causation can come later, at this point it is a notable and actionable correlation.


some light rewrites, in case you’re accepting notes…being new and all…


I’m so glad that you comment on this forum regularly; you really are a gem.


What are you basing that on? What would be a good percentage based on epidemiology?

I’m guessing they are doing something like taking the definition of “full capacity” as persons per square foot and then giving a percentage of that such that it comes out with what they are considering an acceptable amount of social distancing. In fact, they probably just tried 50% and decided that wasn’t enough and then halved it again, but it definitely seems like they would account for the amount of social distancing they want to achieve when coming up with this number. If full capacity is 12 square feet per person, 25% would be 48 square feet per person. If you put people in a square room with 48 square feet per person they will be easily able to arrange themselves such that there will be a little less than 7 feet between them (or a wall). I assume the thinking went something like that. If you count the fact that people will be dining together and the particularities of room shapes as a wash then the logic doesn’t seem that bad, or do you have empirical evidence that it is?

1 Like

If anything the actual language used by the report would imply the opposite direction of causation than you are interpreting it as suggesting, i.e. having covid causes people to report that they had eaten at a restaurant.

1 Like

It’s not just restaurants. Schools that re-open for in-person education will require students to eat lunch at their desks – unmasked – while a teacher tries to teach them. NYC is planning to re-open schools Sept 16. Arts teachers will be expected to rotate through lunch rooms – thus exposing themselves to all the kids in turn, unmasked. The teachers’ union bosses cut a deal to stop the rank & file from striking. NYC students are planning a strike for Sept 16.


Those are separate results. They analyzed “dining at a restaurant” and “bar / cafe” as two variables that independently correlated with positive COVID results in the absence of known contact with a positive individual.

They don’t separate indoor/outdoor dining. They also don’t separate take out from a coffee shop vs. a bar. As @docosc explains, any correlation is interesting as a starting point.


I see @ejeffrey . Thank you!

1 Like