Latest COVID-19 statistics

0

Well, the core news sources I use show that the tail down in Washington State is happening. Even Uncle Dave is talking about it and California is tailing down as well, so we are to the “dance” part while we wait for an effective treatment or a vaccine.

For basic reading, there is an absolute deluge of documents, but I’ve found Tomas Pueyo to be great at distilling things down. He has a series of articles that you should read in order about the epidemic and how to manage it:

  1. Coronavirus, The Hammer and the Dance: The strategy every country
  2. Coronavirus: Out of Many, One: A deep dive in the US
  3. Coronavirus: Learning How to Dance: How the best countries are fighting the epidemic
  4. Coronavirus: The Basic Dance Steps Everybody Can Follow

The Status of the Pandemic

91-Divoc Time Series Historical

91-Divoc.com. The best site for looking at the infection so far. They give a nice timeline and normalized by population view so you can see how folks are doing relative to the infection.

For Washington, make this happen, these are confirmed deaths over time and as they flatten, businesses are starting to reopen. The question is will this be like Hokkaido where reopening caused uncontrolled infection or like Singapore where there is reinfection but… As an aside, this uses an iframe to show the image, so it looks a bit funky

And the US still has infections but this is also flattening and reopening is in the wind as well so deaths are finally starting to trail down.


CDC Model Ensemble

This is a list the CDC has of models that they consider decent. Now Washington State has a Dashboard that includes some of these as well. They have their own models which include all the data in one. But there are quite a few with different methods. This is called ensemble modeling and shows what many different models are showing. As an side, the models are very broad but they do include 95% confidence interval. A straight averaging probably doesn’t make much sense but the best guess for the US as of May 2 is that by end of May, we will have 75-100K deaths. Ugh.


Gu SEIS Model

From Gu Youyang, there’s a machine learning forecast that also uses the SERI/SEIS model happening that is another form of curve fitting. It’s a different view. It uses a model of infection called


DELPHI

This is an Delphi open source model done by MIT Operations Research led by Dimitris Bertsimas using SEIR and shows the predicted counts and so forth on a given day. There doesn’t seem to be an easy way to embed this one. it is not a sanctioned model like the Gu YouYang one or Columbia, but has about the same data.


Columbia Model

This also is an Epidemiological model but it has a big tunable parameter which is the effect of social distancing. The main effect here is to model when hospitals run out of capacity (as happened in Italy), when that happens, the CFR jumps to 10% so flattening the curve is all about that. For low effects like 20% social distance, you have different surge responses. That is effective in dealing with infections. This is a GIS model, so you can toggle. The most interesting thing they don’t have which is differential social distancing effects as the states are non-uniform. But to look at a worst-case try Hospital capacity exceeded in 42 days with no social distancing.

What you see is massive hot spots in rural Washington, rural New York, the South, and the midwest.


UW IHME Future Curve Fitting

It’s definitely sobering when 2,000 people a day dying is considered good and even normal. The IHME figures which show death per day. Of course, this is all dependent on the containment measures working. As you can see there’s a huge error band for deaths that depends a lot on containment strategies. The main issue here is that it is using curve fitting so it has to make many assumptions about the future. Let’s hope that happens…

Note the IMHE model is no longer on the CDC site and has some method problems since it really doing curve fitting rather than using an underlying infection model


John Hopkins Dashboard

Finally, Johns Hopkins has the best current status dashboard and is the basis for the visualizations in 91-Divoc and the global view is pretty sobering 210K deaths overall and climbing.

Related Posts

This site uses Akismet to reduce spam. Learn how your comment data is processed.

© All Right Reserved