t-threads
private beta
search about

Prof. Biol. @UW. Information flow in biology, society, & science. I love crows and ravens. he/him

Jun 30, 2020 2:47 AM
copied to clipboard

Graphs like the ones below scare me. I can’t see how one can spin this as anything other than a disaster in the making. https://www.nytimes.com/interactive/2020/us/coronavirus-us-cases.html

tweet media tweet media


People often complain that I’m looking at cases instead of some other metric: hospitalizations, deaths, contestants on the Bachelor who test antibody positive, whatever.

“Positive cases are determined by testing.”

“You can’t compare April to June.”

“Deaths are what matter.”


I look at cases for two primary reasons.

  1. They are a leading indicator. Deaths lag cases by 2-3 weeks, at which point it’s too late.

  2. Cases are the causal drivers of new cases. Cases now are the best predictor of cases in a week.


It’s true that changes in testing make it difficult to directly compare case numbers between early April and late June.

But that’s not what I’m interested in. I don’t all that much care whether we’re above the peak of the initial spike. What I care about is where we are going.


So what I want to do is compare cases in Florida today with cases in Florida a week ago. Barring extraordinary circumstances, testing doesn’t ramp up exponentially the way that cases have. On the timescale of a week, testing changes are usually a wash.

tweet media


“What about deaths? Why shouldn’t we care about these at least as much as cases? Deaths are not increasing in Florida.”

Not yet, that’s true.

But deaths are a severely lagging indicator or at least three reasons.


  1. It takes 2-3 weeks to progress from infection to death.

  2. It often takes a week or more for deaths to be registered.

  3. If the leading edge of a spike is driven by younger people, you need one or more rounds of transmission to reach a vulnerable population. But you will.


So the trajectory of deaths tells me about how who was infected a month ago was changing, a month ago.

I want to know how who is infected now is changing now, and what that bodes for the future.