So I linked to one important source I used in my non-professional but (hopefully) mathematically literate guesswork about how COVID-19 deaths in the US will increase over the next while. What I didn’t appreciate at that time was that the graphic that they used to discuss geometric growth in deaths in various countries is a dynamic graphic. Every time a country updates its total deaths on its own official websites, which happens once a day or more for Eurpoean and North American countries, the graphic itself updates. This means you can’t click on the link from a past article (or in this paragraph) and get easy access to the numbers I used when writing a particular blog post.
When I linked that article, it showed total deaths in the USA had doubled over the previous 4 days. But two days past and the graphic is now different. Realizing that I couldn’t use that as a stable record of my thinking, I have decided to take screen shots to have a record of how well the death doublings track with my guesses about how the speed of those doublings will be affected by changing quarantine orders and individual behaviors.
Because stay-at-home orders take 2-3 weeks to show up in death rates, and because Washington, NY, and California have been on stay-at-home orders for a while but large states like Texas and Florida hadn’t issued them until last week, I suspected that death doublings would continue at the same rate until about the 15th to 19th. But I’ve been checking in to see if maybe the states that issued early stay-at-home orders haven’t had their effect yet. Maybe there will be a change in doublings before the red state holdouts begin clamping down on transmission risks?
What I found is that after 2 days the doubling rate is very nearly identical. Total deaths were multiplied by 2.1 over the last 4 days of CDC data. You can see that in the following snapshot of the chart, after clicking the circle-i icon next to the US data:
COVID-19 death totals for select countries as of April 5, 2020
Death-doubling rate detail for USA COVID-19 deaths as of April 5, 2020
For the scenario I thought likely (a largely steady doubling rate for the next 12 days, then lengthening slightly to a 5 day then a 6 day doubling period) based on when stay-at-home orders were introduced, the USA would hit a quarter million deaths before the end of the month. Obviously at that point, even if we greatly extend the doubling period, any doubling is a lot more people we lose to death. Also, if you’ve been tracking Trump’s predictions, you know he said the US should expect about 100k deaths and no more than 240k deaths. Hitting the quarter million number on April 29, my guesstimate, would mean we exceed his most pessimistic prediction before the end of this month, with many months more to come before we can expect a vaccine.
I hope I’ve been clear to everyone that I’m not an epidemiologist. I’m just using the publicly available data to track what seems reasonable to me. We’re not doomed to double US deaths three times from April 3rd to April 15, I’m just following what the disease experts say the COVID-19 infection-to-death lag is supposed to be and reckoning that we can’t get any help from the final red state stay-at-home orders until at least then.
This continues to look really, really bad.
I continue to believe that we’re in for a million deaths or more, but all those predictions change if the behavior of US residents changes. More masks, more hand washings, more care when helping others (like leaving packages on a doorstep and backing away instead of bringing food or other necessities into the home of someone we’re trying to help), fewer businesses open, fewer non-mandatory meetings like book clubs and shabbat services: these are all reasonable, possible, incremental changes that can slow that doubling rate. I don’t expect them to have much effect before the end of the month, though.
Why? Because I believe that there are a large number of very stubborn people in the US determined to ignore the risk, some of them for political reasons, some for religious reasons, some for economic or other reasons. I do think that eventually deaths will be so common that if you don’t have one among your immediate family and close friends, you’ll have a friend who does. When we get to the point where everyone knows someone who is grieving I think most resistance to social distancing will collapse. Now, these aren’t predictions that can be checked against math. I don’t know how hard core the resistance to distancing really is, and I don’t actually know how long it will last. I can’t be certain it will stay strong until grieving becomes ubiquitous.
What I can do, though, is keep checking in with that CDC website and provide you an update every few days on how close things are tracking to my original prediction based on April 3rd numbers. In another 6 days we’ll have a good idea about whether 3 doublings by April 15-16 will be borne out. By April 20th I think we’ll be able to be fairly sure whether or not we’ll hit a quarter million US dead by the end of the month. So I’ll keep checking in with the website. If appropriate, I’ll keep updating my guesses about the toll on US residents.
I wish I didn’t have to do this, but it’s obvious that the federal government isn’t releasing updated projections based on real computer models tested against past outbreaks and created by people far more knowledgable than I. So I’m going to keep on updating the numbers, keep on massaging my spherical cow, and keep on hoping that my ignorance of epidemiology really is leading me to dramatically overestimate the ultimate cost to human life and health and happiness.
May you stay safe and healthy.
I hope you are wrong, but I’m pretty sure you’re right.
I discovered yesterday that someone I met briefly has died, my third cousin once removed’s* father-in-law died of COVID-19 last Wednesday at seventy-four; adding to the tragedy is that my TCOR gave birth after he had been diagnossed, none of that side of the family have been able to see young Ezra Atticus, and his father can not go to his father’s funeral unless he then stays away from his wife and son for two weeks. This is the sort of story that may change stubborn people’s minds, but far too many will die before someone close enough to them dies to resonate sufficiently to change their minds.
*I don’t have any nearer relatives on that side of the family, and the several generations have all got on very well, so contact has been more than maintained.
The people who are going to die in the next couple of weeks are already infected. That’s already happened, and there’s not much we can do about it now. They are, effectively, already dead. They just don’t know it yet.
The real worry is whether the number of cases requiring ICU treatment starts exceeding the available ICU capacity, because that’s when the case fatality rate starts going up, as people who would have lived if they’d received adequate treatment start dying too.
If you’re interested in yet another data tracker, I have started following this one from the University of Washington: http://covid19.healthdata.org/
Featured at the link is a forecast of hospitalizations and deaths. The model predicts that the first wave of infections will peak in mid-April, end by June, and kill between 50,000 and 140,000 people. I think it’s optimistic, because the places where I would expect people to be least likely to take social distancing seriously (mostly the South and the Midwest) are also the places that were hit later.
But is it optimistic to expect deaths to peak at all by summer? Or will they continue to grow exponentially until we reach herd immunity, and deaths have reached the millions?
It looks like they have already started to peak in several European countries, and even Iran. I’d like to believe that we will reach peak deaths here by the end of April, but I’m bracing myself to be wrong.
And then there are those extra factors that can push the needle one way or the other. If we find an effective treatment that shortens hospital stays, we can save a hundred thousand lives. If the virus ends up finding a natural reservoir (there are reports it can infect domestic cats), that can make social distancing substantially more difficult.
One would hope. But when dealing with the snooty and self righteous, deaths may be blamed on “not enough faith”, and “you should have…” statements. They’re capable of it. I am not wishing for it, but it might take the deaths of a few big name con artists…I mean, pastors or half a megachurch to die to get their attention.
invivoMark (#3)
And not just domestic cats. What if this gets into feral and wild populations (e.g. cougars)?
I continue to believe that we’re in for a million deaths or more, but all those predictions change if the behavior of US residents changes.
The whole “flatten the curve” thing is not about keeping people from getting sick – it’s about smoothing out the rate at which new patients show up in the ER. It’s a safe bet that everyone on Earth is going to encounter the coronavirus and in the big picture, a percentage of them are going to die. That percentage can get nudged up or down a tiny bit depending on the availability of advanced medical care, but mostly it’s going to be immune systems doing the work, until/unless there’s a vaccine – and then we’ll have to deal with all the ‘autism’ that results from that. :/
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 shortly if ICUs are not overwhelmed. That is a big if.
Worldometer lets you see total cases and deaths on a semilog scale.
Correction to 6
That should be: The growth rate of deaths should start falling below the projected exponential growth rate shortly if ICUs are not overwhelmed.
@militantagnostic
Well, that’s good news. It’s a couple days earlier than I would have thought, but of course it depends on how MUCH below the regression line we’re talking.
@Marcus Ranum:
Well, yes, in the short term that’s true. In the longer term it has other benefits.
Flattening the curve doesn’t keep people from dying eventually unless it a) allows previous cases to clear ICU beds before they’re needed for the next case, and/or b) it delays exposure long enough that the vaccine comes out first. What you say is absolutely correct about ER resources, but if we can slow exposure long enough to get widespread testing, then we can quarantine fewer people, get a lot of folks back to work, and limit the percentage of people who contract COVID-19 before the vaccine can prevent it.
Now, a usable vaccine isn’t inevitable. Some diseases are simply hard to create vaccines for. But we do manage to get vaccines for a lot of them, eventually. So I’m hoping behavioral changes delay mass transmission until testing availability which delays mass transmission until the vaccine. I can’t be certain of anything, but that’s my hope.
@Marcus Ranum,
No, not everyone on Earth is going to get the virus. With no reaction on our part (i.e., no social distancing), the virus would spread until it slows down due to herd immunity. That depends on the virus’s infection dynamics, which we still don’t have precisely nailed down, but might be around 70% of humanity.
But social distancing changes the virus’s infection dynamics. It doesn’t just reduce the rate at which new infections occur. It also lowers the threshold for herd immunity. Mathematically, the herd immunity threshold is:
1 – (1/R0)
When the proportion of immune + infected is greater than the above expression, the number of infections will decrease. Social distancing decreases R0 (the number of times an infected person is expected to infect someone new).
Herd immunity is our worst-case scenario. That’s what happens if our efforts at social distancing aren’t enough to stop the virus’s spread. Even in this case, social distancing can prevent people from getting the virus.
Crip Dyke @8
On March 27, Canada was 13% below the regression line and US was 19% below the extrapolated regression line (both barely significant maybe*). As of yesterday (April 5) Canada is 66% below the line is US is 81% below the the line. Fractions are distanaces on a log scale. This means Canadian cases are a third of the extrapolated number and US acases are a fifth of the extrapolated number (the US line is much steeper than the Canadian line). Data is from Wikipedia.
* The Candian data was a tighter fit to the line for the exponential period.
I just like to mention another graph (collection) here, also on “hot” (daily updated) data, where time is actually time, not an axis (it becomes much clearer once you look at it): https://aatishb.com/covidtrends/