David Dowdy and Gypsyamber D'Souza's three- or four-paragraph comment on the Johns Hopkins website is a response to what they consider to be irresponsible claims in the media that California and other places have achieved or are close to herd immunity. They wrote that "As infectious disease epidemiologists, we wish to state clearly that herd immunity against COVID-19 will not be achieved at a population level in 2020, barring a public health catastrophe."
This comment certainly is not written for statisticians. It contains little in the way of detailed analysis, let alone a model for making reasonably accurate predictions of timing in a changing political landscape. FWIW, they do cite to "a semi-mechanistic Bayesian hierarchical model to attempt to infer the impact of these interventions [such as national lockdowns] across 11 European countries."
But is their conclusion too dire? The estimates of SARS-CoV-2 infections on the order of 50 (or more) times the official numbers in California are interesting. Several statisticians have
savagely criticized these estimates as miscalculated and as unwarranted given the imperfect sensitivity of the serological tests. However, the criticism
may be overstated. So let's assume these estimates are correct. A 50-fold underestimate may sound shocking, but the denominator in the ratio is small, and the relevant parameter for the herd-immunity issue is the true prevalence (assuming, of course, that an immune response confers reasonably long-lasting immunity). Where are we today? The LA and Santa Clara county convenience sampling studies in April by the California researchers (used to produce the ratio of 50 so) give estimates of 2.8% to 5.6% and 2.5% to 4.2%, respectively, for the proportion of the population that has been infected. The highest raw number I have seen is 31.5% positive serologic tests administered to 200 passersby on a street corner in Chelsea, Massachusetts. Better designed studies are underway in various locations.
Dowdy and D'Souza assert that "To reach herd immunity for COVID-19, likely 70% or more of the population would need to be immune." If this is correct, the gap between where we are and the herd-immunity level seems substantial. Of course, that is just a gut reaction, and I wish I could point to a professionally respectable, statistically defensible modelling effort that gives a more optimistic answer to the question of whether we are approaching herd immunity in 2020.
------------------------------
David Kaye
http://www.personal.psu.edu/dhk3/index.htm------------------------------
Original Message:
Sent: 05-08-2020 10:36
From: Sven Serneels
Subject: Great article on misconceptions about herd immunity to COVID19
I have noticed that some other threads on COVID-19 have gotten highly emotional and political, which is not what ASA Connect is meant for. I do not intend to ignite any such discussion here. However, from a professional perspective, I can only disagree with Deirdre's opinion on this article. In its present state, it is poor science.
The article is nothing more than an opinion piece. It makes statements such as "with over 25,000 confirmed cases a day – it will be well into 2021 before we reach herd immunity". This is essentially what we in this forum would call a "naive forecast": just taking the last value of the time series as the forecast for all future cases. In other threads, more sophisticated models are discussed, which put this approach into ridicule.
Furthermore, some assumptions are hardly defensible. In the forecast, they use the confirmed cases a day as the metric to estimate future developments, whereas just a paragraph earlier they admit that the number of real infections is orders of magnitude higher. Actually an estimate from LA county, meanwhile dated by a few weeks, estimates the number of real cases at that time to be 28 to 55 times higher than the confirmed cases:
https://news.usc.edu/168987/antibody-testing-results-covid-19-infections-los-angeles-county/
There is obviously a difference between our densely populated areas and different onset times across the country, but not taking this information into account and using a naive forecast corresponds to some of the poorest data science I've seen reported in the media in recent times.
------------------------------
Sven Serneels
(Diretor, Data Science)
-posting as a private person-
Original Message:
Sent: 05-07-2020 10:01
From: Deirdre Middleton
Subject: Great article on misconceptions about herd immunity to COVID19
This article is a 2-minute read and clears up a lot of epidemiological misconceptions about COVID19.
Early Herd Immunity against COVID-19: A Dangerous Misconception
David Dowdy, Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health
Gypsyamber D'Souza, Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health
"As infectious disease epidemiologists, we wish to state clearly that herd immunity against COVID-19 will not be achieved at a population level in 2020, barring a public health catastrophe."
https://coronavirus.jhu.edu/from-our-experts/early-herd-immunity-against-covid-19-a-dangerous-misconception?fbclid=IwAR0GAXZaJCYr2WIpLO7WSq2GJDbfSdYnAyNID5PwkXIujgzRsnPE3A-t1sM
------------------------------
Deirdre Middleton, MPH
------------------------------