A. Projection Models
There are basically two types of projections: one based on mathematical models and one based on empirical data. Mathematical models can be done early and provide guidance on mitigation efforts and planning. However, mathematical models are notoriously unreliable.
Projections based on empirical data (actual observations) are more accurate. The problem is that empirical data does not exist at the beginning of a pandemic, and it requires a significant amount of time and data to make a projection. It’s like a vaccine – it may well be the best remedy for a new pandemic, but it comes after the pandemic is over.
There is a combination of both models. That involves analyzing prior pandemics and then comparing that data with early data points in a current pandemic. As more information about the slope and disease progression becomes available the greater the accuracy in the predictions. The drawback to using a combination of mathematical and empirical models is that it takes time to acquire the data.
Death numbers were selected as the primary indicator, rather than infections, because deaths are typically more accurate numbers. Infection numbers would be the best data, if such information was known, but antibody tests showed that the infection numbers (i.e. those tested) were 7 to 85 times lower than the actual number. Hence, infection numbers are not reliable.
Unfortunately, there are some problems with the death rate data.
There is some published data suggesting that CDC decreased flu deaths after week 10 below residual flu deaths reported in previous years and reclassified them as COVID-19 deaths.
In addition, Medicare pays more money for COVID-19 patients over all other patients. This provides an economic incentive to classify hospital deaths as COVID-19 related. As discussed in Question 4 section D, a comparison of the up-slope down-slope ratios for Italy, Spain and the UK, showed that the US hospital COVID-19 deaths could have been overstated by 41%.
This makes the projection model results unreliable.
Figure 10. Actual and projected COVID-19 Deaths for 2020. The horizontal axis is the number of Weeks from March to August 2020. The Vertical Axis is the Weekly Deaths in the US. Data from the CDC.
There appears to be an upturn in the COVID-19 deaths in mid July. The CDC stopped providing information about the influenza Flu Map in May 2020. This is normal because the Flu Map only covers the period of October to May. But, it appears that they stopped reporting any flu deaths and classified them as COVID-19 deaths after May. This is not normal and it is highly suspect since the graph in Figure 5 showed that for a normal flu year during the summer months from all prior years there were approximately 2,800 to 3000 flu deaths each year. For that to stop for no apparent reason, does not make sense. The CDC does not explain why they reclassified flu deaths as Covid-19 deaths or do they explain why their changed their procedures for 2020.
These numbers are small compared to a host of other diseases and events. For example, the number of accidental deaths each year is 169,936. The 2018 flu/pneumonia deaths were 191,000. Then there are the big numbers, i.e. heart related deaths at 902,270, respiratory deaths at 290,774, dementia at 258,586, cancer at 699,394, and the list is extensive (brain, liver, kidneys, digestive, diabetes, etc). This is a serious viral death total, but not in relation to other causes of death.
Figure 4 puts the various death rates into perspective. COVID-19 is slightly above the 100,000 line.
Figure 4 is a bar graph of the Annual Death Rate in the United States from various causes. The horizontal axis are the years from 2010 to 2017. The vertical axis is the total number of deaths. The bars represent various diseases. Data from the CDC.