04-23-2020, 06:59 PM
(04-23-2020, 06:31 PM)stupac2 Wrote: Yeah, what's the good news here? Even the worst-hit place in the US isn't all that close to herd immunity (which I've seen as starting at 50-70%, though that could be wrong). If NYC's experience is what it takes to get 1/3 of the way to herd immunity in a short timeframe... eesh, that's not a happy thought.
So, I will preface this by saying that this is rampant speculation on my part, but it is based on my knowledge of the models and their weaknesses. So, the SEIR model predicts exponential growth while R is > 1, and to a first approximation, R is reduced proportionally to the proportion of people who are immune. So, if you have an R0 of 2.5, you would need 60% of patients to be immune so that Rt goes down to 1, which is how you get a prediction that herd immunity will kick in at 50 - 70%. But one of the weaknesses of the crude SEIR is that it treats all people the same, and R0 as an average number of infections that each infected person causes. But here is the thing, the same characteristics that make a person more likely to be infectious (many close social interactions, poor hand hygiene, etc.) also make them more likely to become infected. So, if you figure that there is a distribution of individual R's, it follows that the people with the highest R are also the ones most likely to become infected early. So, the one upshot is that as you develop herd immunity, the people who are most infectious when they are infected start to develop immunity first, so you not only reduce the pool of susceptible people, but remaining susceptible people are both less likely to get infected and less infectious, so Rt goes down not just because of the decline in the susceptible pool but also because over time the remaining susceptibles are less susceptible and when they do become infectious, they are less infectious, so you might expect a smaller proportion of people to need to be immune before you start seeing decreased infectiveness of the virus.
Now, I have no idea how big this effect is, or if it even is a thing (remember this is speculation on my part). I suspect, however, that some of the more complicated models may do a better job accounting for this effect.
BC
