As of April 10, the Institute for Health Metrics and Evaluation is predicting 61,545 COVID-19 deaths by August 4, 2020.[1] That prediction is up a bit from their prediction of a few days ago for August 4, which was 60,415, but not up enough to materially change their message.
I haven’t studied how they produce their model, but I assume it involves the fact that even as fatalities have been increasing, hospitalizations have been decreasing.
I’m exploring the hypothesis that a stupid-simple model (SSM) which ignores that kind of detail might actually be better than a model that is too prone to short-term distractions from short-term phenomena. For instance: while NY is driving the overall numbers in the US, and it is doing a great job at social distancing, hospitalizations go down. But exponential growth rates mean that rural areas may soon be driving the overall US numbers, and we don’t know what kind of social distancing we’ll be seeing there.
If you’re not interested in exploring such a hypothesis, there’s no reason for you to read further. it’s interesting to me; it may not be to you. I’m not claiming that it’s correct; just that it’s interesting.
So: Maybe it’s better to make no assumptions about the usefulness of various details, and stick solely to the most basic known facts: what the current fatality rates are, and what will happen if they are an indicator of the short-term future. They can’t be used to predict the long-term future, because we know that, because of herd immunity and running out of people being newly exposed to the virus, the long-term curve will be basically logistic rather than exponential. But even short-term predictions based on that approach imply, if correct, that the IHME prediction for Aug 4 is drastically wrong.
Here is the recent growth in fatality rates. If varies from day to day, but appears to still be basically exponential:
2020-04-05: 1.14
2020-04-06: 1.04
2020-04-07: 1.18
2020-04-08: 1.16
2020-04-09: 1.12
2020-04-10: 1.13
We average the last 4 days’ exponential growth and get 1.146. We then take yesterday’s numbers:
2020-04-10: Deaths: 2,104; Accumulated deaths: 18,578
And project forward using the SSM, seeing that the SSM anticipates more fatalities by 9 days from now than the IHME predicts for Aug 4:
2020-04-11: Increase: 2,411 Total: 20,989;
2020-04-12: Increase: 2,763 Total: 23,752;
2020-04-13: Increase: 3,166 Total: 26,919;
2020-04-14: Increase: 3,628 Total: 30,547;
2020-04-15: Increase: 4,158 Total: 34,706;
2020-04-16: Increase: 4,765 Total: 39,472;
2020-04-17: Increase: 5,461 Total: 44,934;
2020-04-18: Increase: 6,259 Total: 51,193;
2020-04-19: Increase: 7,173 Total: 58,366;
2020-04-20: Increase: 8,220 Total: 66,587;
2020-04-21: Increase: 9,420 Total: 76,007;
2020-04-22: Increase: 10,795 Total: 86,803;
2020-04-23: Increase: 12,371 Total: 99,175;
Comments