Now THIS is how you calculate risk!

[dreamshark]

I recently spent some time ranting and raving to quadong (because everybody else is tired of listening to this stuff) about a terrible NY Times article purporting to prove that the risk of breakthrough infection is minuscule because the average person's chance of infection on any given day is 1/5000. If you don't see what's wrong with that logic, I'm not going to bother explaining, but it's not a very honest or useful use of statistics.

Here's exactly the opposite - the Microcovid Project. It's a thoughtful, detailed, and  completely transparent risk calculator that attempts to calculate the risk for a specific event and then multiplies the risk out over time. It's also an actual calculator taking a whole bunch of factors into account: where you live, what kind of mask you were wearing, what you know about other people present, etc. And then you can add in your vaccine status, if any. They show you exactly what assumptions they are making, so if you don't agree with their assumptions you can make your own adjustments to the results (e.g., they haven't updated for the new studies on Pfizer vs Moderna). 

This thing is really fun to play with. The well-designed interface remembers your last inputs and lets you make changes one factor at a time and watch how the risk changes. You can choose a premade scenarios (shopping for groceries, visiting a bar, etc.) or design your own. It is fascinating and really well done, IMHO. 

Matt, I am eager to hear what you think of this!

Comments

[redbird]

Aargh. I don't expect most of the press to be good at science reporting, but that's egregiously bad, and dangerously so. There are too many people who are eager for anyone and anything that will tell them that what they want to do is safe enough.

The Microcovid project's calculator is limited, but that's probably inevitable. Things it couldn't handle well included getting a ride from two fully vaccinated people, one of whom is being tested weekly, and anything in Canada. With both of those, I was mostly poking at it to see how it worked: The calculations for flying to and from Canada are different from a hypothetical domestic flight, because of the vaccination and testing requirements for crossing the border.

[dreamshark]

I think the most useful exercise is starting with a scenario and modifying one factor at a time to see how much difference it makes if everybody is masked, how long the exposure is, etc. You could perhaps simulate an international flight by assuming everybody on the plane is vaccinated?

They do not include the effect of weekly testing, but I'm not sure they should. Just because someone has been tested in the past week doesn't mean they aren't infected right now. Personally I think that's in the neighborhood of hygiene theater. But I'll bet they'll add it if someone finds a way to do a study that quantifies the effect of weekly testing.

And since you seem to be agreeing with me about the NY Times article, just let me point out that I calculated the chances for an UNVACCINATED person being infected on any given day and guess what? It was about 2.5 in 5000. So I guess COVID just isn't very dangerous at all, right? I think he just proved the opposite of what he was trying to prove. What a nimrod.

[quadong]

I bet that weekly testing accomplishes about as much as a typical cloth mask. As you've said, it's not magic, but it cuts the risk. Assuming people immediately isolate when they get a positive test (and before they have symptoms), the testing reduces the amount of contact they have with others. It doesn't reduce it to zero, but it reduces it. Daily testing would reduce it further (but still not to zero).

[quadong]

Seems like a good calculator. I tested a few things I thought it might screw up, like whether it increased the risk level linearly with the length of an interaction, and it did not screw them up. It seems to come to about the same conclusions that I did before, which is that biweekly Minnstf picnics are a reasonable risk except for particularly vulnerable people. That assumes everyone is vaccinated, and holds with or without masks. My couple-times-a-month indoor hangouts with small groups of friends come with about the same total risk, which is again reasonable.

I'm a little confused by "Describe the scenario", which only lets you say that you'll be with 2 other people, but then later you can set the number of people? I think it is saying that it will start you at 2 and you can adjust it. Or is it saying that you'll be near 2 people, but others will be filtering past?

The biggest weak point is that it only models the risk of getting COVID, not the risk of getting sick, or being hospitalized. There should be an additional multiplier for that, at least optionally.

[dreamshark]

"The biggest weak point is that it only models the risk of getting COVID, not the risk of getting sick, or being hospitalized. There should be an additional multiplier for that, at least optionally."

Those would be good additions. Maybe you should suggest that?