Saturday, May 9, 2020

COVID-19 Modeling and the Future

I hit the modeling topic in the prior blog.  But I've also had more time to think about it.  The IMHE models at Healthdata.org were frozen (unchanged) for about four days, from a couple of days before May 1 to a couple of days after.  When they were frozen, the graph for Texas was showing a bunch of up and down bounces in the past data, just like you would expect.  The death rate was low enough that daily variations showed up distinctly.  Once the freeze stopped, those bounces disappeared.  The data was now smooth with one exception.  On May 1 there was a minimum and then a sharp rise thereafter.

So a death rate that is at least 14 days behind the curve in infections, decided to jump up the day that Texas began reopening.  The same thing happened with the US mortality.  I didn't actively follow the other states, so cannot say what happened there.

So either IMHE drastically changed their model or their data.  My suspicion is they did both.  They used to use actual data for plotting the curve in the past and projected data for plotting the curve in the future.

It looks to me like they've now fit the past data into a 'best-fit' curve of low order (high order would have more inflections).  So you cannot see what real data is actually being reported or recorded.

Since they showed an inflection point the day reopening began, I suspect they jumped the infection rate (R) value significantly and they did it at least two weeks before May 1.  Either that, or they artificially changed the model to reflect an immediate rise in mortality on May 1.

To me, they are lying scum.  Maybe whoever was providing their data changed it, and I should apologize.  But I won't.  They should have made a big statement on their prediction website with a big warning icon.  Maybe it's in the FAQ, but I won't look.

They could have continued to show their previous model output as it evolved after reopening so we could see how the reopening actually affected the data and their model.  They did not do that.  Instead, they drastically changed their model and data on the reopening day.  How can anybody trust them after that?  I WILL NOT.

On the future.  I don't believe it matters now whether the infections and mortality rate rise.  People and businesses cannot continue to survive without income.  The feds cannot continue to print money without causing a financial failure.  States do not have the income to continue to provide services.  The economy will reopen, and it will not close again.  

The economy will be sputtering for a while.  States are limiting opening by social distancing and capacity.  That will hurt most businesses.  And many people will still want to hide at home.  The mainstream media will still be trying to scare the public to help Joe Biden beat Donald Trump.

You may see a few cities or counties go into lock down again if their infection and mortality rates decide to jump up.  But state wide, people won't take it.  I'm not even confident that if a city or county opens, for more than a week, they will be willing to go into lock down again unless it gets really bad.

Unfortunately, I don't think people will wear masks in places where it's not required.  Especially if they are mad at the government for other reasons; and a lot of people are mad at the government.  And most especially if they do not feel at risk.  And risk is kind of limited to those over 65 and those with multiple comorbidities.  The vast majority of people believe they have only the risk of a bad flu.

So like my simple math predicted, I think the virus will continue to spread.  I think the hospitalization and mortality rates will go up some, but not a lot.  It will spread slowly to the states that weren't really infected.  The cumulative death rate is going to continue to rise nationwide, but we won't see dramatic jumps like we did with New York.

I see little chance of highly effective new therapeutics or vaccines.  We'll have to get to herd immunity the hard way.  Keep those high risk people safe!

No comments:

Post a Comment