So just recently I wrote a post pointing out how reasonable it was to expect 6 doublings of COVID-19 deaths in the USA by early May (which would have given us a bit over a quarter million total persons dead), and how after that even one more doubling would be an unthinkable tragedy. But we’re not seeing doublings on the rate I feared (and which seemed reasonable given a comparison to other nations then-current rates of doubling when they had initiated nationwide stay-at-home policies before the USA). Commenter militantagnostic was early on the case, informing me on April 6th:
I have been plotting cummulative cases for Canada and the US on a semilog graph (Cases on the log scale) versus time. This gives a straight line for exponential growth. I get good fits r^2 = 0.999 (Canada) and 0.998 (USA) for and intervals of over 2 weeks ending March 27. The doubling period during this exponential growth was 2.9 days for Canada and 2.4 days for the US. In both cases the data starts failling below the regression line after March 27. The death rates should start falling [below the projected exponential growth rate] shortly if ICUs are not overwhelmed.
And indeed, while I thought we might get 3 more doublings at the rate of 1 per 4 days, the next doubling took 5 to 6 days, not 4. That original article went up early morning on April 5 using data from April 4 (and quoted the then-current doubling rate of once every 4 days), but on most days since the same website (using data collected by the European Centers for Disease Prevention & Control, or ECDC) has listed a doubling rate of once every 5 days and today it’s at once every 6:
Official data on deaths and death rates for multiple countries from the European CDC
I said at the time, of course,
Since the doubling rate depends on human behavior, it can and will change.
But I also said that I didn’t expect that change to show up for about 2 weeks. I had reasons for that which had a lot to do with when specific shelter-in-place orders came down, but I didn’t have data on what the actual human beings of the USA were actually doing, and it seems like those humans have been much more wise than their leaders.
Given all this, my minimum-6-doublings estimate seems likely to be wrong. Of course, there’s a big but coming: this still depends on human behavior. And guess what? The right wingers are now arguing overtly for an immediate or near-immediate return to “normal”. The problem of course is that there are still many cases at large in communities across the country. If we actually did immediately return to normal behavior, we wouldn’t start out again where we were in January. We would have multiple Patient-Zeros across the country in cities large and small, and in every single state, and in a few other locations like DC and Puerto Rico as well.
Texas Governor Greg Abbott proposed a reopening of his state during a speech he gave this weekend. He doesn’t propose an immediate return to normal, but he is proposing an immediate (or near-immediate) return to … something. From Newsweek:
“We will focus on protecting lives while restoring livelihoods,” Abbott said on Friday at a news conference.
“We can and we must do this. We can do both, expand and restore the livelihoods that Texans want to have by helping them return to work. One thing about Texans, they enjoy working and they want to get back into the workforce. We have to come up with strategies on how we can do this safely.”
Abbott said details of the executive order would be available next week and it is expected to provide businesses with a list of guidelines on how to safely reopen.
“We will operate strategically,” Abbott added. “If we do it too fast without appropriate strategies, it will lead to another potential closure.”
So we won’t reach the quarter-million dead by early May that I feared, but if the conservatives get their way and push for ever fewer limits on public businesses and gatherings before the medical situation warrants, we may still get there.
Here’s hoping that people that know a lot more about infectious disease than i do end up controlling the decision making process.
“One thing about Texans, unlike those lazy loafers in all the other states, they enjoy working and they want to get back into the workforce.” I guess he didn’t quite say that out loud, but we all heard him mumbling under his breath.
Those of us in the folk dance community shut things down a little over 4 weeks ago.
Looking back through a dance leader group on Facebook dance cancellations started around March 6.
@robertbaden:
Interesting! Thanks. I wonder if there’s a way to turn publicly available records of groups like this into data that can help refine models of pandemic spread, so the next time there’s an outbreak of something significant enough to require a state or national shelter-in-place response we can use cancelations of sports and hobby groups in a more predictive way.
Big data certainly can be evil, but if you’re not monitoring individuals, just public groups’ public event schedules, that doesn’t seem bad to me.
Interesting! Thanks. I wonder if there’s a way to turn publicly available records of groups like this into data that can help refine models of pandemic spread, so the next time there’s an outbreak of something significant enough to require a state or national shelter-in-place response we can use cancelations of sports and hobby groups in a more predictive way.
There was some work being done in predictive detection, but it mostly got budget-cut. Currently, the state of affairs is that there are AIs scanning through all the reports that they can get their hands on, looking for outliers. Build a model of what normal human medical interventions look like, and a sudden spike of something sticks out like a whooping cough.
The problem with all of this is having quality data that is delivered in a standard way. Unfortunately, there are people who value such information for financial reasons, or political ones.
A disaster detector would be a good “go short the stock market!” indicator.