Saturday, 7 March 2020

A frightening model


ALERT: Pandemic projection 

model shows 2.16 million 

deaths in USA by July 4th if 

nothing done to halt nearly 

all domestic travel (air, rail 

and roads)

6 March, 2020


Today I finished tweaking the first draft of a pandemic projection model that simulates the spread of the coronavirus in the United States. The assumptions of the model are explained here, and you will find they are extremely conservative (using R0 value of just 1.82, for example).

The model’s predictions are nothing short of apocalyptic if the virus is allowed to spread without restraint across the United States. According to the model, there will be 2.16 million dead Americans by July 4th if domestic travel is not aggressively halted very soon (see the full projections below). This is not a prediction, since I believe that state governments and the federal government will intervene long before July 4th to declare, essentially, medical martial law.

That is now the only remaining option to stop the spread. It must be done soon. It may already be too late to prevent millions from getting infected.

Before we get to the conclusions of the model, here are the assumptions. As you will quickly see, these are very conservative assumptions:

– R0 value of 1.82, spread out across 14 days. In other words, every 1 infected person spreads the virus to another 1.82 additional people over the course of 14 days. This addresses the “velocity” of the spread. The daily velocity here is 0.13, meaning 1 person spreads to an average of 0.13 additional people per day. (Over 14 days, that’s 1.82 people.)

After 14 days, a 98% recovery rate for previously infected people.

No “recurring infections” for previously infected people.

2% mortality rate among those who are infected. The WHO currently says mortality is 3.4% but we are going with a lower number to keep these projections more conservative. This model assumes they die after 14 days and then stop spreading once dead.

Under this model, the number of total aggregate infections roughly doubles in a little over 7 days.

We assume a newly infected person become infectious after 24 hours of being infected and remains infectious for the full 14 days, then stops being infectious because they are either immune or dead.

Those who recover or are dead are removed from the “actively infected” pool so that the simulation only accounts for those who have acquired a new infection within the last 14 days.

We are aware that 14 days may not be the final correct number, but it is a working number for the current draft of this model. This model will be updated with new data over time.

The number of newly infected is calculated daily. In essence, this projection model is “compounded” daily.


Caveats and possible weaknesses:

This model is still in a rough draft form and will no doubt be improved as additional data become available. The model is not yet peer reviewed and should not be cited as predictions or facts.


The 2% mortality rate used in the simulation has a lag time of 14 days, so if you look at the raw numbers, it doesn’t appear to be 2%. That’s because we assume it takes 14 days for infected people to expire.


The model assumes no strong intervention from anti-viral drugs, vaccines, government quarantines, medical martial law or similar reconfigurations on social mobility. Thus, this model predicts a scenario where nothing is done to contain the virus. In reality, this model is therefore not predictive for the real world over the long run for the simple reason that such strong interventions are inevitable. (In other words, state and national leaders must intervene or everyone will eventually get infected.)

Here’s what the “Adams model” shows: 2.16 million dead Americans by July 4th if strong travel restrictions are not put in place

As with every exponential growth curve, the numbers appear quite small at first, then suddenly explode as time goes by.

Additional notes / disclaimers:


The are NOT predictions. These are mathematical projections that assume no strong interventions such as martial law, travel restrictions, social isolation, etc. These numbers are nowhere near “worst case,” however, since they only use an R0 value of 1.82 (the real virus is believed to be closer to 6.6).

If the numbers don’t seem to “add up,” that’s because there’s a 14-day lag time for the newly infected. Only after 14 days are they either alive and cured (98%) or dead and non-infectious (2%). But during the 14 days they are “actively infectious” and spreading the virus via R0=1.82.

We will update these projects as the model acquires new data. We are calibrating the model based on the number of reported deaths, which we (foolishly) assume is accurate. This assumption may be false, meaning the reality of the situation may be far worse than this model predicts.

ADAMS pandemic model, Revision 1, using data up to March 5, 2020 to calibrate projections. Geographic scope is the United States of America only (doesn’t include Canada):


By April 4th, 2020, if no travel restrictions are put in place:
  • 8,645 actively infected and transmitting on this day
  • 580 cumulative dead since day one, with 58 deaths on this day
  • 5,432 cumulative recovered since day one
By May 4th, 2020, if no travel restrictions are put in place:
  • 122,529 actively infected and transmitting on this day
  • 10,432 cumulative dead since day one, with 910 deaths on this day
  • 85,332 cumulative recovered since day one
By June 4th, 2020, if no travel restrictions are put in place:
  • 1.7 million actively infected and transmitting on this day
  • 153,000 cumulative dead since day one, with 12,960 deaths on this day
  • 1.2 million cumulative recovered since day one
By July 4th, 2020, if no travel restrictions are put in place:
  • 24.3 million actively infected and transmitting on this day
  • 2.16 million cumulative dead since day one, with 183,000 deaths on this day
  • 17.1 million cumulative recovered since day one
From here, the numbers only become more apocalyptic and irrelevant, since no governor or president will allow this exponential growth to continue unabated. For example, if allowed to run without restraint, the simulation says the number of infected Americans by August 5th will exceed the population of the entire country, which is obviously an impossibility since it would run out of people to infect first.

The upshot of this model is that you can now safely assume there will be severe travel restrictions put in place across America.

The chances of that occurring are as close to 100% as you can get, barring some miraculous intervention such as a spontaneous cure or act of God. Because if travel restrictions are not put in place, America falls.


No comments:

Post a Comment

Note: only a member of this blog may post a comment.