4 ms·

I've seen lots of sites[1] that show the risk at someone will end up being covid-positive during a large gathering (like Thanksgiving). This data is, as far as

I've seen lots of sites[1] that show the risk at someone will end up being covid-positive during a large gathering (like Thanksgiving). This data is, as far as I can tell based on an estimate of how many people in a given population are likely to have an infection. Does anyone know of any sites/data that offer the same risk analysis, but can take into account people who receive negative covid tests? For example, how would you estimate the risk of a 15 person Thanksgiving gathering if all 15 people receive negative covid tests prior to gathering? [1] https://covid19risk.biosci.gatech.edu/ https://covid19risk.biosci.gatech.edu/

brewdad 6y ago
Considering the White House has been doing daily COVID tests long before Trump caught it and has still had multiple large spread events, I'd say it doesn't tell you much. If you could all get tested a week before, isolate completely for the next 7 days, and continue to test negative the day before or day of, then you can be confident. Otherwise, with the 5-7 lag where you can be infected and spreading the disease while still testing negative and having no symptoms, a single test isn't a guarantee. Depends on your group's specific risk factors.
- lacker 6y ago
  I mean, it’s a little better because the danger window is 7 days instead of 14 days. A back of the envelope estimate would be that 2X people with clear coronavirus tests have equivalent risk of someone having coronavirus as X people who have had no testing.
- mixmastamyk 6y ago
  WH folks were traveling all over the country for election events.
- usrusr 6y ago
  What's the point of sandwiching the 7 days between tests? Wouldn't 7 days isolation, then test be just as good? The only scenario (ignoring actual technical false negatives) where you'd get positive on the first, negative on the second would be a very late stage where infectivity was mostly over already at the time of the first test.
matsemann 6y ago
Are tests so easily available now that they can be "wasted" on people just using them for convenience, instead of reserving the testing capacity for better use? The test may anyway be worthless if you take it and then anyways have been exposed during a flight or so.
- lacker 6y ago
  Yes. Near me at least in Oakland CA there are several providers that offer no-appointment, no-referral covid testing.
- sethhochberg 6y ago
  NYC and NY State are running TV ads reminding everyone to get tested often, because you can be contagious and asymptomatic and frequent testing helps detect that earlier. I'm sure there is regional variability... but in at least some places, it seems like tests are indeed plentiful. If I had to guess, the regions that have decided to handle the pandemic by largely doing nothing different than normal don't have great availability, but probably more for political reasons than logistical ones.
- renewiltord 6y ago
  Depends on the region. Urban California, for instance, has high testing capacity. I get tested here in SF pretty often and any time I'll be hanging out with people. Rural and semi-urban areas like SLO etc. are quite strapped on the other hand. EDIT: See below, my info was out of date by a week. I would hypothesize it's everyone who wants to be tested for the holidays.
  - matsemann 6y ago
    > I get tested here in SF pretty often and any time I'll be hanging out with people Given the time it takes from exposure to symptoms and tests register it, isn't that kind of a false sense of security? Exposed on sunday - takes a test thursday that's negative - hangs out with friends on friday, kinda deal.
    - 6y ago
timr 6y ago
There's no one way to estimate it, so any such tool is just going to turn into a reflection of the assumptions used to create it. Setting aside the probability that someone could be asymptomatic and contagious after a negative test (which actually seems to be fairly low), you'd need to know the false-negative characteristics of the different tests that were used. Unfortunately, people seem to be far more concerned with the former risk than the latter -- they assume that their tests are infallible (wrong), but that the risk of asymptomatic transmission is relatively high (also wrong).
- eminence32 6y ago
  > you'd need to know the false-negative characteristics of the different tests that were used Good point about this. Do we have enough data to compute this? my layman's guess is that if someone is asymptomatic and receives a negative test, they're not likely to follow it up with another test (which would be one way to learn the first test was a false-negative). But maybe there are other clever ways at getting at the false-negative rate for each type of test.
  - pbronez 6y ago
    My organization has put together some data on this: https://covidtestdb.com/performance/ https://covidtestdb.com/performance/ We pulled together performance data for dozens of COVID-19 diagnostics, which you can download. We also created a stats tool to help you understand the Predictive Value of each test based on how widespread the disease is in the tested population. The key metrics we look at are: Positive predictive value (PPV), the fraction of positive tests where the person actually had the disease. Negative predictive value (NPV), the fraction of negative tests where the person actually did not have the disease.
im3w1l 6y ago
I think it's not just about the "first-order "risk of infection itself. It's also the fact that a disease is a graph phenomenon, and gatherings will connect parts of the graph that are normally separate. This means that infections will lead to more than average amounts of secondary, tertiary infections. But on the other hand this may not matter as much anymore since we are in a late stage of the epidemic, and the disease is everywhere already. Also I don't mean to dissuade anyone from attending with this.
sjg007 6y ago
You can do the math, take the GA Tech prevalence data and combine it with the false negative rate (has covid, tests negative). If you want to go further, consider the time you will spend together in an indoor environment mask-less. I think there are some estimates of indoor transmission probabilities per time of exposure to a positive case. Bear in mind the average 5-8 day incubation period so even if you test the day before you are still almost a week behind in exposures. https://www.statnews.com/2020/08/20/covid-19-test-accuracy-supplement-the-math-of-bayes-theorem/ https://www.statnews.com/2020/08/20/covid-19-test-accuracy-s...
- chadash 6y ago
  Very hard to estimate false negative rate for a group given that the false negative rate itself is fluctuates considerably based on when you get the test. False negatives are very high the day before symptoms and go down dramatically at the third day after symptoms (which is the best day to get tested).
  - sjg007 6y ago
    Yep this is basically why people need to not gather.
itronitron 6y ago
You may also want to consider at this point that social gatherings of people do not represent a random sample of the population but are instead self-selecting for higher exposure to covid.

brewdad 6y ago

Considering the White House has been doing daily COVID tests long before Trump caught it and has still had multiple large spread events, I'd say it doesn't tell you much. If you could all get tested a week before, isolate completely for the next 7 days, and continue to test negative the day before or day of, then you can be confident. Otherwise, with the 5-7 lag where you can be infected and spreading the disease while still testing negative and having no symptoms, a single test isn't a guarantee. Depends on your group's specific risk factors.

lacker 6y ago
I mean, it’s a little better because the danger window is 7 days instead of 14 days. A back of the envelope estimate would be that 2X people with clear coronavirus tests have equivalent risk of someone having coronavirus as X people who have had no testing.
mixmastamyk 6y ago
WH folks were traveling all over the country for election events.
usrusr 6y ago
What's the point of sandwiching the 7 days between tests? Wouldn't 7 days isolation, then test be just as good? The only scenario (ignoring actual technical false negatives) where you'd get positive on the first, negative on the second would be a very late stage where infectivity was mostly over already at the time of the first test.

matsemann 6y ago

Are tests so easily available now that they can be "wasted" on people just using them for convenience, instead of reserving the testing capacity for better use? The test may anyway be worthless if you take it and then anyways have been exposed during a flight or so.

lacker 6y ago
Yes. Near me at least in Oakland CA there are several providers that offer no-appointment, no-referral covid testing.
sethhochberg 6y ago
NYC and NY State are running TV ads reminding everyone to get tested often, because you can be contagious and asymptomatic and frequent testing helps detect that earlier. I'm sure there is regional variability... but in at least some places, it seems like tests are indeed plentiful. If I had to guess, the regions that have decided to handle the pandemic by largely doing nothing different than normal don't have great availability, but probably more for political reasons than logistical ones.
renewiltord 6y ago
Depends on the region. Urban California, for instance, has high testing capacity. I get tested here in SF pretty often and any time I'll be hanging out with people. Rural and semi-urban areas like SLO etc. are quite strapped on the other hand. EDIT: See below, my info was out of date by a week. I would hypothesize it's everyone who wants to be tested for the holidays.
- matsemann 6y ago
  > I get tested here in SF pretty often and any time I'll be hanging out with people Given the time it takes from exposure to symptoms and tests register it, isn't that kind of a false sense of security? Exposed on sunday - takes a test thursday that's negative - hangs out with friends on friday, kinda deal.
  - 6y ago

timr 6y ago

There's no one way to estimate it, so any such tool is just going to turn into a reflection of the assumptions used to create it. Setting aside the probability that someone could be asymptomatic and contagious after a negative test (which actually seems to be fairly low), you'd need to know the false-negative characteristics of the different tests that were used. Unfortunately, people seem to be far more concerned with the former risk than the latter -- they assume that their tests are infallible (wrong), but that the risk of asymptomatic transmission is relatively high (also wrong).

eminence32 6y ago
> you'd need to know the false-negative characteristics of the different tests that were used Good point about this. Do we have enough data to compute this? my layman's guess is that if someone is asymptomatic and receives a negative test, they're not likely to follow it up with another test (which would be one way to learn the first test was a false-negative). But maybe there are other clever ways at getting at the false-negative rate for each type of test.
- pbronez 6y ago
  My organization has put together some data on this: https://covidtestdb.com/performance/ https://covidtestdb.com/performance/ We pulled together performance data for dozens of COVID-19 diagnostics, which you can download. We also created a stats tool to help you understand the Predictive Value of each test based on how widespread the disease is in the tested population. The key metrics we look at are: Positive predictive value (PPV), the fraction of positive tests where the person actually had the disease. Negative predictive value (NPV), the fraction of negative tests where the person actually did not have the disease.

im3w1l 6y ago

I think it's not just about the "first-order "risk of infection itself. It's also the fact that a disease is a graph phenomenon, and gatherings will connect parts of the graph that are normally separate. This means that infections will lead to more than average amounts of secondary, tertiary infections. But on the other hand this may not matter as much anymore since we are in a late stage of the epidemic, and the disease is everywhere already. Also I don't mean to dissuade anyone from attending with this.

sjg007 6y ago

You can do the math, take the GA Tech prevalence data and combine it with the false negative rate (has covid, tests negative). If you want to go further, consider the time you will spend together in an indoor environment mask-less. I think there are some estimates of indoor transmission probabilities per time of exposure to a positive case. Bear in mind the average 5-8 day incubation period so even if you test the day before you are still almost a week behind in exposures. https://www.statnews.com/2020/08/20/covid-19-test-accuracy-supplement-the-math-of-bayes-theorem/ https://www.statnews.com/2020/08/20/covid-19-test-accuracy-s...

chadash 6y ago
Very hard to estimate false negative rate for a group given that the false negative rate itself is fluctuates considerably based on when you get the test. False negatives are very high the day before symptoms and go down dramatically at the third day after symptoms (which is the best day to get tested).
- sjg007 6y ago
  Yep this is basically why people need to not gather.

itronitron 6y ago

You may also want to consider at this point that social gatherings of people do not represent a random sample of the population but are instead self-selecting for higher exposure to covid.