Death, on a pale horse (revisited)
Note from Tom: I put up this post on March 28. I didn't know it then, but that day turned out to be the peak of the 7-day growth rate in deaths. The rate was 640%, meaning total deaths so far were almost doubling every day. Had they continued at that rate, the entire US population would have been dead by early May. This ended up being (I think) my most-read post. You don't need to ponder the details too much; what drew people to this post was the last four paragraphs.
With the huge increases in new Covid cases and hospitalizations lately, we're at a similar point now. What will keep the great majority of the population alive is the fact that the ratio of deaths to closed cases is now 3.6% vs. 40% on March 28. That being said, we're in for a horrific next few months, with deaths perhaps reaching 400,000 by Biden's inauguration. The big difference between today and March 28 is that we all know very well what we should be doing. But we're not doing it. Starting at the top.
- Going forward, I plan to grow the total cases number by the average of the 7-day growth rates over the past 7 days. Of course, total expected deaths over the pandemic is directly tied to total cases, since expected deaths (as of day X) equal total cases (as of day X) times the case mortality rate (i.e. the number of cases that end in death – this number can only be known after the pandemic is over).
- But when I put together the spreadsheet this morning, the projections based on the 7-day average of 7-day growth rate in tests (785%, based on yesterday’s figures) were too horrible to publish here, and I allowed myself to be swayed by the fact that yesterday’s 7-day growth rate in cases was the lowest since the beginning of March (when my data begin). That rate was 438%, while previous 7-day rates were between 522% and 839%. So rather than use the 7-day average of the 7-day growth rates, I used yesterday’s number, on the theory that maybe the limited social distancing that’s been put in place in a number of states was beginning to have an effect. Hopefully, the numbers will bear me out shortly, and this trend will continue.
- I’ve been using 3% as the case mortality rate so far, but that may be too low for three reasons. First, the rate worked out to be around 3.9% in China, which is more or less at the end of the pandemic. Second, if you look at the data so far (at the bottom of my numbers above), the mortality rate for closed cases so far in the US is 40%. This is presumably artificially inflated by the fact that closing a case when the person is still alive will normally lag a long time, since no doctor will want to close a case and then have it turn out the patient was still sick. It will come down, but will it come down to 3%? 4%? 10%? Too soon to say. Third and most importantly, the mortality rate is very dependent on the degree of hospital overcrowding – shortage of beds, shortage of ventilators, shortage of masks for the staff members, etc. In the next two weeks, as the full onslaught hits NYC, we’ll see how serious this overcrowding and lack of resources turns out to be.
- I was previously publishing a number for estimated deaths across the pandemic, based simply on the day’s case number times the 3% mortality rate. But I realized this morning that this is a cop-out. It assumes that, by some truly miraculous occurrence, there will be no new cases after today. Of course, given the number of people out there who have the virus but haven’t been tested yet (either because they haven’t showed any symptoms, or because they’ve tried to get tested and haven’t been able to), this is completely unrealistic.
- My new realistic minimum number assumes that as of the day in question, a total lockdown on the country is imposed – in fact, every member of a family is locked down separately from every other member of that family (as eventually happened in Wuhan, since whole families ended up infected before they did this), and the lockdown works perfectly. Given the 14-day incubation period, this means that by the end of 14 days, the authorities will have identified every case anywhere (because it will have shown symptoms or else it will have gone away). They can then lock down each of these people; they don’t have to trace contacts, since after the person has been isolated for 14 days, they won’t be able to spread the virus because they don’t have it – at least that’s how it works in theory. In other words, everything goes perfectly. This is of course not realistic, but hopefully a close-to-total lockdown would lead to a close-to-total identification of all current cases. We’re currently far away from being able to do either of those, of course.
- So I get my new estimate for total cases during the pandemic by finding the total projected cases 14 days later; total deaths during the pandemic are 3% of that case number. This means that, by my previous method, I would have said total expected cases given yesterday’s case number were 3,128. By the new method, total expected deaths as of yesterday are equal to the projected case number on April 10 (2,000,089) times 3%, or 60,003.
- For the first time, I’ve started projecting actual daily deaths (i.e. the number you hear on the news – it was 1,704 yesterday), not just deaths over the course of the epidemic. I previously didn’t do this because the only good way to estimate it would be based on medical knowledge of how the virus will progress in each person, etc. But nobody has this knowledge yet anyway, so I decided to grow the daily reported number by – again – the 7-day average of the 7-day growth rates. This yields an estimate for each day going into the future.
- As I already said, just using the actual case figure from yesterday and the projection based on a lowballed growth rate, the minimum number of US deaths over the course of the pandemic are 60,000. And that’s based on the assumption that the entire country was locked down as of yesterday, which obviously didn’t happen. So 60,000 is the bare minimum number of deaths we can expect going forward. This is 2,000 more than the total number of Americans that died of all causes in the Vietnam War. Note that actual deaths reported yesterday were 1,704.
- Let’s say we wait until April 3 (next Friday) and do our total lockdown that day. Then, total deaths over the course of the pandemic are 262,812. However, the projected deaths on that day are 2,150.
- Now we’ll go to April 10 (two weeks from yesterday), and start the lockdown that day. Then, total pandemic deaths are – are you sitting down? – 1,151,115. Just as horrific, on that day alone, reported deaths will be 11,530.
Comments
Post a Comment