Exponential deaths

The 1Point3Acres COVID-19 in US and Canada / real time updates with credible sources is a great website; probably the best place to track how the pandemic is growing in the US. While it has great visuals, I firmly believe in playing with / plotting the data yourself to feel like you really understand it.

As of writing this (3/22), despite the social distancing measures and shutdowns happening in a lot of places, it’s still clear that the cases are still growing just as exponentially as it was a week ago (we won’t see if the social distancing measures are working for some number of days after they’re enacted). You can clearly see it rising exponentially when you plot the data on a log scale, where it essentially forms a straight line.

With that linear pattern (in log space), you can make a pretty simple prediction of what the numbers will be if the trend continues. But while the log scale is great at showing just how predictably the cases / deaths are rising (in an exponential manner), it’s a bit hard to fathom what that actually means in real life. That’s why I’ve decided to write out how many people will die in the next week, or the next two weeks, if this trend continues. The numbers are quite harrowing (see below).

I’ll keep updating this plot as long as the trend seems to hold. But wow. I really hope the social distancing and other societal measures really break this trend, because the future is VERY bleak otherwise.

3/23/20: The numbers two days ago were 26899 cases and 348 deaths. The model predicted that yesterday we’d have 35805 cases and 461 deaths. Actual numbers were 35229 and 472 deaths. Pretty close… Based on my ultra simple model, the prediction for the end of day today is 47247 cases and 633 deaths.

3/24/20: Actual numbers for yesterday were 46451 and 597 deaths, so definitely up from the day before, but not quite as much as expected based on the last two weeks of exponential growth. Perhaps a good sign that spread is slowing a little? Predictions for end of day today are now 62551 cases and 792 total deaths.

3/25/20: Actual numbers yesterday were 55258 cases and 802 deaths. So the number of cases was much lower than estimated, but deaths were slightly higher. I had actually estimated the numbers based on cases (and extrapolated deaths based on a constant case fatality rate), but it’s possible CFR is going to start changing soon, so I’m going to estimate both numbers independently. Estimates for tomorrow are now 74258 cases and 1054 deaths by end of day today. Oh, and here’s now a plot of daily case fatality rate, to track how that may be changing over time…

3/26/20: Actual cases yesterday was 69222 (~5000 less than I predicted), but deaths were 1055 (spot on with my prediction). So it really does seem like rate of reported cases are falling (maybe b/c we’re starting to catch up a bit more with testing?) but deaths are still growing in an exponential manner. Predictions for end of day today are 89976 cases (likely going to be another overestimate) and 1375 total deaths.

3/27/20: Actual cases yesterday was 86045, and deaths was 1304. The cases has topped the total cases in China and it’s going to keep climbing, though yesterday’s cases and deaths are less than the estimate (so hopefully that’s a good sign?). Next estimates are 109,690 total cases, and 1701 total deaths.

3/28/20: Actual case total from yesterday was 104848 and deaths 1715. Deaths again below predicted, but deaths again spot on. There has definitely been an uptick in death rate over the last 4 days, and it’s not an anomaly like earlier in the epidemic b/c the n values for each day are now in the hundreds each day. New predictions are 131,271 total cases and 2234 total deaths at the end of the day.

3/29/20: Actual case total from yesterday were 124,626 cases and 2202 deaths. So I think it’s quite clear at this point that our rate of new cases is dropping off a little from a week or two ago, but are deaths haven’t really done that. part of that may be the fact that deaths are more delayed than cases (since the order is case -> death). It’s also possible that with the health care capacities straining, the death rate per person infected is actually increasing. New predicted totals for end of day today is 153854 cases and 2844 deaths.

3/30/20: Well, good news is that both the rate of new cases and new deaths is dropping off of the exponential. Total cases as of yesterday was 142629 and total deaths 2578. Estimates for today are 173129 and 3299.

3/31/20: Yesterday was 164671 and 3180. So still fewer new cases, but plenty of deaths. Will be nice to see the death numbers start decreasing in pace… new numbers are 197504 and 3999.

4/1/20: Yesterday was 189802 and 4086. So again a fair number of fewer new cases than expected, but MORE deaths than expected. So uh…. that’s still not a great sign. New predicted numbers are 223917 and 5102. We’re in the phase now where each day is estimated to bring about a thousand new deaths…

4/2/20: Yesterday was 216762 cases and 5148 deaths. Same overall trend; it’s good I’m still overestimating cases, but bad that deaths have been slightly higher than predicted. New predictions are 252157 cases, 6421 deaths.