My rough forecast for reported COVID-19 deaths on June 1st was based on the rolling 7 -day average of deaths as of that day and the idea that states loosening lockdown restrictions would translate into deaths that high or higher for the next 28 days. In particular, I said I'd bet the over on every one of the forecasts projecting 120K-128K total deaths as of June 28th. Turns out the actual total was 126K, so I would have lost a couple of those bets. Now that we have more data to look at, I can see more where I went wrong--essentially it boils down to a couple main factors:
- Given ~5% test positivity, there is a roughly 4-week delay between confirmed cases and reported new deaths.
- Before early April, essentially none of the US testing data is complete enough to be predictive of future new deaths.
If you look at April-May cases and compare to May-June deaths, in cases where there is ~5% test positivity, we see about 32 confirmed cases per reported new death, and if you look at individual states, this holds pretty well and it also holds pretty well that it is about a 28-day delay between the two.
So that last week of deaths in May that I used to project for June was really indicative of cases from late April, but for the US in aggregate, cases were lower in late May than they were in late April, so because there is so much lag between confirmed cases and reported new deaths, June was set to be better than May.
But now things are entirely different in terms of confirmed cases. Cases have increased significantly since mid-June and the impacts on reported new deaths are only now starting to be seen from that. For instance, looking at Minnesota:
628 cases -- May 17
369 cases -- June 14
19 deaths - June 14
5 deaths - July 12
The 5 deaths is actually better than you would have expected, but given such low expected totals (~11/day), there is likely to be some variance. The current problem we face--if deaths are of primary concern--is that cases bottomed out for Minnesota on June 19th, and that's less than 4 weeks ago. With Minnesota back up to 564 cases/day, deaths/day are likely to be increasing gradually over the next four weeks. Granted, Minnesota is not itself in such bad shape, but with test positivity so high in large states like Texas, Florida, and Arizona, and with people moving in and out of those states around the US, it's hard to be very optimistic about anything being well controlled in the near future.
By August 10th, I think the US will be up to 169K+ total reported deaths and 1900 new deaths/day. That's lower than the peak of 2100 new deaths/day we saw in mid-April when NYC was the epicenter, but potentially 1900/day is an underestimate given that the test positivity is so high across large portions of the south--when the test positivity has been high, we have seen a higher ratio of deaths to confirmed cases. Once COVID-19 gets past 169K deaths in the US, it will virtually be assured of being at least the 3rd-leading cause of death in the US for 2020, behind only heart disease and cancer, and that would be with roughly 5% of the US having been infected (symptomatic or otherwise).
5 percent positivity is a dream for a lot of states right now.
Some guy publicly complains about (and threatens legal action against) a restaurant because he believes it is discriminating against him because he wears pro-Trump clothing. In response, said restaurant airs its dirty laundry about guy to City Pages. Whatever, who knows who is in the right.
But said guy won't stop commenting in the comments section of the online article. It's like watching a car accident!
Holy bananas. Someone actually offered him legal advice that his lawyer should have —-> don’t comment about the case on social / media.
Well that was a comment section full of barely literate writing....
🙁 RIP Mythbuster Grant Imahara and father of Geoff Peterson, 49, of a brain aneurism
Here's the game. They should have included the entire bottom of the ninth! Looks like it was an exciting finish.
One double clutch away from an 8th no-hitter.
There should really be a rule against the White Sox wearing red and the Red Sox wearing black.
I think the league punished them by making them wear shorts.
Scraping old paint off my parents' shed - feels good to do some honest work for a change
Try not to inhale too much of the lead.
It's been painted every 2 years (because not scraped) and is post-lead. It has other problems, though. Think squirrels
Back on June 1st, I was feeling pretty pessimistic about the next 28 days. I wanted to revisit that comment and say both why I was wrong and why I'm way more pessimistic now.
My rough forecast for reported COVID-19 deaths on June 1st was based on the rolling 7 -day average of deaths as of that day and the idea that states loosening lockdown restrictions would translate into deaths that high or higher for the next 28 days. In particular, I said I'd bet the over on every one of the forecasts projecting 120K-128K total deaths as of June 28th. Turns out the actual total was 126K, so I would have lost a couple of those bets. Now that we have more data to look at, I can see more where I went wrong--essentially it boils down to a couple main factors:
- Given ~5% test positivity, there is a roughly 4-week delay between confirmed cases and reported new deaths.
- Before early April, essentially none of the US testing data is complete enough to be predictive of future new deaths.
If you look at April-May cases and compare to May-June deaths, in cases where there is ~5% test positivity, we see about 32 confirmed cases per reported new death, and if you look at individual states, this holds pretty well and it also holds pretty well that it is about a 28-day delay between the two.
So that last week of deaths in May that I used to project for June was really indicative of cases from late April, but for the US in aggregate, cases were lower in late May than they were in late April, so because there is so much lag between confirmed cases and reported new deaths, June was set to be better than May.
But now things are entirely different in terms of confirmed cases. Cases have increased significantly since mid-June and the impacts on reported new deaths are only now starting to be seen from that. For instance, looking at Minnesota:
628 cases -- May 17
369 cases -- June 14
19 deaths - June 14
5 deaths - July 12
The 5 deaths is actually better than you would have expected, but given such low expected totals (~11/day), there is likely to be some variance. The current problem we face--if deaths are of primary concern--is that cases bottomed out for Minnesota on June 19th, and that's less than 4 weeks ago. With Minnesota back up to 564 cases/day, deaths/day are likely to be increasing gradually over the next four weeks. Granted, Minnesota is not itself in such bad shape, but with test positivity so high in large states like Texas, Florida, and Arizona, and with people moving in and out of those states around the US, it's hard to be very optimistic about anything being well controlled in the near future.
By August 10th, I think the US will be up to 169K+ total reported deaths and 1900 new deaths/day. That's lower than the peak of 2100 new deaths/day we saw in mid-April when NYC was the epicenter, but potentially 1900/day is an underestimate given that the test positivity is so high across large portions of the south--when the test positivity has been high, we have seen a higher ratio of deaths to confirmed cases. Once COVID-19 gets past 169K deaths in the US, it will virtually be assured of being at least the 3rd-leading cause of death in the US for 2020, behind only heart disease and cancer, and that would be with roughly 5% of the US having been infected (symptomatic or otherwise).
5 percent positivity is a dream for a lot of states right now.
Some guy publicly complains about (and threatens legal action against) a restaurant because he believes it is discriminating against him because he wears pro-Trump clothing. In response, said restaurant airs its dirty laundry about guy to City Pages. Whatever, who knows who is in the right.
But said guy won't stop commenting in the comments section of the online article. It's like watching a car accident!
Holy bananas. Someone actually offered him legal advice that his lawyer should have —-> don’t comment about the case on social / media.
Well that was a comment section full of barely literate writing....