COVID-19 two week forecast

How many active cases will there be
based on the recent daily growth?
Sections on this page:
 

What's going on here?

Getting through COVID-19 with minimal loss of life will come down to our ability to not-overwhelm our healthcare workers ("flattening the curve"). My goal is to help communicate two things related to achieving this goal:

  1. Make it easier to keep track of the recent growth rate (RGR) in active cases (not total historical cases) for each country. This is the best single-number to watch since it:
    • forecasts the total active cases we're going to have in the future, and
    • informs us to what degree we are heading in the right (or wrong) direction.
  2. Provide a detailed model of how the numbers change to help illustrate:
    • How small-ish numbers get so big, so quickly.
    • Why this is going to be a long fight.
    • Why taking the important (if inconvenient) steps to physically isolate from each other, and wash/sanitize things that come home from outside, is not an "over-reaction" but absolutely required.
Read more about why reducing the recent growth rate is so important

As long as there is any growth in active cases, we move the healthcare system closer to its breaking point. If that happens, the mortality-rate will climb sharply, since no help will be available even for those in need of critical care.

It is hard for most to understand how quickly this virus is spreading because few of us ever encounter compounded daily growth like we're experiencing with COVID-19. This simple forecasting tool is designed to visualize what the total cases will be in just two more weeks, using the recent growth rate.

Even when best-efforts to contain the virus are working, the growth-rate in active cases will only decline slowly (what is referred to as the "optimistic growth rate" in the forecast detail-view.

 

How accurate are these numbers?

Robust modeling tools exist for predicting how a well-understood virus spreads, but this virus is not yet well understood, and those models are hard for most people to understand. The goal of the site is to help people understand approximately what's going to happen, in order to reduce the chance of leaving them puzzled about precisely what's going to happen.

Using the recent growth rate* to plot our future at that rate is meaningful in terms of communicating why we need to get the current growth rate down. The model used here is not for our cherished epidemiologists; It's for everyone else.

Note that it's possible for a country to add new active cases without any growth in active cases, if it had an equal number of people move into the 'recovered' category. I'm focusing only on total active cases, because this is the number that forecasts how burdened the healthcare system is.

* The Recent Growth Rate is the 7-day average (mean) in the daily growth of active cases over the most recent 9 days, dropping the highest and lowest values over that period. This smooths out anomalous spikes or drops in the daily numbers, caused when countries have to suddenly revise their numbers based on new methodologies or testing.
 

Show me the mathy!

The most relatable place most of us would see agressive compounded growth like we're experiencing with COVID-19 might be an unpaid credit card bill. Once the interest is charged on a balance, the next time interest is calculated it's based on the previous balance PLUS the interest that was just added. It's this "interest on interest" that creates the "compounding" effect that drives the numbers up so quickly.

Description Change Cases
Starting value - 1,000
1st-day growth of 26% +260 1,260
2nd-day growth of 26% +327 1,587
3rd-day growth of 26% +413 2,000

In the first table, we start with 1,000 cases. On the first day the cases have grown by 26% (not an uncommon growth rate for this virus, even when countries have started taking serious measures to curb the disease) resulting in 260 more cases. But on the 2nd day, it doesn't just grow by 260 again; This time it grows by the same 260 plus 26% of the-1st-day's-growth of 260 (67) bringing the 2nd day's increase to 327 and total active cases to 1,588. Growing 1,588 by 26% one last time, takes us up to 2,000. A doubling in just 3 days.

Description Cases
Starting value1,000
1st doubling (@3 days)2,000
2nd doubling (@6 days)4,000
3rd doubling (@9 days)8,000
4th doubling (@12 days)16,000
5th doubling (@15 days)32,000

BUT WAIT THERE'S MORE! That was just the first doubling. Doubling itself is counter intuitive and is going to happen every 3 days when the growth rate is 26%, as we just saw.

If we start with 1,000 cases and double them every 3 days, in just over 2 weeks we've got 32x as many cases as we started with. That is why it is so critical to pay attention to the growth rate, even if the number of cases currently seems manageable. It's also why even single-digit growth rates can still be "bad" since even they can quickly lead to an unmanageable number of cases.

 

Contact

Hi! I'm not sure how busy this site or I may become, but I'll try to get back to you in a reasonable amount of time.