Hilda Bastian and John Ioannidis on coronavirus decision making; Jon Zelner on virus progression models

1. Hilda Bastian writes:

Doing nothing for which there is no strong evidence is doing something: it’s withholding public health interventions that, on the balance of what we know, could save a lot of lives and trauma – including the lives of a lot of healthcare workers.
Secondly, the need for societies to be able to monitor the impact is an argument for putting more effort into monitoring. Weaknesses in that is not a reason to not act. . . .

We do not know “the” case fatality rate, but that won’t be the same everywhere, dependent as it is on regional differences like health system capacity and levels of antibiotic resistance for secondary pneumonia. And while it means best and worst case scenarios are far apart, that does not of itself give best case scenarios greater weight. . . .

Could there be fiascos from over-reaction? Yes, there could, but several countries have introduced measures that are draconian, appear to have pegged outbreaks back, and are loosening the measures. Could there be fiascos from under-reaction? Well, we already have some of those. . . . there is broad consensus that this is a public health emergency, and we have to take action, not just sit there studying the situation and waiting for better information before acting. I think the stakes are too high to ignore the public health community urging us to act in favor of a “hot take” from someone who doesn’t seem to have done his homework.

She was clarifying some issues raised by this post from John Ioannidis, who wrote:

At a time when everyone needs better information, from disease modelers and governments to people quarantined or just social distancing, we lack reliable evidence on how many people have been infected with SARS-CoV-2 or who continue to become infected. Better information is needed to guide decisions and actions of monumental significance and to monitor their impact. . . . Three months after the outbreak emerged, most countries, including the U.S., lack the ability to test a large number of people and no countries have reliable data on the prevalence of the virus in a representative random sample of the general population. . . .

2. I’ve been talking a bit with Jon Zelner about coronavirus progression models. Zelner writes:

I think in some ways that the whole transmission modeling approach, when done well, shares a lot of DNA w/MRP [multilevel regression and poststratification, also known as Mister P]. Like this paper, Substantial undocumented infection facilitates the rapid dissemination of novel coronavirus (SARS-CoV2), by Ruiyun Li, Sen Pei, Bin Chen, Yimeng Song, Tao Zhang, Wan Yang, and Jeff Shaman that just came out—one of the times where I think we can say the material is indeed tabloid-worthy—directly models both the transmission process and the observation process that together give rise to hard-to-interpret patterns of who is observed testing positive for coronavirus.

One of the things that makes these nonlinear/diff-eq models kind of interesting in this way is also that sometimes you can’t explain the geometry of the epidemic curve with just the observed data, i.e. the peak is too high to be explained by just the small fraction of observed cases. So modeling the dynamic process actually informs the estimates of the age-specific reporting rates, etc.

You know that saying, There’s nothing so practical as a good theory? That’s what’s going on here. Latent-variable models are necessary both to understand the process and to make predictions about observables.

See here, here, and here for more on these models.