This is an important new finding and one that requires some quick reflections. Thread
We have all been very interested in seroprevalence studies
Why?
Because we're wondering whether the proportion of the population that has been infected is much larger than we might think
1/n
We have all been very interested in seroprevalence studies
Why?
Because we're wondering whether the proportion of the population that has been infected is much larger than we might think
1/n
If a much larger proportion has been infected, that would be great for two reasons:
1. Disease milder than we think (lower fatality rate, etc.)
2. We are closer to herd immunity than we thought.
2/n
1. Disease milder than we think (lower fatality rate, etc.)
2. We are closer to herd immunity than we thought.
2/n
So far, 270K NYers have been found to be infected. That's 1.3% of NY State population (of 19.5 million).
We all know that the true rate is much higher? Why?
Because we've been under testing!!
How much higher?
I've argued that we identify about 1 in 10 folks with COVID19
3/n
We all know that the true rate is much higher? Why?
Because we've been under testing!!
How much higher?
I've argued that we identify about 1 in 10 folks with COVID19
3/n
Seroprevalence study out today says that the underlying infection rate is 13.9%
That's about 10X what we've identified.
In line with the idea that we've been testing only about 10% of infected folks
That's about 10X what we've identified.
In line with the idea that we've been testing only about 10% of infected folks
One more important caveat:
The test characteristics matter a lot.
The NYS website doesn't give out the sensitivity but says specificity between 93% and 100% (huge range).
5/n
The test characteristics matter a lot.
The NYS website doesn't give out the sensitivity but says specificity between 93% and 100% (huge range).
5/n
So what does this mean?
If we assume 100% sensitivity and 100% specificity (no test is perfect but hey), then 13.9% is right (with CIs).
But what if we assume 100% sensitivity and 93% specificity, then what?
6/n
If we assume 100% sensitivity and 100% specificity (no test is perfect but hey), then 13.9% is right (with CIs).
But what if we assume 100% sensitivity and 93% specificity, then what?
6/n
Then, the underlying prevalence may be closer to 7% and nearly 1 in 2 positive tests are false positives.
Either way, this is NOT consistent with the idea that the true numbers are 40-85X the number of folks diagnosed (as was suggested by Santa Clara study).
7/9
Either way, this is NOT consistent with the idea that the true numbers are 40-85X the number of folks diagnosed (as was suggested by Santa Clara study).
7/9
Recap:
1.3% of NYers have been infected based on testing
Seroprevalence says 13.9% -- or about 10X.
In line with expectations
But it could be as low as 7% (if specificity of test is 93%)
There are a whole host of other issues but....
8/9
1.3% of NYers have been infected based on testing
Seroprevalence says 13.9% -- or about 10X.
In line with expectations
But it could be as low as 7% (if specificity of test is 93%)
There are a whole host of other issues but....
8/9
Sampling issues always important.
But NYS smart enough to tap into the brilliant @nataliexdean so I'm more comfortable.
Seroprevalence from NYS in line with expectations.
This doesn't say we are missing very large numbers of asymptomic patients...but clearly some
Fin
But NYS smart enough to tap into the brilliant @nataliexdean so I'm more comfortable.
Seroprevalence from NYS in line with expectations.
This doesn't say we are missing very large numbers of asymptomic patients...but clearly some
Fin
