General News
April 3 Update: How modeling helps
- April 3, 2020
April 3, 2020
Hannover, Germany
Good morning,
Yesterday I had a conference call with Ed Liu from Jackson Lab, Darron Collins from College of the Atlantic, and Art Blank from MDI Hospital. We discussed the present situation on MDI and in Maine. At the hospital everybody is well prepared, and it is currently quiet. College of the Atlantic is preparing for and transitioning to digital education. Jackson Lab has established an impressive new testing site for COVID-19 in Connecticut. We talked about the problems and obstacles of getting accreditation for a new test. Hopefully, it will also be introduced in Maine. Such a test system will become very helpful when we return to normal.
The discussion moved rapidly to the big question: how long will it take before we can return to normal? Numbers were exchanged and critically discussed. It is obvious that our predictions depend on the data we get. Everybody is using the data available to calculate outcome and time span. A very good website for information and prediction models, viewable by state, is from the Institute for Health Metrics and Evaluation site: IHME COVID-19 Projections.
It’s important to realize that the numbers being used are not data from a laboratory experiment. For example, take the numbers of infected people. The reporting of infected people is mandatory and fairly accurate. However, these numbers critically depend on how many and which people have been tested. If we look at the daily increase of infected patients, we should also be looking at how many people have been tested. Otherwise, a reported increase in infected people may only reflect that many more people have been tested. It will be of considerable interest in the future to know about infection with the virus in people with no symptoms. Presently, asymptomatic people are not tested in most countries and we have no idea how many asymptomatic people are carrying the virus. One research project has started in Germany to look at this problem in an isolated village with many cases of COVID-19 that occurred after a local wedding with most of the villagers being present. Although I am very much interested in such research, it is good that we have not seen a scenario such as this in Northeastern Maine yet.
It is also interesting to analyze prediction models with regard to the death rate – comparing the actual numbers with predicted numbers. In most states, the death rates are small and highly variable. These numbers do not reflect a gradual increase in the population but reflect mostly local “epicenters”. As an example, in the city of Wolfsburg near Hannover (where Volkswagen is located), seven elderly patients died in one nursing home within the last couple of days. This incident pushed up the daily death rate for the state considerably and gave the impression that in general more people are dying in our state.
On the other hand, is it very reassuring to work with these models. Compare actual numbers and predicted numbers for today and when you look carefully it is very good to see that the predictions are worse than compared to the actual numbers.
There is light on the horizon. I wish you a sunny weekend.
Best,
HH