Covid-19 Data Sites
We are literally swimming in a sea of confused data about the Corona Virus data sites. Net, net, the growth in death rates is key. Eventually, hopefully, deaths will start to shrink. See UMN for a collection of links or just use Google. Here are the three sites, you should monitor every day if you are interested
- 91-Divoc. Ok, this is Covid-19 spelled backwards and it shows with time how the infection is spreading in terms of growth rates. Right now the US is leading the way in terms of the number of cases and growth rate as well for developed countries (and actually for most countries now not just developed).
- New York Times. Amazingly there is no county by county count in the United States, so the New York Times is trying to do this.
- Healthdata.org. This updates every Monday so you can see how estimates march actuals by comparing with 91-divoc. I could stare at these charts all day long, they show what the (very optimistic, see below) expected rate of infection is given current countermeasures and more importantly when each state runs out of ICU beds. Pretty scary. The main problem with this model is that it assumes that implementing measures has the same effect as China, that is probably not going to be true. Certainly not true for Italy and not true for the US. Also, this model as all of them starts with very little data, so expect the curves to change dramatically literally day by day since it is a “curve fitting” model. It’s also not updated regularly
- Washington State Hospital Association. This site is good because it shows you the historical death counts. Not good news to see the last three days has seen that grow from 15 then to 9 then to 27. Could be some reporting issues, but the numbers this weekend are going to really tell an important story. It lets you playback on the state map deaths both cumulative and new each day.
- Atishb. This site is really nice because it shows the growth rates of cases as a point that is moving from time, so it is way easier to see how things are working. It shows how dramatically better Korea did for instance. It caught it early and pushed the curve down quickly. These are just confirmed cases, but it also shows how the US has just rocketed so far ahead in total cases globally. And that this growth is accelerating.
If you have more time, these are additional good charts.
- Stat news. This is a really beautiful chart set (grim, but well done). You can use it to see the new deaths for each country and also for each state in the US. The death rates are going bounce some for states because there are so few, but you can get an idea of the trends. Washington just spiked in deaths by 3x from yesterday, so hopefully, that is a reporting anomaly and not a whole new level of death. It even allows a county by county view which is even more granular.
- Johns Hopkins. This is a nice overview of the current state, but it’s hard to see the growth trends.
- Ncov2019.live. Written by a high school student in Mercer Island. Pretty cool data, but not split up enough to really predict what will happen.
- Flourish. Covid is now the no. 3 cause of death in the US! Sadly at 1,774 deaths per day at heart disease, we are 554/day now so at current rates we will that in a week 🙁 as of March 28.
- CT Bergstrom. UW Professor in Biology. His debunking is statistical inaccuracies is incredible. Great for understanding what’s true and what’s not.
Then some good backgrounders
- Bill Gates on the whole issue in a town hall and this was all going to happen.
- South Korea. I realize that this is a very different country but in the spirit of learning something. This is a must-watch.
Analysis of the above as of 27 March
- Deaths vs. Cases. Most places, particularly in the US, are undertested, so we don’t know the real rates of the disease. As an example, today in the State of Washington, the percentage of positives for testing is 7%. That is 7% of the people tested are showing positives. That seems super high when the estimate incidence is 1-2%. This probably means we are testing very few people (3-5x lower). It also means we are missing cases as well.
- Death rate growth as a more solid metric. Since we are missing cases, death rates are a better way to see what is happening, although some governments are not counting deaths as Corona or just not reporting them. Right now that rate is estimated at 3% death rate (so 20-30x higher than the normal flu). The main issue with death rates is they are lagging indicator as it takes a week for the virus to become symptomatic and then, of course, the disease has to run its course. That any preventative action you take will take weeks to show up in the statistics. And of course, decreasing the growth rate is the first step, then you have to actually lower the deaths hopefully to zero.
- Grow rates, doubling times and log scales. This is a disease that is growing quickly, it has an R0 (R-nought) of over 2 (that is over two people get it for every person infected), so you need to look at growth rates. This is sometimes couched as doubling time. So if a disease is growing 33% a day, then it is doubling every three days. That’s super fast. That means that you can not be too concerned at the early parts of an epidemic, so people use a logarithmic scale, so you can see the early parts quickly.
- Smoothing the curve and overwhelming the system. This is one of the big concerns, a big spike in cases, means you run out of intensive care beds and ventilators. That will spike the death rate quite a bit, so that’s why having measures in place matter.