# Slides for Covid-19 talk

Here are my slides for my recent COVID-19 talk at the Centre for Applied Mathematics in BioScience and Medicine (CAMBAM). It’s an updated version of the one I gave to the FDA.

## 2 thoughts on “Slides for Covid-19 talk”

1. yohannes shiferaw says:

I am a bit skeptical about the way fatality rate is modeled. My reasoning is the following. If you look at the % deaths in Ca (see link bellow), you will see that about 70% of deaths are for the 70+ age group. Bellow 49 years is about 6%. So death rates are highly dependent on age. Now, post lockdown, the infection rates of these age groups is also very different. Younger people are readily infected while the vulnerable are sheltering in place–so the average age of new infections is actually bellow 50. This is clearly because both of these groups did a risk assessment and adjusted their behavior accordingly.

This explains two things: The first is the pronounced lag we are seeing now between deaths and new infections. Note that this lag was not there early in the epidemic (in NY for example) because the infection rate of the vulnerable was essentially the same as that of the young (nursing homes got hit relatively early since they weren’t ready…etc) . This, also explains how Sweden can proceed without masks and moderate measures while their death rate is plummeting. Essentially they are finally effectively protecting their vulnerable after screwing it up at the beginning.

So my feeling is that an accurate model of fatality rate will have to have to break the population into at least 2 groups–essentially young and old, each with very different infection rates and vastly different fatality rates. I believe this will explain the lag we are seeing the Southern States and will give a better model of the time dependence of the fatality rate. Note that the coupling between the two groups will be crucial.

Anyway, hope this helps the modeling effort. I can’t think of anything else to add.

Like

2. @yohannes Yes, age stratification is certainly important and something we have thought about. The data wasn’t there for most places in March when we started this but that has changed so we could try to do this in the next generation of the model.

Like