In order to evaluate whether the number 140,000 is significant or not we need a baseline. As a
starting point, let’s estimate the expected total number of deaths for the same period in a non-crisis
situation. We can estimate it from publicly available data.
The world population is around 7.7 billion. The globally averaged baseline crude death rate is
estimated at 7.6 deaths per 1000 population per year. In normal conditions we expect 59 million
expected deaths per year. There are 137 days in December 1st, 2019 to April 15th, 2020 period. It
means that we expect 22.2 (137/365*59) million deaths during this period. 140,000 deaths related
to coronavirus is 0.6% (140/22200) of total deaths. [37]
We are more interested how coronavirus related fatalities look in the context of other influenza-
like infections. The estimated number of deaths from lower respiratory infections (mostly pneu-
monia) is about 2.5 million [], and deaths from various types of influenza is about 300,000. It does
not include deaths from chronic respiratory diseases which would add another 4 million expected
fatalities per year. [14, 17]
During the relevant period, we expect 1.05 (137/365*2.8) million deaths from lower respiratory
infections and influenza. 140,000 sounds like a huge number, but it is buried within the noise of
the estimated number of deaths from pneumonia and influenza-like illnesses.
Because our attention is focused on the perceived danger from a single pathogen, we close our
eyes to 85% of fatal cases caused by other viruses and bacterias producing similar symptoms.
We did a simplistic comparison that has many problems
• it does not tell if the 140,000 we consider is part of the estimated 1.05 million or the excess
over this baseline
• it assumes uniform distribution while mortality rate fluctuates during the year. If we look
at shorter period like week or month we expect fluctuations from the yearly average. The
question is what divergence from the average is normal and what abnormal.
To make it easier to compare new data to baseline, statisticians use the z-score. It’s a metric
that shows how much a single data point stands out from the average. The z-score is normalized
by the standard deviation and is often adjusted for seasonality. A positive z-score indicates that
data point is above the average, a negative - below the average. It simplifies tracking time series
and comparing new data with historical data. In particular, in the ongoing mortality monitoring,
the z-score allows detecting when excess mortality occurs.
We are going to use the European monitoring of excess mortality for public health action
(Euromomo) [11]. It aggregates actual mortality reports for Europe.
Without relying on coronavirus data, we can check if the crisis of last weeks increased total
mortality. For the most affected countries like Italy, Spain and France, the z-score for the last
week of March and the first week of April 2020 rose to 18, 17 and 14 respectively, what means
that all-cause mortality is much higher than average. Neither is it extraordinary high. In January
2017, the z-score in the mentioned countries were 12, 14 and 12 respectively.
The effect of increased mortality at the turn of March and April is clearly visible in statistics
but not unusual. It is comparable to the increased mortality during seasonal flu outbreaks in the
past years.
2.3
Missing baseline
Specific claim
This coronavirus is unlike anything in our lifetime. The current situation is a once-in-a-century
pandemic. It’s the worst global crisis since World War II, etc.
These are basically content-free statements, but it’s worth to point out the same error - missing
baseline. With the difference that this time we cannot estimate baseline by ourselves because of
lack of data to compare. What is actually unique is that it’s the first time in history, when the
5