The COVID Report with David Blake: 'Cases per fatality'

Monday, Aug. 17

We’ve gone over how to measure the percent infected in each state, and Rt. Today I want to go over another measure of testing effectiveness, the measure “cases per fatality,” or the reciprocal of the case fatality ratio.

Usually this is reported as a cumulative measure, as the total number of cases divided by the total number of deaths. However, cases per fatality changes over time.

We have, from the CDC, an estimate that the infection fatality ratio is 0.65 percent now, which implies there will be 153 infections per fatality.

I plotted time-dependent cases per fatality at different time-points in the pandemic to show that the number we are finding is going up, and that still we are not finding enough.

If you ever really want to suppress COVID-19, you need to find nearly all the infections and find them within 48 hours of symptoms showing up. Given our current testing setup, that is a tall order. But nothing less will work, and other countries routinely are successful in this measure.

Spain right now finds 400 cases per fatality.

Before I go on to show plots of the first four months, know these things.

If you find all the infections, your cases per fatality will be close to 153. If your state has more infections in older people, or more nursing home problems, in the period of the plot, your cases per fatality will be lower.

I did adjust these numbers for demographics changes across states, so that states with a larger proportion of older population are not treated unfairly. This measure is simple: the number of deaths from COVID in the past two weeks, divided by the number of cases over a two week period shifted 20 days into the past.

Here are the plots.

The dates are at the top. And we wonder, is New York really finding four times as many cases as Massachusetts and New Jersey? Isn’t that an addressable problem?