Content uploaded by Marine Baudin
Author content
All content in this area was uploaded by Marine Baudin on Oct 25, 2021
Content may be subject to copyright.
1
Nature of the COVID-era public health disaster in the
USA, from all-cause mortality and socio-geo-economic
and climatic data
Denis G. Rancourt1,*, Marine Baudin2, Jérémie Mercier2
1 Ontario Civil Liberties Association (ocla.ca) ; 2 Mercier Production (jeremie-mercier.com) ;
* denis.rancourt@alumni.utoronto.ca
This article has not been peer-reviewed by a journal.
It is published simultaneously at the following websites.
https://ocla.ca/
https://denisrancourt.ca/
https://archive.today/
https://www.researchgate.net/profile/Marine-Baudin
https://www.globalresearch.ca/
25 October 2021
2
Abstract
We investigate why the USA, unlike Canada and Western European countries, has a
sustained exceedingly large mortality in the “COVID-era” occurring from March 2020 to
present (October 2021). All-cause mortality by time is the most reliable data for
detecting true catastrophic events causing death, and for gauging the population-level
impact of any surge in deaths from any cause. The behaviour of the USA all-cause
mortality by time (week, year), by age group, by sex, and by state is contrary to
pandemic behaviour caused by a new respiratory disease virus for which there is no
prior natural immunity in the population. Its seasonal structure (summer maxima), age-
group distribution (young residents), and large state-wise heterogeneity are
unprecedented and are opposite to viral respiratory disease behaviour, pandemic or
not. We conclude that a pandemic did not occur. We infer that persistent chronic
psychological stress induced by the long-lasting government-imposed societal and
economic transformations during the COVID-era converted the existing societal
(poverty), public-health (obesity) and hot-climate risk factors into deadly agents, largely
acting together, with devastating population-level consequences against large pools of
vulnerable and disadvantaged residents of the USA, far above preexisting pre-COVID-
era mortality in those pools. We also find a large COVID-era USA pneumonia epidemic
that is not mentioned in the media or significantly in the scientific literature, which was
not adequately addressed. Many COVID-19-assigned deaths may be misdiagnosed
bacterial pneumonia deaths. The massive vaccination campaign (380 M administered
doses, 178 M fully vaccinated individuals, mainly January-August 2021 and March-
August 2021, respectively) had no detectable mitigating effect, and may have
contributed to making the younger population more vulnerable (35-64 years, summer-
2021 mortality).
3
Table of contents
Abstract ...................................................................................................................................................... 2
Summary .................................................................................................................................................... 4
List of figures ............................................................................................................................................ 7
Table of abbreviations and definitions ............................................................................................. 13
1. Introduction ......................................................................................................................................... 17
2. Data and methods ............................................................................................................................. 19
3. Results, analysis and discussion .................................................................................................. 25
3.1. All-cause mortality per year, USA, 1900-2020 .................................................................... 25
3.2. ACM by week (ACM/w), USA, 2013-2021 .............................................................................. 33
3.3. ACM by week (ACM/w), USA, 2013-2021, by state ............................................................. 38
3.4. Late-June 2021 heatwave event in ACM/w for Oregon and Washington .................... 41
3.5. ACM-SB/w normalized by population (ACM-SB/w/pop), by state .................................. 42
3.6. ACM-SB by cycle-year (winter burden, WB) by population (WB/pop), USA and
state-to-state variations ................................................................................................................... 54
3.7. Geographical distribution and correlations between COVID-era above-SB seasonal
deaths: cvp1 (spring-2020), smp1 (summer-2020) and cvp2 (fall-winter-2020-2021) ...... 62
3.8. Associations of COVID-era mortality outcomes with socio-geo-economic and
climatic variables ............................................................................................................................... 69
Obesity .............................................................................................................................................. 71
Poverty .............................................................................................................................................. 76
Climatic temperature ..................................................................................................................... 81
Obesity, poverty, and climatic temperature ............................................................................ 84
Age structure of the population ................................................................................................. 87
Population density ......................................................................................................................... 95
All-cause mortality by week (ACM/w) by age group ........................................................... 103
Comparing all-cause excess mortality and COVID-assigned mortality ........................ 112
Vaccination .................................................................................................................................... 122
4. Comparison with Canada, and implications............................................................................. 124
5. Mechanistic causes for COVID-era deaths ............................................................................... 129
6. Conclusion ........................................................................................................................................ 137
References ............................................................................................................................................. 141
Appendix: ACM/w, 2013-2021, with colour-differentiated cycle-years, for all the individual
states of continental USA .................................................................................................................. 147
4
Summary
We studied all-cause mortality (ACM) by time (week, year) 2013-2021 for the USA,
resolved by state, or by age group, in relation to several socio-geo-economic and
climatic variables (poverty, obesity, climatic temperature, population density,
geographical region, and summer heatwaves).
We calculate “excess” mortality, by calendar-year or (summer to summer) cycle-year or
selected ranges of weeks, as the week-by-week ACM above a summer baseline (SB)
ACM, which has a monotonic and linear variation on the decadal timescale, 2013-2019,
extrapolated into 2021.
Unlike Canada and Western European countries, the USA has a dramatic anomalous
increase in both ACM by year and “excess” ACM by year in 2020 and 2021, which
started immediately following the World Health Organization (WHO) 11 March 2020
declaration of a pandemic. Nothing of this magnitude occurs in other nations. The
USA’s yearly mortality in 2020-2021 is equal to (2020) and greater than (2021) the
mortality by year occurring in its domestic population just after the Second World War.
Regarding geo-temporal variations in ACM by week (ACM/w) and in excess (above-SB)
ACM by week (ACM-SB/w), we find that there are two distinct periods: the “COVID-era”
(March 2020 to present), and the “pre-COVID-era” (prior to March 2020). Normal
epidemiological variations occur in the pre-COVID-era, as has been observed for more
than a century, in all mid-latitude Northern hemisphere jurisdictions having reliable data;
whereas there is unprecedented state-wise jurisdictional and regional geographical
heterogeneity in ACM by time in the COVID-era, which is contrary to theoretical
pandemic behaviour caused by a new virus for which there is no prior natural immunity
in the population.
COVID-era time-integrated seasonal and yearly features of ACM-SB/w significantly
correlate with poverty (PV), obesity (OB), and climatic temperature (Tav), by state; and
5
differ by age group. The correlations account for the state-to-state heterogeneity, with
notable outliers in one feature (March-June 2020) of the ACM-SB/w; and such
correlations do not occur in pre-COVID-era cycle-year excess mortality. The co-
associations of excess deaths with PV, OB and Tav occur only in the COVID-era. We
show that normal (pre-COVID) excess (winter season) deaths — largely attributed to
viral respiratory diseases occurring in the elderly — occur irrespective of PV, OB and
climate, and that there is solely a correlation to age structure of the population in the
state.
An example of a co-correlation is the relation between the summer-2020 excess
mortality normalized by population (smp1/pop) and the product of OB and PV (OB.PV),
state-by-state (see article for details):
A similar large excess of deaths occurred in the summer 2021, which is also strongly
co-correlated with poverty, obesity and regional climate. In addition, we showed that
these 2020 and 2021 summer mortalities and massive fall-winter-2020-2021 mortality,
unlike with viral respiratory disease deaths, occur in younger people, over broad age
categories.
In the correlations that we identified, the 2020 and 2021 summer excess (above-SB)
mortalities extend to zero values for sufficiently small values of poverty, obesity or
6
summer temperatures, or their combinations, such as the product of poverty and
obesity.
We also found, for example, that the onset of the COVID-era is associated with an
increase in deaths of 15-34 year olds to a new plateau in ACM/w (approximately 400
more deaths per week), which does not return to normal over the period studied.
The behaviour of all-cause mortality in the COVID-era is irreconcilable with a pandemic
caused by a new virus for which there is no prior natural immunity in the population.
On the contrary, we concluded that the COVID-era deaths are of two types:
• A large narrow peak (in ACM/w) occurring immediately after the WHO
declaration of a pandemic apparently caused by the aggressive novel
government and medical responses that were applied in certain specific state
jurisdictions, against sick elderly populations (34 states do not significantly exhibit
this feature).
• Summer-2020, fall-winter-2020-2021, and summer-2021 peaks and excesses (in
ACM/w), which co-correlate with poverty, obesity and regional climate,
presumably caused by chronic psychological stress induced by the government
and medical responses, which massively disrupted lives and society, and
affected broad age groups, as young as 15 year olds.
Therefore, a pandemic did not occur; but an unprecedented systemic aggression
against large pools of vulnerable and disadvantaged residents of the USA did occur. We
interpret that the persistent chronic psychological stress induced by the societal and
economic transformation of the COVID-era converted the existing societal (poverty),
public-health (obesity) and hot-climate risk factors into deadly agents, largely acting
together, with devastating population-level consequences, far beyond the deaths that
would have occurred from the pre-COVID-era background of preexisting risk factors.
7
List of figures
Figure 1. All-cause mortality by calendar-year in the USA from 1900 to 2020………..…………25
Figure 2a. All-cause mortality by year in the USA for the 1-4, 5-14, 15-24 and 25-34 years age
groups, from 1900 to 2016……………………………………………………………………………..26
Figure 2b. All-cause mortality by year in the USA for the 35-44 and 45-54 years age groups,
from 1900 to 2016……………………...……………………………………………………………….27
Figure 2c. All-cause mortality by year in the USA for the 55-64, 65-74, 75-84 and 85+ years
age groups, from 1900 to 2016……..…………...…………………………………………………….27
Figure 3a. Population of the USA from 1900 to 2020..…...……………………………………..…29
Figure 3b. Population of the USA by age group from 1900 to 2016.……………………………..29
Figure 4a. All-cause mortality by year normalized by population for the USA from 1900 to
2020……………...…………………………………………...……………………………………….…30
Figure 4b. All-cause mortality by year normalized by population for the USA for the 15-24 years
age group, for each of both sexes, from 1900 to 1997.……………………………………..………31
Figure 4c. All-cause mortality by year normalized by population for the USA for the 25-34 years
age group, for each of both sexes, from 1900 to 1997…...…………………………………………32
Figure 5. All-cause mortality by week in the USA from 2013 to 2021…...……………..…………34
Figure 6. Difference between all-cause mortality and summer baseline mortality for the USA
from 2013 to 2021…...…………………………………………...……………………………………..35
Figure 7. Difference between all-cause mortality and summer baseline mortality for the USA
from 2018 to 2021…...…………………………………………...………………………………….….37
Figure 8. Map of COVID-era features pattern in the USA…...…………………………………….40
Figure 9a. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Connecticut, Maryland, Massachusetts, New Jersey and New York
from 2013 to 2021……………………………………...………………………………………….……43
Figure 9b(i). Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Colorado, Illinois, Indiana, Michigan and Pennsylvania from 2013
to 2021…...…………………………………………...…………………………………………….……44
Figure 9b(ii). Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Colorado, Illinois, Indiana, Michigan and Pennsylvania from 2019
to 2021.…...…………………………………………...…………………………...………………….…44
Figure 9c. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Iowa, Kansas, Missouri, Montana, Nebraska, North Dakota,
Oklahoma and South Dakota from 2013 to 2021…….…...…………………………………………45
8
Figure 9d. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Idaho, Nevada, New Mexico, Utah and Wyoming from 2013 to
2021…...…………………………………………...……………………………………….……………46
Figure 9e. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Oregon and Washington from 2013 to 2021…...………………….46
Figure 9f. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for California and Georgia from 2013 to 2021…...…………..………..47
Figure 9g. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Arizona, Florida, Mississippi, South Carolina and Texas from 2013
to 2021…...…………………………………………...………………………………………………….48
Figure 9h(i). Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Louisiana and Michigan from 2013 to 2021…...……...……………48
Figure 9h(ii). Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Louisiana and Michigan from 2019 to 2021…...………...…………49
Figure 10a. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for California, Florida, Michigan, Nevada, New York and South Dakota
from 2013 to 2021…...…………………………………………...……………………………..………51
Figure 10b. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for California, Florida, Michigan, Nevada, New York and South Dakota
from 2013 to 2019…...…………………………………………...………………………..……………51
Figure 10c. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for California, Florida, Michigan, Nevada, New York and South Dakota
from 2019 to 2021…...…………………………………………...……………………..………………52
Figure 11a. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Colorado, Connecticut, Illinois, Louisiana, New Jersey and New
York from 2013 to 2021…...…………………………………………...……….………………………53
Figure 11b. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Colorado, Connecticut, Illinois, Louisiana, New Jersey and New
York from 2013 to 2019…...…………………………………………...……….………………………53
Figure 11c. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Colorado, Connecticut, Illinois, Louisiana, New Jersey and New
York from 2019 to 2021…...…………………………………………...……….………………………54
Figure 12a. Winter burden normalized by population in the USA for cycle-years 2014 to
2021…...…………………………………………...………………………….…………………………55
Figure 12b. Winter burden normalized by population for each of the continental states of the
USA for cycle-years 2014 to 2021…...……………………………………………...………………..56
9
Figure 12c. Winter burden normalized by population in Alabama, Arizona, Florida, Louisiana,
Mississippi, South Carolina and Texas for cycle-years 2014 to 2021…...……..…………………57
Figure 12d. Winter burden normalized by population in Connecticut, Maryland, Massachusetts,
New Jersey and New York for cycle-years 2014 to 2021…...………………………………...……58
Figure 13. Frequency distributions of state-to-state values of WB/pop for each cycle-year, 2014-
2021…...…………………………………………...…………………………...………………………..59
Figure 14. Statistical parameters of the WB/pop distributions of the 49 continental states of the
USA for cycle-years 2014 to 2021…...……………………………………..…………………………60
Figure 15. Map of the intensity of the cvp1 mortality normalized by population for the continental
USA...……………………………………………………………………….....…………………………63
Figure 16. Map of the intensity of the smp1 mortality normalized by population for the
continental USA...………………………………………………………….....…………………………64
Figure 17a. smp1/pop versus cvp1/pop...……………………………………………………………65
Figure 17b. cvp2/pop versus cvp1/pop...…………………………………………………………….66
Figure 17c. cvp2/pop versus smp1/pop...……………………………………………………………66
Figure 17d. smp2/pop versus smp1/pop...…………………………………………………………..68
Figure 18. cvp2/pop versus smp1/pop, with the radius size determined by cvp1/pop………….69
Figure 19a. cvp1/pop versus obesity...……………………………………………………………….72
Figure 19b. smp1/pop versus obesity...……………………………………………………………...72
Figure 19c. cvp2/pop versus obesity...………………………………………….……………………73
Figure 19d. WB/pop for cycle-year 2019 versus obesity...……………………...…………………74
Figure 19e. WB/pop for COVID-era cycle-year 2020 versus obesity……………………………..75
Figure 19f. WB/pop for COVID-era cycle-year 2021 versus obesity……………………….……..76
Figure 20a. cvp1/pop versus poverty...………………………………………………………………77
Figure 20b. smp1/pop versus poverty...……………………………………………………………..77
Figure 20c. cvp2/pop versus poverty...………………………………………………………………78
Figure 20d. WB/pop for cycle-year 2019 versus poverty...………………………………….……..79
Figure 20e. WB/pop for COVID-era cycle-year 2020 versus poverty...…………………………..80
Figure 20f. WB/pop for COVID-era cycle-year 2021 versus poverty...……………………...……81
Figure 21. Mean daily average temperature: Mean of daily minimum and maximum, averaged
over the year, and for three decades (1970-2000) ...……………………………………...………..82
10
Figure 22. Average temperature, per state of the continental USA, for August 2020…………..83
Figure 23. smp1/pop versus average daily maximum temperature over July and August 2020,
Tmax Jul-Aug 2020...……………………………………………………………………………….…..84
Figure 24. Obesity versus poverty...………………………………………………………………….85
Figure 25. smp1/pop versus the product of obesity and poverty, with the radius size determined
by Tmax Jul-Aug 2020...…………………………………………...………………………….……….86
Figure 26. Tav 2020 versus the product of obesity and poverty, with the radius size determined
by smp1/pop...…………………………………………………………....……………………………..87
Figure 27a. WB/pop versus 85+/pop for cycle-year 2014...………………………………………..88
Figure 27b. WB/pop versus 85+/pop for cycle-year 2015...……………………………...………..89
Figure 27c. WB/pop versus 85+/pop for cycle-year 2016..………………………………...………90
Figure 27d. WB/pop versus 85+/pop for cycle-year 2017..…………………………………...…...90
Figure 27e. WB/pop versus 85+/pop for cycle-year 2018..……………………………………...…91
Figure 27f. WB/pop versus 85+/pop for cycle-year 2019..…………………………………………91
Figure 28a. cvp1/pop versus 85+/pop..……………………………………...……………………….93
Figure 28b. smp1/pop versus 85+/pop..………………………………………………………….….93
Figure 28c. cvp2/pop versus 85+/pop..………………………………………………………………94
Figure 28d. WB/pop versus 85+/pop for cycle-year 2020..……………………………………......94
Figure 28e. WB/pop versus 85+/pop for cycle-year 2021..………………………………………...95
Figure 29a. WB/pop for cycle-year 2014 versus population density..…………………………….96
Figure 29b. WB/pop for cycle-year 2015 versus population density..…………………………….97
Figure 29c. WB/pop for cycle-year 2016 versus population density..…………………………….97
Figure 29d. WB/pop for cycle-year 2017 versus population density..…………………………….98
Figure 29e. WB/pop for cycle-year 2018 versus population density..…………………………….98
Figure 29f. WB/pop for cycle-year 2019 versus population density..…………………………….99
Figure 30a. cvp1/pop versus population density..…………………………………………………100
Figure 30b. smp1/pop versus population density..………………………………………………..100
Figure 30c. cvp2/pop versus population density..…………………………………………………101
Figure 30d. WB/pop for cycle-year 2020 versus population density..…………………………..101
Figure 30e. WB/pop for cycle-year 2021 versus population density..…………………………..102
11
Figure 31. All-cause mortality by week, fully vaccinated individuals by day and COVID vaccine
doses administered by day, in the USA, from 2020 to 2021..…………………………………….123
Figure 32a. All-cause mortality by week in the USA for the 18-64 and 65+ years age groups,
from 2014 to 2021……………………………………………………………………………………..103
Figure 32b. Difference in all-cause mortality by week in the USA between the 65+ years and
the rescaled 18-64 years age groups, from 2014 to 2021………………………………………...104
Figure 33a. All-cause mortality by week normalized by population for the USA for the 14 years
and less age group, for each of both sexes, from 2020 to 2021………………………………….105
Figure 33b. All-cause mortality by week for the USA for the 15-34 years age group, both sexes,
from 2020 to 2021……………………………………………………………………………………..106
Figure 33c. All-cause mortality by week normalized by population for the USA for females of the
15-34 years age group, from 2020 to 2021……………………………………………………...….106
Figure 33d. All-cause mortality by week for the USA for the 35-54 years age group, both sexes,
from 2020 to 2021……………………………………………………………………………………..107
Figure 33e. All-cause mortality by week normalized by population for the USA for females of the
35-54 years age group, from 2020 to 2021…………………………………………………….…...108
Figure 33f. All-cause mortality by week normalized by population for the USA for the 55-64
years age group, for each of both sexes, from 2020 to 2021……………………………….….…108
Figure 33g. All-cause mortality by week normalized by population for the USA for the 65-74
years age group, for each of both sexes, from 2020 to 2021……………………………….….…109
Figure 33h. All-cause mortality by week normalized by population for the USA for the 75-84
years age group, for each of both sexes, from 2020 to 2021……………………………….….…110
Figure 33i. All-cause mortality by week normalized by population for the USA for the age group
85 years and older, for each of both sexes, from 2020 to 2021……………………………….…110
Figures 34a. All-cause, COVID-19, influenza, pneumonia and PIC mortality by week for the
USA from 2014 to 2021……………………………………………………………………………....113
Figure 34b. All-cause, COVID-19, influenza, pneumonia and PIC mortality by week for the USA
from 2019 to 2021……………………………………………………………………………………..114
Figure 34c. All-cause above-SB, COVID-19, influenza, pneumonia and PIC mortality by week
for the USA from 2014 to 2021……………………………………………………………………….115
Figure 34d. All-cause above-SB, COVID-19, influenza, pneumonia and PIC mortality by week
for the USA from 2019 to 2021……………………………………………………………………….115
Figure 34e. All-cause above-SB, COVID-19, influenza, pneumonia-pSB and PIC-pSB mortality
by week for the USA from 2014 to 2021…………………………………………………………….116
12
Figure 34f. All-cause above-SB, COVID-19, influenza, pneumonia-pSB and PIC-pSB mortality
by week for the USA from 2019 to 2021…………………………………………………………….117
Figure 34g. All-cause above-SB, COVID-19, influenza, pneumonia-pSB and ACM-SB minus
PIC-pSB mortality by week for the USA from 2014 to 2021………………………………...…….118
Figure 34h. All-cause above-SB, COVID-19, influenza, pneumonia-pSB and ACM-SB minus
PIC-pSB mortality by week for the USA from 2019 to 2021……………………………...……….119
Figure 34i. All-cause above-SB, COVID-19, influenza and pneumonia-pSB mortality by week,
and the ratio of COVID-19 deaths with pneumonia to all COVID-19 deaths by week, for the USA
in the COVID-era (March-2020 into 2021) …………………………………………………………120
Figure 35. All-cause mortality by week in Canada from 2010 to 2021……..…………………...125
Figure 36a. All-cause mortality by cycle-year for Canada, cycle-years 2011 to 2021…..…….126
Figure 36b. Winter burden for Canada for cycle-years 2011 to 2021……………………...……127
Figure 37. All-cause mortality by calendar-year, calendar-years 2010 to 2020, shown with all-
cause mortality by cycle-year, cycle-years 2011 to 2021, for Canada………………..…………128
Figure 38a. Map of life expectancy at birth for USA states, from census tracts 2010-
2015……………………………………………………………………………………….…………….130
Figure 38b. Antibiotic prescriptions per 1,000 persons by state (sextiles) for all ages, United
States, 2019…………………………………………………………………………...……………….131
Figure 39. Estimated number of outpatients with dispensed antibiotic prescriptions, USA, 2019-
2020…………………………………………………………………………………………….……….136
13
Table of abbreviations and definitions
Abbreviation
Name
Units
Description
Notes
85+ 85+ People
Population estimate of people of 85 years old an over as of
July 1st of the year
85+/pop 85+ by population %
Proportion of the people of 85 years old and older in the
population
ACM All-cause mortality Deaths
Total deaths from all causes (occurring in a defined period and
for a defined place)
ACM/w
All-cause mortality by
week
Deaths/w Total deaths from all causes occurring per week
ACM/w/pop ACM/w by population Deaths/w/pop
Total deaths from all causes occurring per week normalized by
population
ACM/y
All-cause mortality by
year Deaths/y Total deaths from all causes occurring per year
ACM/y/pop ACM/y by population Deaths/y/pop
Total deaths from all causes occurring per year normalized by
population
ACM-SB
All-cause minus summer
baseline mortality Deaths
Difference between total deaths from all causes and deaths
from all causes of the summer baseline 1
ACM-SB/w ACM-SB by week Deaths/w
Difference between total deaths from all causes and deaths
from all causes of the summer baseline per week
ACM-SB/w/pop ACM-SB/w by population Deaths/w/pop
Difference between total deaths from all causes and deaths
from all causes of the summer baseline per week normalized
by population ("Proportion of excess mortality per week")
av
Average
Arithmetic mean of all the values of a data set
(av-med)/av
Average minus median
divided by average
Ratio between the difference between the average and the
median and the average of the values of a data set
av-sd Average minus standard
deviation
Difference between the average and the standard deviation of
the values in a data set
14
COVID-19
Coronavirus disease
2019 N/A
“Coronavirus disease 2019 is a contagious disease caused by
severe acute respiratory syndrome coronavirus 2”
cvp1 COVID-peak 1 Deaths
Integrated deaths of ACM-SB between week 11 of 2020 (week
of March 9, 2020) and week 25 of 2020 (week of June 15,
2020), inclusively
2
cvp1/pop
COVID-peak 1 by
population Deaths/pop COVID-peak 1 normalized by population
cvp2 COVID-peak 2 Deaths
Integrated deaths of ACM-SB between week 40 of 2020 (week
of September 28, 2020) and week 11 of 2021 (week of March
15, 2021), inclusively
3
cvp2/pop
COVID-peak 2 by
population Deaths/pop COVID-peak 2 normalized by population
med
Median
The 50th percentile of values in a data set
neg-cor
Negative correlation
OB Obesity %
Prevalence of self-reported obesity by state and territory
(BRFSS (Behavioral Risk Factor Surveillance System), 2020)
OB.PV
Obesity times poverty
Product of obesity and poverty
pSB
Pneumonia summer
baseline mortality Deaths Pneumonia assigned-deaths baseline trend
Pneumonia-
pSB
Pneumonia minus
pneumonia summer
baseline mortality
Deaths Difference between total pneumonia-assigned deaths and
summer baseline pneumonia-assigned deaths
PIC
Pneumonia, Influenza
and/or COVID-19
mortality
Deaths Deaths from the following causes: pneumonia and/or influenza
and/or COVID-19
PIC-pSB
PIC minus pneumonia
summer baseline
mortality
Deaths Difference between PIC-assigned deaths and summer
baseline pneumonia-assigned deaths
ACM-SB -
PIC-pSB ACM-SB minus PIC-pSB Deaths
Difference between ACM-SB ("excess") and PIC-pSB ("PIC
above pneumonia-baseline") deaths
15
pop Population People
Resident population estimate for the states of the USA as of
July 1st of the year
popD Population density People/mile²
Number of inhabitants per unit surface area (average
population per square mile)
pos-cor
Positive correlation
PV
Poverty
%
Estimated percent of people of all ages in poverty
SB Summer baseline Deaths
Linear baseline of mortality independent of winter mortality
estimated from the summer trough weeks 26 to 37, inclusively,
of summers 2013 to 2019, inclusively
sd Standard deviation
Measure of the amount of variation or dispersion of values in a
data set
sd/av
Standard deviation
divided by average
Ratio between the standard deviation and the average of the
values of a data set
smp1 Summer-peak 1 Deaths
Integrated deaths of ACM-SB between week 26 of 2020 (week
of June 22, 2020) and week 39 of 2020 (week of September
21, 2020), inclusively
4
smp1/pop
Summer-peak 1 by
population Deaths/pop Summer-peak 1 divided by population
smp2 Summer-peak 2 Deaths
Integrated deaths of ACM-SB between week 26 of 2021 (week
of June 28, 2021) and week 37 of 2021 (week of September
13, 2021), inclusively
5
smp2/pop
Summer-peak 2 by
population Deaths/pop Summer-peak 2 divided by population
Tav Average temperature ° F
Average daily average temperature, where an average daily
temperature is the average between the max and min daily
temperatures
Tav 2020
Average temperature in
2020
° F
Average daily average temperature over the calendar-year
2020
Tmax
Maximum temperature
° F
Average daily maximum temperature
Tmax Jul-Aug
2020
Maximum temperature in
July and August 2020 ° F
Average daily maximum temperature over July and August
2020
16
USA United States of America N/A
Here USA means continental USA, which are 49 states,
including the District of Columbia and excluding Alaska and
Hawaii
WB Winter burden Deaths/y
Integrated deaths of ACM-SB between the week 31 of a year
N and the week 30 of a year N+1, inclusively (which is the
definition of a cycle-year)
6
WB/pop
Winter burden by
population
Deaths/y/pop Winter burden normalized by population 7
1 Also called "all-cause above-SB" or "excess" deaths in the text
2 Also called "March-June 2020 peak" or "covid peak" or "spring-2020 peak" or "spring-2020 excess mortality" in the text
3 Also called "fall-winter-2020-2021 excess mortality" in the text
4 Also called "summer-2020 excess mortality" in the text
5 Also called "summer-2021 excess mortality" in the text
6 If a year is placed in front, it means it's the WB of this cycle-year
7 If a year is placed in front, it means it's the WB/pop of this cycle-year
N/A stands for not applicable
17
1. Introduction
A small but growing number of researchers are recognizing that it is essential to
examine all-cause mortality (ACM), and excess deaths from all causes compared with
projections from historic trends, to make sense of the events surrounding COVID-19
(Jacobson and Jokela, 2021) (Kontopantelis et al., 2021) (Rancourt, 2020) (Rancourt et
al., 2020) (Rancourt et al., 2021) (Woolf et al., 2021).
In our prior analyses of ACM by time (by day, week, month, year) for many countries
(and by province, state, region or county), we showed that the data in the COVID-era
(March 2020 to present) is inconsistent with a viral respiratory disease pandemic, in that
the mortality is highly heterogeneous between jurisdictions, with no anomalies in most
places, and hot spots or hot regions with deaths that are synchronous with aggressive
local or regional responses, both medical and governmental (Rancourt, 2020) (Rancourt
et al., 2020) (Rancourt et al., 2021).
The surges in all-cause deaths are highly localized geographically (by jurisdiction) and
in time, which is contrary to pandemic behaviour; but is consistent with the surges being
caused by the known government and medical responses (Rancourt, 2020) (Rancourt
et al., 2020) (Rancourt et al., 2021).
In particular, Canada shows no evidence of a pandemic, since ACM by year (ACM/y) in
the COVID-era is squarely on the linear trend of the previous decade. In addition, the
ACM by week (ACM/w) data for Canada shows large province-level heterogeneity of
temporal and seasonal changes in ACM, by sex and by age group, that must be
ascribed to the impacts of medical and governmental measures (Rancourt et al., 2021).
We have also extensively studied ACM by time (day, month, year) for France, at many
jurisdictional levels (regions, departments, communes), in comparison to high-resolution
18
data for institutional occupancies and drug use (Rancourt et al., 2020) (and
unpublished), and examined data for European countries, to various degrees of detail.
We reported on the USA in our prior articles about ACM, concentrating on the
spectacular hot-spot anomalies that occurred in March through May 2020 (Rancourt,
2020) (Rancourt et al., 2020). Here, we extend our analysis for the USA, up to presently
available data, and include socio-geo-economic and climatic data.
The ACM data for the USA in the COVID-era has shocking features, unlike anything
else in the world. The USA is unique in this regard. Above-decadal-trend deaths in the
COVID-era are massive. Nothing like this occurs in neighbouring Canada. Nothing like
this occurs in Western European countries. Similar surges occur in Eastern European
countries, but are not of the same large magnitudes as in the USA.
Our goal was to describe the most that can be rigorously inferred from ACM by time,
jurisdiction, age group, and sex, in order to elucidate the nature of the massive excess
mortality that occurred in the USA in the COVID-era, and delimit its likely causes, with
an eye to known mechanisms of disease vulnerability (psychoneuroimmunology, and
stress-immune-survival relationships for humans). Therefore, we examined socio-geo-
economic data, including:
• Age structure of the population
• Population density
• Racial considerations
• Obesity
• Poverty (also median household income)
• Climatic temperatures
• Vaccination status (COVID-19 and flu vaccines)
• Antibiotic prescription rates
19
2. Data and methods
Table 1 describes data used in this work and the sources of the data.
Data
Country
Period
Time scale
Filters
Source
ACM USA 2013-2021* Week State CDC, 2021a
ACM USA 2013-2021* Week Age group1 CDC, 2021a
ACM USA 2020-2021** Week Age group2,
sex CDC, 2021b
ACM USA 1900-2020§ Year Age group3,
sex
CDC, 2021a
CDC, 2021c
CDC, 2021d
ACM USA 1900-1998 Year Age group3,
sex CDC, 2021c
ACM USA 1968-2016 Year Age group4,
sex CDC, 2021d
Obesity USA 2020 Year State CDC, 2021e
P-I-C USA 2013-2021* Week - CDC, 2021a
Population USA 1900-2020§§ Year Age group3,
sex
CDC, 2021c
CDC, 2021d
US Census
Bureau,
2021b
Population USA 1900-1997 Year Age group5,
sex CDC, 2021c
Population USA 1968-2016 Year Age group4,
sex CDC, 2021d
Population USA 2010-2020 Year State US Census
20
Bureau,
2021a
Population USA 2010-2020# Year State, age
group6, sex
US Census
Bureau,
2021b
Density USA 1910-2020## Decade State
US Census
Bureau,
2021c
Poverty USA 2019 Year State
US Census
Bureau,
2021d
Temperature USA 1895-2021*** Month State7 NOAA, 2021
Vaccines USA 2020-2021+ Day - CDC, 2021f
ACM Canada 2010-2021++ Week - StatCan,
2021
Table 1. Data retrieved. USA means continental USA, composed of 49 states, including the
District of Columbia and excluding Alaska and Hawaii, unless otherwise stated in the text.
* At the date of access, data were available from week-40 of 2013 to week-40 of 2021. Usable
data are until week-37 of 2021, due to insufficient data in later weeks, which gives a large
artifact (anomalous drop in mortality, see Appendix). For the work on USA at the state level, we
could add the missing weeks of 2013 (week-1 of 2013 to week-39 of 2020) thanks to a
previously downloaded file (downloaded on June 24, 2020) from the same website (CDC,
2021a), which was including those weeks back then.
** At the date of access, data were available from week-1 of 2020 (week ending on January 4,
2020) to week-40 of 2021 (week ending on October 9, 2021). Usable data are until week-37 of
2021 (week ending on September 18, 2021), due to insufficient data in later weeks, which gives
a large artifact (anomalous drop in mortality).
*** At the date of access, data were available until August 2021.
§ These data are a combination of the data found in CDC 2021a, CDC 2021c and CDC 2021d.
§§ These data are a combination of the data found in CDC 2021c, CDC 2021d and US Census
Bureau 2021b.
# In our work, we use the population data of the year 2020 (census estimate).
## In our work, we use the population density data of the year 2020.
+ At the date of access, data were available from December 14, 2020 (week-51 of 2020) to
September 27, 2021 (week-39 of 2021).
21
++ At the date of access, data were available from week-1 of 2010 (week ending on January 9,
2010) to week-30 of 2021 (week ending on July 31, 2021). Usable data are until week-20 of
2021 (week ending on May 22, 2021) due to not consolidated data in later weeks, which gives a
large artifact (anomalous drop in mortality).
1 3 age groups: <18, 18-64, 65+
2 11 age groups: <1, 1-4, 5-14, 15-24, 25-34, 35-44, 45-54, 55-64, 65-74, 75-84, 85+
3 12 age groups: <1, 1-4, 5-14, 15-24, 25-34, 35-44, 45-54, 55-64, 65-74, 75-84, 85+, unknown
4 14 age groups: <1, 1-4, 5-9, 10-14, 15-19, 20-24, 25-34, 35-44, 45-54, 55-64, 65-74, 75-84,
85+, not stated
5 19 age groups: <1, 1-4, 5-9, 10-14, 15-19, 20-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54,
55-59, 60-64, 65-69, 70-74, 75-79, 80-84, 85+
6 86 age groups: by 1 year age group, from 0 to 85+
7 Temperatures are not available for the District of Columbia.
StatCan (2021) defines a death as “the permanent disappearance of all evidence of life
at any time after a live birth has taken place” and excludes stillbirths. StatCan specifies
that the ACM for 2020 and 2021 is provisional and subject to change, and that the
counts of deaths “have been rounded to a neighbouring multiple of 5 to meet the
confidentiality requirements of the Statistics Act”.
According to CDC (CDC, 2021a):
• “[…] pneumonia, influenza and/or COVID-19 (PIC) deaths are identified based on
ICD-10 multiple cause of death codes.”
• “NCHS Mortality Surveillance System data are presented by the week the death
occurred at the national, state, and HHS Region levels, based on the state of
residence of the decedent.”
• “Not all deaths are reported within a week of death therefore data for earlier
weeks are continually revised and the proportion of deaths due to P&I or PIC
may increase or decrease as new and updated death certificate data are
received by NCHS.”
• “The COVID-19 death counts reported by NCHS and presented here are
provisional and will not match counts in other sources, such as media reports or
numbers from county health departments. COVID-19 deaths may be classified or
defined differently in various reporting and surveillance systems. Death counts
reported by NCHS include deaths that have COVID-19 listed as a cause of
22
death and may include laboratory confirmed COVID-19 deaths and
clinically confirmed COVID-19 deaths. Provisional death counts reported by
NCHS track approximately 1-2 weeks behind other published data sources on
the number of COVID-19 deaths in the U.S. These reasons may partly account
for differences between NCHS reported death counts and death counts reported
in other sources.”
• “In previous seasons, the NCHS surveillance data were used to calculate the
percent of all deaths occurring each week that had pneumonia and/or influenza
(P&I) listed as a cause of death. Because of the ongoing COVID-19 pandemic,
COVID-19 coded deaths were added to P&I to create the PIC (pneumonia,
influenza, and/or COVID-19) classification. PIC includes all deaths with
pneumonia, influenza, and/or COVID-19 listed on the death certificate.
Because many influenza deaths and many COVID-19 deaths have pneumonia
included on the death certificate, P&I no longer measures the impact of influenza
in the same way that it has in the past. This is because the proportion of
pneumonia deaths associated with influenza is now influenced by COVID-19-
related pneumonia. The PIC percentage and the number of influenza and
number of COVID-19 deaths will be presented in order to help better understand
the impact of these viruses on mortality and the relative contribution of each virus
to PIC mortality.”
For all the scatter plots presented in this article, the following colour-code is applied for
the 49 continental states of the USA (including District of Columbia, excluding Alaska
and Hawaii).
23
The main points of our methodology are as follows.
We work with all-cause mortality (ACM), deaths from all causes, in order to avoid the
uncertainty and bias in attributing a cause of death, in this context of COVID-19 in which
cause of death is not simple nor obvious. ACM data is available by jurisdiction (state,
country, county), by age group, by race, by sex, and by time (day, week, year). We can
normalize group-specific ACM totals by the respective populations of the relevant
groups, in order to allow comparisons between jurisdictions or different groups, on a
per-population basis.
Generally, in jurisdictions that exhibit seasonal winter maximums of mortality, the
bottom-values of mortality in the summer troughs follow a straight-line trend on a
decadal or shorter timescale. We call this trend-line the “summer baseline” (SB), and we
use it to count above-SB deaths, when we wish to thus quantify “excess deaths”.
In other words, we are following our previous methodology in which we argued that
mortality by time (day, week, month) is best analyzed using a SB, and winter burden
(WB) deaths above the SB, over a (natural) cycle-year from summer to following
summer, rather than use assumed underlying sinusoidal seasonal variations of any
presumed component(s), since such sinusoidal theoretical curves fail to represent the
data or any of its inferred principle components (e.g., Simonsen et al., 1997). Although
the summer trough mortality values follow a linear local trend by time (in normal, pre-
COVID-era, circumstances), above-SB features have significant randomness in their
season to season variations, suggesting that summer baseline mortality is
representative of “stable” mortality not influenced by the many different and seasonally
variable winter-time life-threatening health challenges (Rancourt, 2020) (Rancourt et al.,
2020) (Rancourt et al., 2021).
SB estimation at the state level
The linear summer baseline (SB) is a least-squares fit to the summer troughs for
summer-2013 through summer-2019, using the summer trough weeks 27 to 36,
24
included, for all the states of the continental USA, except for Alabama and Wisconsin for
summer-2014 and summer-2015, respectively, and corrected by 1 % (see below). For
Alabama, only the weeks [30-32] were used for summer-2014 as drops in data are seen
for weeks [27-29] and weeks [33-36] of 2014 (see Appendix). For Wisconsin, only the
weeks [27-29] and [33-36] were used for summer-2015 as a drop in data is seen for
weeks [30-32] of 2015 (see Appendix). We corrected the SB by 1 % so as to lower the
SB and make it match the bottoms of the summer troughs. We also estimated the SB
taking different summer periods, from the shortest to the largest: weeks [30-32], weeks
[29-33], weeks [28-35] and weeks [27-36], to determine our 1 % correction. We found
that the larger the period, the better the estimate of the SB slope, but also the higher the
estimate of the SB intercept, as the last weeks towards the previous winter season and
the first weeks towards the next winter season are included. We thus decided to
estimate the SB with the largest summer period (weeks [26-37]) and lower the intercept
by 1 % (no correction leading to a too high intercept and a correction factor of 2 %
leading to a too low intercept). The SB is so estimated between the weeks 26 and 37
(inclusively) of each summer of the pre-COVID-era (summers 2013 to 2019), which
corresponds to the weeks laying from the beginning of July to the beginning of
September.
SB estimation at the national level
• For work involving the states, the SB estimate of the USA is a sum of the SB
estimates of each individual state.
• For work not involving the states, the SB is a least-squares fit to the summer
troughs for summer-2014 through summer-2019, using the summer trough
weeks 27 to 36, included, for the whole USA (including Alaska and Hawaii) with
no correction, since none was needed.
In the same way that we thus quantify a winter burden of deaths in a given cycle-year,
we can also quantify an excess (above-SB) of deaths over any period of time, such as
over a period that captures any prominent features in ACM by time. We defined such
periods of interest occurring in the COVID-era: a spring-2020 peak (cvp1),
25
summer-2020 (smp1), the fall-winter-2020-2021 maximum (cvp2), and summer-2021
(smp2), as specified in the text.
3. Results, analysis and discussion
3.1. All-cause mortality per year, USA, 1900-2020
We start by examining ACM/y (per calendar-year) in the USA, for the years 1900
through 2020. This is shown in Figure 1.
Figure 1. All-cause mortality by calendar-year in the USA from 1900 to 2020. Data were
retrieved as described in Table 1.
The ACM/y 1900-2020 has the following main features. First, it has a generally
increasing trend over the entire period, with a slope of approximately 16K deaths per
year per year (16K/y/y) in the region 1920-2010. The overall increasing trend is due to
population growth. One needs to normalize by population to remove this dominant effect
(see below). Second, there is a large increase in 1918, which corresponds to the so-
26
called “1918 Flu Pandemic”. Third, there is a large increase in 2020, which corresponds
to the first year of the COVID-era. Fourth, there are notable increases in the late-1920s
and mid-1930s, which correspond to the hardships associated with The Great
Depression and the accompanying decade-long Dust Bowl droughts of the Midwest.
Fifth (by omission), there are no detected increases that would correspond to any of the
major 20th-21st century influenza pandemics that are described to have occurred in
1957-58, 1968, and 2009 (Doshi, 2008) (Doshi, 2011).
These main features in ACM/y are clarified and enhanced on examining ACM/y by age
group (available for 1900-2016). This is shown for all the ages, excluding <1 year,
divided into 10 age groups in Figure 2.
Figure 2a. All-cause mortality by year in the USA for the 1-4, 5-14, 15-24 and 25-34 years
age groups, from 1900 to 2016. Data are displayed per calendar-year. Data were retrieved as
described in Table 1.
27
Figure 2b. All-cause mortality by year in the USA for the 35-44 and 45-54 years age
groups, from 1900 to 2016. Data are displayed per calendar-year. Data were retrieved as
described in Table 1.
Figure 2c. All-cause mortality by year in the USA for the 55-64, 65-74, 75-84 and 85+ years
age groups, from 1900 to 2016. Data are displayed per calendar-year. Data were retrieved as
described in Table 1.
The ACM/y 1900-2016 by age-group data allows the following observations to be made.
28
Regarding 1918, the event was devastating for the age groups 15-24 years and 25-34
years, much less so for the age groups 35-44 years and 45-54 years, and virtually
undetected for those 55 years and older, which would be very surprising for influenza. In
fact, we know that most of the deaths were associated with massive bacterial lung
infections (Morens et al., 2008) (Chien et al., 2009) (Sheng et al., 2011), in an era
predating antibiotics, in a period massively perturbed by a world war, and that the event
was concomitant with typhoid epidemics in Europe and Russia.
Regarding The Great Depression and the Dust Bowl devastation, the late-1920s and
mid-1930s increases in ACM/y are prominent for the 15-24, 25-34, 35-44 and 45-54
years age groups, but are not detected for 55 year olds and older.
Regarding 20th-21st century purported influenza pandemics, there is no trace of
increased mortality for 1957-58, 1968, and 2009, in any age group, including the older
age groups of 55-64, 65-74, 75-84, and 85+ years. Clearly, these 20th century declared
pandemics had negligible impacts on all-cause mortality; not comparable to the large
impacts of the events of 1918, late-1920s-mid-1930s, <1945, and 2020, which are
associated with major socio-economic upheavals (the First World War, The Great
Depression and Dust Bowl, the Second World War, and the medical and government
response to the declared COVID-19 pandemic, respectively).
The ACM/y by age group has long-period (decadal) variations with notable broad
minima occurring at approximately:
~1975-1980: 35-44 years age group
~1985-1990: 45-54 years age group
~1995-2000: 55-64 years age group
~2005-2010: 65-74 years age group
~2010-2015: 75-84 years age group
These variations are due to the post Second World War baby boom effects on
population.
29
The population of the USA varied from 1900 to 2020 as shown in Figure 3 (and from
1900 to 2016 for the age groups).
Figure 3a. Population of the USA from 1900 to 2020. Data are displayed per calendar-year.
Data were retrieved as described in Table 1.
Figure 3b. Population of the USA by age group from 1900 to 2016. Data are displayed per
calendar-year. Data were retrieved as described in Table 1.
30
Here (Figure 3a), we see a large dip in population at 1943-1945, related to the Second
World War. The slope to population versus time also changes dramatically at 1943-
1945, increasing after the war, in accordance with the known baby boom. The
population by age group (Figure 3b) confirms that the dip at 1943-1945 is solely from
the 15-24 and 25-34 years age groups, especially 15-24 years. This figure (Figure 3b)
also shows the dramatic consequences of the baby boom, showing itself, age group
after age group, as the baby boomers age. The monotonic increase in the 85+ years
population (Figure 3b) is directly the cause of the monotonic increase in 85+ years
deaths (Figure 2c).
Next, we normalize ACM/y (Figure 1) by population (Figure 3a), 1900-2020, to obtain
ACM/y/pop shown in Figure 4a.
Figure 4a. All-cause mortality by year normalized by population for the USA from 1900 to
2020. Data are displayed per calendar-year. Data were retrieved as described in Table 1.
This allows us to see ACM/y expressed as a fraction of population. We again see the
gigantic catastrophe that was the 1918 event (pneumonia/typhoid, wartime upheaval),
peaks in the late-1920s and mid-1930s (Great Depression, Dust Bowl), a peak in the
Second World War period (young men, 15-24 and 25-34 years age groups, as per
31
Figure 3b), relatively uneventful mortality after 1945 (no public health catastrophes
detected), no sign of the announced pandemics of 1957-58, 1968, and 2009, and the
COVID-era increase of 2020 (a subject of this article).
The mortality events of the late 1920s, mid-1930s and <1945, and the >1945 uneventful
period, are elucidated further by examining ACM/y/pop resolved by age group and by
sex, as per the following.
Figure 4b. All-cause mortality by year normalized by population for the USA for the 15-24
years age group, for each of both sexes, from 1900 to 1997. The population of the specific
age group and sex is used for each normalization. Data are displayed per calendar-year. Data
were retrieved as described in Table 1.
32
Figure 4c. All-cause mortality by year normalized by population for the USA for the 25-34
years age group, for each of both sexes, from 1900 to 1997. The population of the specific
age group and sex is used for each normalization. Data are displayed per calendar-year. Data
were retrieved as described in Table 1.
Figures 4b and 4c show that both young men and women were impacted by the
hardships of the late-1920s and mid-1930s, but that only young men were impacted to
death by the Second World War. Interestingly, 15-24 year old men had relatively high
mortality between the mid-1960s and the early-1980s.
The 2020 value of ACM/y/pop brings us back to a mortality equal to the mortality by
population that prevailed in 1945 (Figure 4a), which suggests that the socio-economic
upheavals from COVID-19 response are comparable to the upheavals from the last
major war period, with an albeit much older population presently, and possibly greater
class disparity, since The New Deal had already been implemented in 1945, in
response to the hardships of the 1930s.
33
3.2. ACM by week (ACM/w), USA, 2013-2021
The ACM/w for the USA from 2013 to 2021 is shown in Figure 5, with a straight-line
trend for the bottoms of the summer troughs for 2013 through 2019 (of the pre-COVID-
era). We call this trend-line the “summer baseline” (SB), and we use it to count above-
SB deaths (“excess” deaths).
We are following our previous methodology in which we argued that mortality by time
(day, week, month) is best analyzed using a SB, and winter burden deaths (WB) above
the SB, over a (natural) cycle-year from summer to following summer, rather than use
assumed underlying sinusoidal seasonal variations of any presumed component(s),
since such sinusoidal theoretical curves fail to represent the data or any of its inferred
principle components (e.g., Simonsen et al., 1997). It is a general feature with seasonal
mortality data that SB trends are typically linear on the timescale of one decade or so,
whereas above-SB features have significant randomness in their season to season
variations, suggesting that summer baseline mortality is representative of “stable”
mortality not influenced by the many different and seasonally variable winter-time life-
threatening health challenges (Rancourt, 2020) (Rancourt et al., 2020) (Rancourt et al.,
2021).
34
Figure 5. All-cause mortality by week in the USA from 2013 to 2021. Data are displayed
from week-1 of 2013 to week-37 of 2021. The linear summer baseline (SB) is a least-squares fit
to the summer troughs for summer-2013 through summer-2019, using the summer trough
weeks 27 to 36, included, except for Alabama and Wisconsin for summer-2014 and summer-
2015, respectively, and corrected by 1 % (see section 2). Data were retrieved from CDC (CDC,
2021a), as described in Table 1.
Next, for the sake of visualization, we can remove the SB from the ACM, week by week,
to obtain ACM-SB/w. This is shown for the USA from 2013 to 2021, in Figure 6, where
we have used different colours for the different cycle-years.
35
Figure 6. Difference between all-cause mortality and summer baseline mortality for the
USA from 2013 to 2021. Data are displayed from week-1 of 2013 to week-37 of 2021. The
different colours are for the different cycle-years. The cycle-year starts on week-31 of a
calendar-year (beginning of August) and ends on week-30 of the next calendar-year (end of
July). ACM data were retrieved from CDC (CDC, 2021a), as described in Table 1. SB was
estimated as described in section 2.
Many striking features occur in ACM/w (or ACM-SB/w) in the COVID-era period for the
USA (Figures 5 and 6):
• The WB (total above-SB deaths per cycle-year) is much greater in cycle-years
2020 (summer-2019 to summer-2020) and 2021 (summer-2020 to summer-
2021) than in cycle years 2014 through 2019, which is consistent with ACM/y
already discussed above (Figures 1 and 4).
• The 2020 cycle-year exhibits a sharp and intense feature spanning weeks 11
through 25 of 2020, starting when the pandemic was declared by the World
Health Organization (WHO) on 11 March 2020, lasting three months, and which
we have called “the COVID peak” and amply described in our previous articles
(Rancourt, 2020) (Rancourt et al., 2020) (Rancourt et al., 2021). In this article, we
refer to this feature and its integrated intensity as “cvp1”.
• There is “no summer”, in terms of lower mortality, in the summer-2020. The
ACM/w does not descend down to the SB. In fact, the summer of 2020 exhibits a
36
broad mid-summer peak in ACM/w, spanning weeks 26 through 39 of 2020
(approximately mid-June to mid-September), which is unprecedented in any
ACM by time data that we have examined, for data since 1900 for dozens of
countries and hundreds of jurisdictions. In this article, we refer to this feature and
its integrated intensity as “smp1”.
• The 2021 cycle-year exhibits a massive peak, spanning from week-40 of 2020
through to week-11 of 2021 (approximately late-September 2020 to mid-March
2021). The peak extends to 35K deaths per week above SB. It is anticipated that
the ACM/y for 2021 will be larger than for 2020, which in turn brought us back to
mortality of the magnitude that was occurring just after the Second World War,
on a per population basis (Figure 4a). In this article, we refer to this winter 2020-
2021 feature and its integrated intensity as “cvp2”.
• Finally, there is a summer-2021 upsurge of mortality (ACM/w) in the last weeks
of the usable data set, starting in mid-July 2021. This upsurge in ACM/w is
particularly large for Florida, for example. We refer to this feature as “smp2”,
which is interrupted by the end of the data set (week-37 of 2021 for consolidated
data, as described in section 2).
To be clear, the three uninterrupted prominent features in the USA ACM/w for the
COVID-era (cvp1, smp1, and cvp2) are shown, according to their operational definitions
in Figure 7. For each feature, its quantification is achieved by summation of ACM-SB/w
over the weeks spanned by the feature. The late-summer-2021 feature “smp2” is also
indicated.
37
Figure 7. Difference between all-cause mortality and summer baseline mortality for the
USA from 2018 to 2021. Data are displayed from week-1 of 2018 to week-37 of 2021. The
cvp1, smp1, cvp2 and smp2 features discussed in the text are indicated. The light-blue vertical
lines represent the weeks 11, 25, 40 of 2020 and 11 of 2021, emphasizing the delimiting weeks
of the cvp1, smp1 and cvp2 features. The constant zero line is in black. ACM data were
retrieved from CDC (CDC, 2021a), as described in Table 1. SB was estimated as described in
section 2.
Although these features in USA ACM (cvp1, smp1, cvp2, smp2; highlighted in Figure 7)
are unprecedented in recent decades and are shocking in themselves; an equally
striking aspect is only seen on examining ACM/w (or ACM-SB/w) by state, for individual
states. The later examination shows (below) that the said features in the COVID-era,
unlike anything previously observed in epidemiology, are often dramatically different, in
both relative and absolute magnitudes, and in shape and position, in going from state to
state. The next section is devoted to illustrating this remarkable state-to-state variability
in COVID-era ACM by time.
cvp1 smp1 cvp2
smp2
38
3.3. ACM by week (ACM/w), USA, 2013-2021, by state
Graphs of ACM/w, from 2013 to 2021, with colour-differentiated cycle-years, for all the
individual states of continental USA (excluding Alaska and Hawaii) are shown in
Appendix (attached below).
In these graphs (Appendix), note that the pre-COVID-era seasonal pattern (2013-2019)
is essentially identical from state to state (more on this further below), whereas there
are large state to state changes in the COVID-era patterns. This concurs with our
previous findings that COVID-era behaviour in ACM by time is abnormally
heterogeneous on a jurisdictional basis, which is the opposite of past seasonal
epidemiological behaviour (Rancourt, 2020) (Rancourt et al., 2020) (Rancourt et al.,
2021). Woolf et al. (2021) also report large USA regional differences in all-cause excess
mortality by time patterns during the COVID-era.
Some comparative and systematic features in these curves (Appendix) are as follows.
• L0M / North-Easterly coastal states: Several of the North-Easterly coastal
states exhibit a pattern in cvp1-smp1-cvp2 (an “L0M” pattern) in which cvp1 is
very large, smp1 is essentially zero (ACM/w comes down to the SB values) and
cvp2 is of medium magnitude: New York, New Jersey, Connecticut,
Massachusetts and Rhode Island, and Maryland and District of Columbia to
some degree.
• LSL / North-Central-Easterly non-coastal states: A group of neighbouring
North-Central-Easterly non-coastal states exhibit a pattern in cvp1-smp1-cvp2
(an “LSL” pattern) in which cvp1 is large, smp1 is small (near-zero) and cvp2 is
large: Colorado, Delaware, Illinois, Indiana, Michigan, and Pennsylvania,
although Michigan has a unique extra peak in ACM/w.
• LSLx / Michigan: Michigan has an LSL pattern and belongs to the latter group,
however its LSL pattern is followed by a unique late peak occurring in March
through May 2021, centered in mid-April. Therefore, we refer to Michigan’s
pattern as “LSLx”.
39
• 00L / prairie states: Seven of the ten prairie or Great Plains states, states that
experienced the Dust Bowl drought of the 1930s, saw no anomalous mortality
whatsoever until late into the COVID-era, until the fall of 2021. Here, cvp1 and
smp1 are essentially zero or near-zero, and the only large feature is cvp2 (“00L”
pattern). Easterly neighbouring states of Iowa, Missouri and Wisconsin also have
this 00L pattern: Iowa, Kansas, Missouri, Montana, Nebraska, North Dakota,
Oklahoma, South Dakota, and Wisconsin. The prairie states of New Mexico and
Wyoming have a similar pattern, 0SL; whereas Texas has 0LL, and Colorado has
LSL.
• 0SL / Central-Westerly and Central-Easterly states: The cluster of adjacent
states of Arkansas, Idaho, Kentucky, North Carolina, Tennessee, West Virginia,
Wyoming, Nevada and Utah, and the prairie state of New Mexico, exhibit a “0SL”
pattern. The 00L and 0SL patterns are similar: in 00L we characterize smp1 as
“near-zero”, whereas in 0SL we characterize smp1 as “small”.
• 0SL / North-Westerly coastal states: The North-Westerly coastal states of
Oregon and Washington also have the 0SL pattern; and a sharp (one-week)
heatwave signal discussed below (section 3.4).
• SBL / North-Easterly states: Minnesota, New Hampshire, Ohio, and Virginia
exhibit an “SBL” pattern, intermediate between SSL and S0L.
• SSL / California and Georgia: California and Georgia exhibit similar patterns to
each other, in which both cvp1 and smp1 are distinct but small or medium, and
cvp2 is very large. We refer to this as an “SSL” pattern. The SSL pattern occurs
in populous states but is otherwise similar to the 00L and 0SL patterns, in that
relatively small or near-zero excess mortality occurs until late into the
COVID-era, until the fall of 2021 when cvp2 starts and becomes a large feature
in ACM/w.
• 0LL / Southern states: Both Florida and Texas exhibit a “0LL” pattern in cvp1-
smp1-cvp2 in which cvp1 is essentially zero, whereas smp1 and cvp2 are both
large. Most of the most southerly states exhibit this pattern: Alabama, Arizona,
Florida, Mississippi, South Carolina, and Texas; whereas Louisiana exhibits a
pattern in which all three features are large, an “LLL” pattern. Thus, the Southern
40
states are generally characterized and distinguished by large mortalities in the
summer of 2020, which is exceptional for these states, followed by large
mortalities in the fall and winter of 2020-2021.
• LLL / Louisiana: Louisiana is the only state that has all three main features in
ACM/w (cvp1, smp1, cvp2) being comparable and large. It is the only Southern
state that experienced a large cvp1 mortality at the start of the COVID-era.
• The remaining states, Vermont and Maine, have borderline patterns to those
described above, which could be characterized as 00S and 0SS, respectively.
• The summer-2021 feature “smp2” occurs in virtually all the states (see
Appendix).
This distribution of cvp1-smp1-cvp2 pattern type is shown, colour coded, on a map of
the USA, in Figure 8.
Figure 8. Map of COVID-era features pattern in the USA. The different colours represent the
different pattern groups discussed in the text: black = L0M, gray = LSL, dark blue = 00L, blue =
0SL, light blue = SSL, purple = SBL, red = 0LL, yellow = LLL, white = 00S and 0SS. The first
character of the pattern characterizes the cvp1 feature, the second the smp1 feature and the
last the cvp2 feature. L stands for large, M for medium, S for small, B for borderline and 0 for
zero / near-zero.
41
3.4. Late-June 2021 heatwave event in ACM/w for Oregon and Washington
There are sharp peaks (a single week or so) in the ACM/w data for Oregon and
Washington, occurring at week-26 of 2021, which is the week of 28 June 2021
(Appendix).
The increased deaths coincide with an extraordinary weather event: The two states and
British Columbia (Canada) experienced a short but record-breaking summer heatwave.
NASA Earth Observatory (2021) described the heatwave as follows:
Taking peak-to-local-baseline values, we estimate excess deaths from the heatwave to
have been 246 and 475 deaths, respectively for Oregon and Washington.
42
This is a reminder of the deadliness of stress from atmospheric heat, which is relevant
to our discussion about the COVID-era anomalies in the USA (below). We previously
quantified such a heat-wave mortality event that occurred in France in 2003 (Rancourt
et al., 2020).
3.5. ACM-SB/w normalized by population (ACM-SB/w/pop), by state
The different state-wise patterns of mortality in the USA during the COVID-era are best
examined using ACM-SB/w normalized by population, ACM-SB/w/pop, and by
reference to the cvp1-smp1-cvp2 patterns identified above. Normalization by population
allows direct comparisons of the data for states with different populations.
In the following figures, normalization was done as follows:
Normalization of a cycle-year N was done with the population estimated just before the
start of the cycle-year. Population estimates are each year on July 1st. The cycle-year
starts on week-31 of a calendar-year (beginning of August). At the date of access,
population estimates were from 2010 to 2020, so the cycle-year 2022 (last weeks of the
data set) was normalized by the last available population estimate, the one for 2020.
When at the state level, the population used for normalization is the population of the
specific state.
ACM-SB/w/pop curves are shown by groups of similar behaviours in Figure 9, as:
(a) L0M / North-Easterly coastal states: Connecticut, Maryland, Massachusetts, New
Jersey, and New York.
(b) LSL / North-Central-Easterly non-coastal states: Colorado, Illinois, Indiana,
Michigan (LSLx), and Pennsylvania.
43
(c) 00L / prairie states: Iowa, Kansas, Missouri, Montana, Nebraska, North Dakota,
Oklahoma, and South Dakota. (Wisconsin is excluded because of bad data
points for 2015, see Appendix.)
(d) 0SL / Central-Westerly non-coastal states: Idaho, Nevada, New Mexico, Utah,
Wyoming.
(e) 0SL / North-Westerly coastal states: Oregon and Washington. (With June-2021
heatwave peak.)
(f) SSL / California and Georgia: California and Georgia.
(g) 0LL / Southern states: Arizona, Florida, Mississippi, South Carolina, and Texas
(Alabama is excluded because of bad data points for 2014, see Appendix).
(h) LLL / Louisiana: Louisiana, shown with Michigan.
Figure 9a. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Connecticut, Maryland, Massachusetts, New Jersey and
New York from 2013 to 2021. Data are displayed from week-1 of 2013 to week-37 of 2021.
The dashed line emphasizes the zero. ACM data were retrieved from the CDC (CDC, 2021a)
and population data were retrieved from the US Census Bureau (US Census Bureau, 2021a),
as described in Table 1. SB was estimated as described in section 2.
44
Figure 9b(i). Difference between all-cause mortality and summer baseline mortality by
week normalized by population for Colorado, Illinois, Indiana, Michigan and Pennsylvania
from 2013 to 2021. Data are displayed from week-1 of 2013 to week-37 of 2021. The dashed
line emphasizes the zero. ACM data were retrieved from the CDC (CDC, 2021a) and population
data were retrieved from the US Census Bureau (US Census Bureau, 2021a), as described in
Table 1. SB was estimated as described in section 2.
Figure 9b(ii). Difference between all-cause mortality and summer baseline mortality by
week normalized by population for Colorado, Illinois, Indiana, Michigan and Pennsylvania
from 2019 to 2021. Data are displayed from week-1 of 2019 to week-37 of 2021. The dashed
45
line emphasizes the zero. ACM data were retrieved from the CDC (CDC, 2021a) and population
data were retrieved from the US Census Bureau (US Census Bureau, 2021a), as described in
Table 1. SB was estimated as described in section 2.
Figure 9c. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Iowa, Kansas, Missouri, Montana, Nebraska, North Dakota,
Oklahoma and South Dakota from 2013 to 2021. Data are displayed from week-1 of 2013 to
week-37 of 2021. The dashed line emphasizes the zero. ACM data were retrieved from the
CDC (CDC, 2021a) and population data were retrieved from the US Census Bureau (US
Census Bureau, 2021a), as described in Table 1. SB was estimated as described in section 2.
46
Figure 9d. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Idaho, Nevada, New Mexico, Utah and Wyoming from 2013
to 2021. Data are displayed from week-1 of 2013 to week-37 of 2021. The dashed line
emphasizes the zero. ACM data were retrieved from the CDC (CDC, 2021a) and population
data were retrieved from the US Census Bureau (US Census Bureau, 2021a), as described in
Table 1. SB was estimated as described in section 2.
Figure 9e. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Oregon and Washington from 2013 to 2021. Data are
displayed from week-1 of 2013 to week-37 of 2021. The dashed line emphasizes the zero. ACM
47
data were retrieved from the CDC (CDC, 2021a) and population data were retrieved from the
US Census Bureau (US Census Bureau, 2021a), as described in Table 1. SB was estimated as
described in section 2.
Figure 9f. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for California and Georgia from 2013 to 2021. Data are displayed
from week-1 of 2013 to week-37 of 2021. The dashed line emphasizes the zero. ACM data were
retrieved from the CDC (CDC, 2021a) and population data were retrieved from the US Census
Bureau (US Census Bureau, 2021a), as described in Table 1. SB was estimated as described in
section 2.
48
Figure 9g. Difference between all-cause mortality and summer baseline mortality by week
normalized by population for Arizona, Florida, Mississippi, South Carolina and Texas
from 2013 to 2021. Data are displayed from week-1 of 2013 to week-37 of 2021. The dashed
line emphasizes the zero. ACM data were retrieved from the CDC (CDC, 2021a) and population
data were retrieved from the US Census Bureau (US Census Bureau, 2021a), as described in
Table 1. SB was estimated as described in section 2.
Figure 9h(i). Difference between all-cause mortality and summer baseline mortality by
week normalized by population for Louisiana and Michigan from 2013 to 2021. Data are
displayed from week-1 of 2013 to week-37 of 2021. The dashed line emphasizes the zero. ACM
49
data were retrieved from the CDC (CDC, 2021a) and population data were retrieved from the
US Census Bureau (US Census Bureau, 2021a), as described in Table 1. SB was estimated as
described in section 2.
Figure 9h(ii). Difference between all-cause mortality and summer baseline mortality by
week normalized by population for Louisiana and Michigan from 2019 to 2021. Data are
displayed from week-1 of 2019 to week-37 of 2021. The dashed line emphasizes the zero. ACM
data were retrieved from the CDC (CDC, 2021a) and population data were retrieved from the
US Census Bureau (US Census Bureau, 2021a), as described in Table 1. SB was estimated as
described in section 2.
Figures 8 and 9 show that there are large state-to-state differences in COVID-era
mortality by time, and that these differences approximately group into four (4) types, by
geographical region, as:
• L0M : North-East coastal states
• LSL : North-East non-coastal states
• 00L / 0SL / SSL / SBL : Central and Western-Eastern states
• 0LL : Southern states
Louisiana is unique, with an LLL pattern, and large mortality in all three periods (cvp1,
smp1, cvp2). Michigan (LSLx) has a unique late peak, occurring in March through May
50
2021, centered on mid-April 2021. Oregon and Washington have unique June-2021
single-week heatwave peaks.
This description is “coarse grain” and is simplified. For example, California has a distinct
cvp1 feature even though it is much smaller than that occurring in the North-East states.
Also, what happened in New York City is literally off-the-charts regarding cvp1
(Rancourt, 2020).
A most striking aspect of mortality during the COVID-era is precisely the state-wise
heterogeneity in ACM by time, which we have described and illustrated above, and in
the Appendix. This is striking because the seasonal cycle of all-cause deaths is usually
remarkably uniform from state to state, from country to country, from province to
province, from county to county… through all the inferred and declared epidemics and
pandemics of viral respiratory diseases. Although the shapes of ACM by time change
from season to season, the shapes for a given year are nonetheless synchronous and
essentially the same across regions, over a global hemisphere, since good data has
been available, since the end of the Second World War in most Western countries
(Rancourt, 2020) (Rancourt et al., 2020) (Rancourt et al., 2021).
Indeed, as an aside, we consider that this empirical fact (geographic homogeneity of
synchronous mortality by time curves) represents a hard challenge against the theory
that viral respiratory diseases spread person-to-person by proximity or “contact” and
that such spread drives epidemics and pandemics, at the population level.
We quantify the said geographical heterogeneity of the COVID-era mortality by time
below, but first we illustrate it further with direct comparisons of the ACM-SB/w/pop
curves for states in different regions, with different cvp1-smp1-cvp2 patterns.
Figure 10 shows ACM-SB/w/pop for one state from each of the following
cvp1-smp1-cvp2 patterns: California (SSL), Florida (0LL), Michigan (LSLx), Nevada
(0SL), New York (L0M), South Dakoda (00L).
51
Figure 10a. Difference between all-cause mortality and summer baseline mortality by
week normalized by population for California, Florida, Michigan, Nevada, New York and
South Dakota from 2013 to 2021. Data are displayed from week-1 of 2013 to week-37 of 2021.
The dashed line emphasizes the zero. ACM data were retrieved from the CDC (CDC, 2021a)
and population data were retrieved from the US Census Bureau (US Census Bureau, 2021a),
as described in Table 1. SB was estimated as described in section 2.
Figure 10b. Difference between all-cause mortality and summer baseline mortality by
week normalized by population for California, Florida, Michigan, Nevada, New York and
South Dakota from 2013 to 2019. Data are displayed from week-1 of 2013 to week-52 of 2019.
52
The dashed line emphasizes the zero. ACM data were retrieved from the CDC (CDC, 2021a)
and population data were retrieved from the US Census Bureau (US Census Bureau, 2021a),
as described in Table 1. SB was estimated as described in section 2.
Figure 10c. Difference between all-cause mortality and summer baseline mortality by
week normalized by population for California, Florida, Michigan, Nevada, New York and
South Dakota from 2019 to 2021. Data are displayed from week-1 of 2019 to week-37 of 2021.
The dashed line emphasizes the zero. ACM data were retrieved from the CDC (CDC, 2021a)
and population data were retrieved from the US Census Bureau (US Census Bureau, 2021a),
as described in Table 1. SB was estimated as described in section 2.
Figure 11 makes the same kind of comparison for states that have large cvp1 features:
Colorado (LSL), Connecticut (L0M), Illinois (LSL), Louisiana (LLL), New Jersey (L0M),
New York (L0M).
53
Figure 11a. Difference between all-cause mortality and summer baseline mortality by
week normalized by population for Colorado, Connecticut, Illinois, Louisiana, New Jersey
and New York from 2013 to 2021. Data are displayed from week-1 of 2013 to week-37 of
2021. The dashed line emphasizes the zero. ACM data were retrieved from the CDC (CDC,
2021a) and population data were retrieved from the US Census Bureau (US Census Bureau,
2021a), as described in Table 1. SB was estimated as described in section 2.
Figure 11b. Difference between all-cause mortality and summer baseline mortality by
week normalized by population for Colorado, Connecticut, Illinois, Louisiana, New Jersey
and New York from 2013 to 2019. Data are displayed from week-1 of 2013 to week-52 of
54
2019. The dashed line emphasizes the zero. ACM data were retrieved from the CDC (CDC,
2021a) and population data were retrieved from the US Census Bureau (US Census Bureau,
2021a), as described in Table 1. SB was estimated as described in section 2.
Figure 11c. Difference between all-cause mortality and summer baseline mortality by
week normalized by population for Colorado, Connecticut, Illinois, Louisiana, New Jersey
and New York from 2019 to 2021. Data are displayed from week-1 of 2019 to week-37 of
2021. The dashed line emphasizes the zero. ACM data were retrieved from the CDC (CDC,
2021a) and population data were retrieved from the US Census Bureau (US Census Bureau,
2021a), as described in Table 1. SB was estimated as described in section 2.
3.6. ACM-SB by cycle-year (winter burden, WB) by population (WB/pop), USA and
state-to-state variations
Next, we analyse ACM-SB/w in terms of integrated intensities over cycle-years. By
definition, the said integrated intensity is the “winter burden”, WB, for the given cycle-
year. WB is the excess (above-SB) mortality per cycle-year. We normalize WB by
population, WB/pop, in order to make state-to-state and state-to-nation comparisons.
Figure 12a shows the WB/pop, for cycle-years 2014 to 2021 (cycle-year 2021 contains
and is approximately centered on January 2021, and so on), for the entire continental
55
USA (49 states). We see the seasonal (year to year) variations 2014-2019, followed by
the large COVID-era increase 2020-2021, which echoes the large 2020 calendar-year
increase shown in Figures 1 and 4.
Figure 12a. Winter burden normalized by population in the USA for cycle-years 2014 to
2021. The cycle-year starts on week-31 of a calendar-year (beginning of August) and ends on
week-30 of the next calendar-year (end of July). ACM data were retrieved from the CDC (CDC,
2021a) and population data were retrieved from the US Census Bureau (US Census Bureau,
2021a), as described in Table 1. SB was estimated and WB calculated as described in section
2.
Figure 12b shows WB/pop versus cycle-year (2014-2021), for all the continental USA
states on the same graph.
56
Figure 12b. Winter burden normalized by population for each of the continental states of
the USA for cycle-years 2014 to 2021. The cycle-year starts on week-31 of a calendar-year
(beginning of August) and ends on week-30 of the next calendar-year (end of July). The 49
continental states include the District of Columbia and exclude Alaska and Hawaii. ACM data
were retrieved from the CDC (CDC, 2021a) and population data were retrieved from the US
Census Bureau (US Census Bureau, 2021a), as described in Table 1. SB was estimated and
WB calculated as described in section 2.
Figure 12c shows WB/pop versus cycle-year (2014-2021) for the “0LL” group of
Southern states (having a cvp1-smp1-cvp2 0LL pattern), and for Louisiana, which has
the cvp1-smp1-cvp2 “LLL” pattern, on the same graph. We note a larger 2020 WB/pop
value for Louisiana, than would be expected for a Southern state, because its large
LLL-pattern cvp1 feature increases its 2020 WB/pop value.
57
Figure 12c. Winter burden normalized by population in Alabama, Arizona, Florida,
Louisiana, Mississippi, South Carolina and Texas for cycle-years 2014 to 2021. The cycle-
year starts on week-31 of a calendar-year (beginning of August) and ends on week-30 of the
next calendar-year (end of July). ACM data were retrieved from the CDC (CDC, 2021a) and
population data were retrieved from the US Census Bureau (US Census Bureau, 2021a), as
described in Table 1. SB was estimated and WB calculated as described in section 2.
Figure 12d shows WB/pop versus cycle-year (2014-2021) for the “L0M” group of North-
East coastal states (having a cvp1-smp1-cvp2 L0M pattern), including Maryland, which
has a limit behaviour to be included in this group. Since this group has exceptionally
large cvp1 features, we see that generally the WB-2020 is larger than the WB-2021.
58
Figure 12d. Winter burden normalized by population in Connecticut, Maryland,
Massachusetts, New Jersey and New York for cycle-years 2014 to 2021. The cycle-year
starts on week-31 of a calendar-year (beginning of August) and ends on week-30 of the next
calendar-year (end of July). ACM data were retrieved from the CDC (CDC, 2021a) and
population data were retrieved from the US Census Bureau (US Census Bureau, 2021a), as
described in Table 1. SB was estimated and WB calculated as described in section 2.
Figure 12b shows that, like the ACM-SB/w/pop curves themselves would suggest
(Figures 10 and 11), the state-to-state spread in WB/pop values is much larger in the
COVID-era than in the previous decade or so. We can illustrate this pre-COVID/COVID-
era difference by plotting the frequency distribution of state-to-state values of WB/pop
for each cycle-year. These distributions are shown together in Figure 13.
59
Figure 13. Frequency distributions of state-to-state values of WB/pop for each cycle-year,
2014-2021, as indicated by the colour scheme. Each distribution is normalized to 49, the
number of continental USA states (including District of Columbia, excluding Alaska and Hawaii).
A bin-width of 2.5E−4 deaths/pop was used. The cycle-year starts on week-31 of a calendar-
year (beginning of August) and ends on week-30 of the next calendar-year (end of July). ACM
data were retrieved from the CDC (CDC, 2021a) and population data were retrieved from the
US Census Bureau (US Census Bureau, 2021a), as described in Table 1. SB was estimated
and WB calculated as described in section 2.
Here (Figure 13), it is interesting to note that the six pre-COVID-era cycle-years (2014-
2019) fall into two distinct distribution types, with the same widths but positions differing
by a set amount, corresponding to “light” (2014, 2016, 2019; less deadly winter) and
“heavy” (2015, 2017, 2018; deadlier winter) years that are also recognized in the
ACM/w or ACM-SB/w patterns themselves (e.g., Figures 5 and 6).
By comparison, the distribution for cycle-year 2020 has larger WB/pop values and a tail
that extends far towards even larger values. The distribution for cycle-year 2021 is
exceedingly wide and extends to extremely large values.
60
Properties of the frequency distributions (Figure 13) can be quantified as follows. For
each distribution (for a given cycle-year) we calculate: the average (“av”), the median
(“med”), the standard deviation (“sd”), and the difference “av-med”. The latter difference
av-med is related to the magnitude of the asymmetry of the distribution, and its sign
indicates whether any extended tail extends toward small (negative) or large (positive)
WB/pop values. These four parameters (av, med, sd, av-med) are shown versus cycle-
year in Figure 14.
Figure 14. Statistical parameters of the WB/pop distributions of the 49 continental states
of the USA for cycle-years 2014 to 2021. The 49 continental states include the District of
Columbia and exclude Alaska and Hawaii. The cycle-year starts on week-31 of a calendar-year
(beginning of August) and ends on week-30 of the next calendar-year (end of July). ACM data
were retrieved from the CDC (CDC, 2021a) and population data were retrieved from the US
Census Bureau (US Census Bureau, 2021a), as described in Table 1. SB was estimated and
WB calculated as described in section 2.
Here (Figure 14), the variations of “av” and “med” are generally those expected, given
the behaviour of WB/pop versus cycle-year for the entire continental USA (Figure 12a).
The “sd” (Figure 14) has a remarkably constant pre-COVID-era (prior to 2020) value of
approximately 1.6(1.2—1.9 range)E−4 deaths/pop, and then shoots up to 4.3E−4
61
(2020) and 6.1E−4 (2021) deaths/pop. In other words, the COVID-era is characterized
by an anomalously large state-to-state heterogeneity in WB/pop values, an
approximately 4-fold increase in absolute magnitude.
In fact, using WB/pop masks the actual state-wise heterogeneity, since the COVID-era
features cvp1 and smp1 have a much larger intrinsic (relative) heterogeneity than WB.
The said large heterogeneity is evident in the ACM-SB/w/pop data itself (Figures 10 and
11), but let us quantify it, and let us examine “asymmetry” (presence of tails) as well.
We use the dimensionless parameters sd/av and (av-med)/av, which are as follows.
Breadth and asymmetry of state-wise distributions of integrated deaths
feature sd/av (av-med)/av
pre-COVID-era WB/pop
2014-2019
0.20—0.31 -0.03—+0.04
2020 WB/pop 0.39 +0.14
cvp1/pop 0.79 +0.27
smp1/pop 0.67 +0.17
cvp2/pop 0.28 0.00
2021 WB/pop 0.30 -0.05
Table 2. Breadth and asymmetry of state-wise distributions of integrated deaths for the
pre-COVID-era WB/pop, and for features in the COVID-era. Features in the COVID-era
include 2020 WB/pop, cvp1/pop, smp1/pop, cvp2/pop and 2021 WB/pop.
The state-wise heterogeneity of cvp1 is massive (sd/av: 0.79 compared to ~0.25)
((av-med)/av: +0.27 compared to ~+0.01), since cvp1 consists of essentially one
extreme region in the North-East coastal states. The state-wise heterogeneity of smp1
is large (sd/av: 0.67 compared to ~0.25) ((av-med)/av: +0.17 compared to ~+0.01),
since smp1 consists of essentially an extreme region in the Southern states.
62
We have observed such COVID-era jurisdictional heterogeneity in many countries, and
country-wise in Europe, and we have argued that it is contrary to pandemic behaviour,
and contrary to any (1945-2021) season of viral respiratory disease burden in the
Northern hemisphere, and arises mainly from jurisdictional differences in applied
medical and government responses to the pronouncement of a pandemic (Rancourt,
2020) (Rancourt et al., 2020) (Rancourt et al., 2021).
In contrast, cvp2, which is entirely within the 2021 cycle-year and is the cycle-year’s
main (winter) feature, has normal pre-COVID-era state-wise homogeneity (sd/av: 0.28
compared to 0.20—0.31) ((av-med)/av: 0.00 compared to -0.03—+0.04). This suggests
that cvp2 is not affected by any widely different state-to-state applied responses, but
rather is the result of a broad, sustained, and state-wise homogenous stress on the
USA population.
3.7. Geographical distribution and correlations between COVID-era above-SB
seasonal deaths: cvp1 (spring-2020), smp1 (summer-2020) and cvp2 (fall-winter-
2020-2021)
Recall that Figure 7 shows how we integrate to obtain the total above-SB deaths in
each of the operationally defined features cvp1, smp1 and cvp2. Since the peak
positions are operationally the same for all states (barring the extra peak for Michigan),
we use the same delimiting weeks throughout, those shown in Figure 7. We normalize
the state-wise deaths by state-wise population, in order to allow state-to-state
comparisons.
Figure 15 shows a map of cvp1/pop for the continental states of the USA.
63
Figure 15. Map of the intensity of the cvp1 mortality normalized by population for the
continental USA. Continental USA includes the District of Columbia and excludes Alaska and
Hawaii. The cvp1 feature is the integrated deaths of ACM-SB between week-11 of 2020 and
week-25 of 2020, inclusively. The darker the blue, the more intense the cvp1/pop. ACM data
were retrieved from the CDC (CDC, 2021a) and population data were retrieved from the US
Census Bureau (US Census Bureau, 2021a), as described in Table 1. SB was estimated as
described in section 2.
Here, we see that a cluster of North-East coastal states were essentially the only
intense hot spot; and notable other states, including Louisiana, Illinois and Michigan, to
a lesser degree. In fact, some 34 of the USA states do not have a resolved or
detectable or significant cvp1 feature. We have described this previously (Rancourt,
2020) (Rancourt et al., 2020). We have argued that the cvp1 feature (the “covid peak”)
is highly jurisdictionally heterogeneous, has a start synchronous with the 11 March 2020
WHO declaration of a pandemic, and is present throughout the mid-latitude Northern
hemisphere, because it is caused by the medical and government responses to the
declaration of a pandemic, especially in hospitals and care homes (Rancourt, 2020)
(Rancourt et al., 2020) (Rancourt et al., 2021). One can say with certainty that there
64
was no detectable or significant “first wave” in most of the USA, a phenomenon which is
contrary to the very concept of a pandemic (Rancourt et al., 2021).
Figure 16 shows a map of smp1/pop for the continental states of the USA.
Figure 16. Map of the intensity of the smp1 mortality normalized by population for the
continental USA. Continental USA includes the District of Columbia and excludes Alaska and
Hawaii. The smp1 feature is the integrated deaths of ACM-SB between week-26 of 2020 and
week-39 of 2020, inclusively. The darker the red, the more intense the smp1/pop. ACM data
were retrieved from the CDC (CDC, 2021a) and population data were retrieved from the US
Census Bureau (US Census Bureau, 2021a), as described in Table 1. SB was estimated as
described in section 2.
This is a remarkable map, which shows that the above-SB deaths in the summer of
2020 were concentrated in the Southern states of Arizona, Texas, Louisiana,
Mississippi, Alabama, Florida and South Carolina. These results can be understood in
terms of climatic, socio-economic and population health effects, as shown below. The
results (Figure 16) are inconsistent with the theoretical concept of a viral respiratory
disease pandemic. Furthermore, no previous large anomalous burden of all-cause
65
mortality has ever been concentrated in the Southern states, in one season, in the
modern history of epidemiology for the USA.
There is no point showing a map of cvp2/pop for the continental states of the USA,
because we showed above that the state-wise distribution of cvp2/pop is essentially
homogeneous (Table 2). A map of cvp2/pop does not show any recognizable pattern.
Next, we examine whether there are any correlations or anti-correlations between the
outcomes cvp1, smp1 and cvp2; and also smp2. Plots of one versus the other are as
follows, in Figure 17.
Figure 17a. smp1/pop versus cvp1/pop. Each point is for one continental USA state. The
colour-code of the 49 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
66
Figure 17b. cvp2/pop versus cvp1/pop. Each point is for one continental USA state. The
colour-code of the 49 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
Figure 17c. cvp2/pop versus smp1/pop. Each point is for one continental USA state. The
trend line is meant merely to illustrate the correlation discussed in the text. It results from the
usual least squares fit, using all the points in the graph. The colour-code of the 49 continental
states is shown in section 2. Data were retrieved and calculations made as described in section
2.
67
Figure 17a shows that near-zero values of smp1/pop occur for the largest values of
cvp1/pop, and that most large values of smp1/pop occur for small values of cvp1/pop.
Similarly, Figure 17b shows that near-zero values of cvp2/pop occur for the largest
values of cvp1/pop, and that most large values of cvp2/pop occur for small values of
cvp1/pop.
This shows that the states with extremely large values of cvp1/pop (New York, New
Jersey, Connecticut, Massachusetts… mainly the L0M pattern) had small (cvp2) or
near-zero (smp1) values of mortality in the seasons that followed (summer-2020, fall-
winter-2020-2021). Possible explanations include: the so-called “dry tinder” effect, in
which those likely to die would have already died in the first “wave”, or socio-geo-
economic and climatic factors that give large smp1 and cvp2 are absent in those states
that have the largest cvp1 peaks. Our analysis shows that the latter explanation is more
likely. Indeed, different age groups, social classes (poverty, obesity) and state
jurisdictions predominantly contribute to cvp1 versus smp1 and cvp2. A dry tinder effect
interpretation for cvp1/smp1-cvp2 is not compatible with the many observed
correlations.
A notable exception (outlier) in the smp1-cvp1 relation (Figure 17a) is Louisiana, which
has both large cvp1 and large smp1. We have interpreted large values of cvp1 (“covid
peak”), occurring heterogeneously and synchronously around the world, as being due to
local-jurisdictional aggressive immediate medical and government responses to the
11 March 2020 WHO pronouncement of a pandemic (Rancourt, 2020) (Rancourt et al.,
2020) (Rancourt et al., 2021). New York City and New York state directives are the
defining examples of such aggression. There is circumstantial evidence that Louisiana
has a medico-government culture approaching that of New York: “Louisiana's largest
hospital system will impose fee on employees if their spouse is unvaccinated”, Blaze
media, 01 October 2021, https://archive.ph/sDfL2.
Figure 17c shows that there is a correlation between cvp2/pop and smp1/pop. Such a
correlation, as opposed to an anti-correlation, is contrary to a “dry tinder” effect
68
occurring between summer-2020 and fall-winter-2020-2021. Rather, it suggests that
some or all of the same socio-geo-economic and climatic effects impact the mortality in
both seasons.
The summer-2021 feature smp2 behaves similarly to smp1 (summer-2020) in many
regards, although it starts later in the summer, and smp2/pop is correlated to smp1/pop,
as shown in Figure 17d.
Figure 17d. smp2/pop versus smp1/pop. Each point is for one continental USA state.
Connecticut, North Carolina and West Virginia are removed from the graph as there are not
enough consolidated data points in ACM/w for smp2 for those states (see Appendix). The trend
line is meant merely to illustrate the correlation discussed in the text. It results from the usual
least squares fit, using all the points in the graph. The colour-code of the 49 continental states is
shown in section 2. Data were retrieved and calculations made as described in section 2.
Figure 18 shows the same data as in Figure 17c, but with added circle-symbol-size
(radius) determined by cvp1/pop.
69
Figure 18. cvp2/pop versus smp1/pop, with the radius size determined by cvp1/pop. Each
point is for one continental USA state. The colour-code of the 49 continental states is shown in
section 2. Data were retrieved and calculations made as described in section 2.
We note that the largest values of cvp1/pop (by state) are clustered at small values of
both smp1/pop and cvp2/pop, with Louisiana as the main exception, followed by
Mississippi.
3.8. Associations of COVID-era mortality outcomes with socio-geo-economic and
climatic variables
The data, in which quantitative mortality outcomes (cvp1, smp1, cvp2, WB) are known
by state, can be compared with state-wise or state-specific socio-geo-economic and
climatic variables, in a search for correlations or relations, since all 49 diverse
continental USA states can be used. This is a unique opportunity to identify factors
which may cause or contribute to the excess (above-SB) USA mortality during the
COVID-era.
70
We found three variables that appear to be determinative of COVID-era summer-2020
(smp1) and fall-winter-2020-2021 (cvp2) excess (above-SB) mortality in the USA. These
are:
1. Climatic temperature (summer-period heatwave effect) (smp1)
2. Poverty (smp1 and cvp2)
3. Obesity (smp1 and cvp2)
The variables are somewhat correlated to each other, but have a significant degree of
independence (one can be obese and rich, etc.). We found that using the product
“OB.PV” of obesity (OB) and poverty (PV) gives a stronger correlation than either
variable alone (being both obese and poor is deadlier than being either obese or poor).
We found that climatic temperature — evaluated using either maximum temperature
(Tmax) or average temperature (Tav), either averaged in July-August-2020 or averaged
over a calendar-year — is highly predictive of the geographical location of smp1
mortality (the hottest states were the most deadly in summer-2020, and dramatically
so).
None of the variables (OB, PV, Tmax) that correlate with smp1 and cvp2 correlate with
cvp1, which shows distinctly different death-causing phenomena in the two periods
(cvp1 versus smp1-cvp2) in the COVID-era. We interpret cvp1 as being due to the
immediate aggressive medical and government measures, whereas later deaths are
apparently due to accumulated social and psychological chronic stress, combined with
climatic stress, and affect younger individuals in broader age groups.
The latter age-dependence was shown by examining correlations between mortality
outcomes and population age structure, by state. The smp1 feature (above-SB deaths
in summer-2020) is uniquely anti-correlated with age of the state-wise population, which
is contrary to WB mortality behaviour in all studied pre-COVID-era cycle-years, 2014-
2019, and contrary to viral respiratory disease epidemiology.
71
Throughout this study, we compare our COVID-era results with a similar search for
correlations in WB/pop mortality outcome in given cycle-years occurring prior to the
COVID-era. Contrary to deaths in the COVID-era, normal epidemiology of the
unperturbed society shows no state-to-state correlations of winter burdens with obesity,
poverty or climatic temperature, whatsoever, in any of the six specific cycle-years 2014-
2019. The only “normal era” correlation we find is with age structure, and it is persistent
from year to year. The same is true for many more cycle-years for France, and so on. It
seems clear to us that the variables obesity, poverty and climatic temperature become
determinative, and have a disproportionate and immediate deadly impact, only in the
significantly socio-economically perturbed and stressed population of the COVID-era
measures.
Here are the details, as follows.
Obesity
Figure 19 shows the scatter plots for obesity (OB), defined as the prevalence of self-
reported obesity among U.S. adults (CDC, 2021e).
72
Figure 19a. cvp1/pop versus obesity. Each point is for one continental USA state. The colour-
code of the 49 continental states is shown in section 2. Data were retrieved and calculations
made as described in section 2.
There is no discernable trend between cvp1/pop and OB.
Figure 19b. smp1/pop versus obesity. Each point is for one continental USA state. The trend
line is meant merely to illustrate the correlation discussed in the text. It results from the usual
73
least squares fit, using all the points in the graph. The colour-code of the 49 continental states is
shown in section 2. Data were retrieved and calculations made as described in section 2.
There is a positive trend between smp1/pop and OB.
Figure 19c. cvp2/pop versus obesity. Each point is for one continental USA state. The trend
line is meant merely to illustrate the correlation discussed in the text. It results from the usual
least squares fit, using all the points in the graph. The colour-code of the 49 continental states is
shown in section 2. Data were retrieved and calculations made as described in section 2.
There is a positive trend between cvp2/pop and OB.
74
Figure 19d. WB/pop for cycle-year 2019 versus obesity. Each point is for one continental
USA state. The colour-code of the 49 continental states is shown in section 2. Data were
retrieved and calculations made as described in section 2.
There is no correlation whatsoever. This is true for all pre-COVID-era cycle-years, 2014-
2019 (data not shown). “Normal-era” winter burden deaths above-SB have no relation to
obesity, on a state-wise basis.
75
Figure 19e. WB/pop for COVID-era cycle-year 2020 versus obesity. Each point is for one
continental USA state. The colour-code of the 49 continental states is shown in section 2. Data
were retrieved and calculations made as described in section 2.
Excluding the six states with highest 2020 WB/pop values and OB < 31 % (Connecticut,
District of Columbia, Massachusetts, New Jersey, New York, Rhode Island), there is a
positive trend for the remaining states. This is consistent with the fact that 2020 cycle-
year includes both cvp1 and approximately half of smp1, and that the excluded states
have extremely large cvp1/pop values in mostly wealthy states.
76
Figure 19f. WB/pop for COVID-era cycle-year 2021 versus obesity. Each point is for one
continental USA state. The trend line is meant merely to illustrate the correlation discussed in
the text. It results from the usual least squares fit, using all the points in the graph. The colour-
code of the 49 continental states is shown in section 2. Data were retrieved and calculations
made as described in section 2.
There is a positive trend between WB/pop for COVID-era cycle-year 2021 and OB.
Poverty
Figure 20 shows the scatter plots for poverty (PV), defined as the estimated percent of
people of all ages in poverty (US Census Bureau, 2021d).
77
Figure 20a. cvp1/pop versus poverty. Each point is for one continental USA state. The colour-
code of the 49 continental states is shown in section 2. Data were retrieved and calculations
made as described in section 2.
There is no discernable trend between cvp1/pop and PV.
Figure 20b. smp1/pop versus poverty. Each point is for one continental USA state. The trend
line is meant merely to illustrate the correlation discussed in the text. It results from the usual
78
least squares fit, using all the points in the graph. The colour-code of the 49 continental states is
shown in section 2. Data were retrieved and calculations made as described in section 2.
There is a positive trend between smp1/pop and PV.
Figure 20c. cvp2/pop versus poverty. Each point is for one continental USA state. The trend
line is meant merely to illustrate the correlation discussed in the text. It results from the usual
least squares fit, using all the points in the graph. The colour-code of the 49 continental states is
shown in section 2. Data were retrieved and calculations made as described in section 2.
There is a positive trend between cvp2/pop and PV.
79
Figure 20d. WB/pop for cycle-year 2019 versus poverty. Each point is for one continental
USA state. The colour-code of the 49 continental states is shown in section 2. Data were
retrieved and calculations made as described in section 2.
There is no correlation whatsoever. This is true for all pre-COVID-era cycle-years, 2014-
2019 (data not shown). “Normal-era” winter burden deaths above-SB have no relation to
poverty, on a state-wise basis.
80
Figure 20e. WB/pop for COVID-era cycle-year 2020 versus poverty. Each point is for one
continental USA state. The colour-code of the 49 continental states is shown in section 2. Data
were retrieved and calculations made as described in section 2.
Excluding the four states with highest 2020 WB/pop values (Connecticut,
Massachusetts, New Jersey, New York), there is a positive trend for the remaining
states. This is consistent with the fact that 2020 cycle-year includes both cvp1 and
approximately half of smp1, and that the excluded states have extremely large cvp1/pop
values in mostly wealthy states.
81
Figure 20f. WB/pop for COVID-era cycle-year 2021 versus poverty. Each point is for one
continental USA state. The trend line is meant merely to illustrate the correlation discussed in
the text. It results from the usual least squares fit, using all the points in the graph. The colour-
code of the 49 continental states is shown in section 2. Data were retrieved and calculations
made as described in section 2.
There is a positive trend between WB/pop for COVID-era cycle-year 2021 and PV. The
outlier at 13.6 % poverty is North Carolina, which is an artifact of incomplete data for the
final weeks for this state (see Appendix).
Climatic temperature
One of the most striking results of our study is that the summer-2020 excess
(above-SB) mortality is concentrated in Southern states (Figure 16). Excess summer
mortality is striking in itself because viral respiratory diseases barely transmit in humid
summer climates (aerosol particles are not stable in high absolute humidity: Harper,
1961; Shaman et al., 2010), and summers “always” exhibit seasonal lows of mortality in
mid-latitude regions, seasonally inverted in the Southern hemisphere. Yet, here in the
USA, there was an actual peaked maximum in ACM/w in the summer-2020 (Figures 5,
6, 7, 9, 10, and Appendix).
82
The geographical pattern of summer-2020 excess (above-SB) mortality, on a map of the
USA (Figure 16), is remarkably well predicted by climatic temperature, shown in
Figure 21.
Figure 21. Mean daily average temperature: Mean of daily minimum and maximum,
averaged over the year, and for three decades (1970-2000). This represents “climatic mean
temperature” for the continental USA (spatial average is achieved using weighted cells, with the
available surface air weather stations). Source: Climate Atlas of the United States, developed by
NOAA's National Climatic Data Center in Asheville, NC., Version 2.0, CD-ROM, released
September 2002. Figure accessed at http://www.virginiaplaces.org/climate/ on 26 September
2021. (Typo: “< 70.0” should be “> 70.0”).
We illustrate this on a state-by-state basis, using the state-wise average August-2020
temperature, shown in Figure 22.
83
Figure 22. Average temperature, per state of the continental USA, for August 2020.
Continental USA excludes Alaska and Hawaii. The darker the red, the higher the average
temperature. Climatic temperature data were retrieved from the NOAA (NOAA, 2021), as
described in Table 1. (The reader is asked to compare this map with the map shown in Figure
16.)
Essentially the same pattern occurs for July 2020, or for any month, or for yearly
averages, or using daily maximum temperatures rather than daily average
temperatures. Basically, all the average temperatures (averages of daily averages, or
averages of daily maxima; on July or August, or on July and August, or on any
calendar-year or cycle-year) chosen to represent climatic temperature are highly
correlated to each other. For our purpose, these different averages are interchangeable.
The correlation between climatic temperature and summer-2020 excess (above-SB)
mortality (smp1/pop, by state) is illustrated in Figure 23, using the July-August 2020
average daily maximum temperature (averaged by state and over the two-month
period).
84
Figure 23. smp1/pop versus average daily maximum temperature over July and August
2020, Tmax Jul-Aug 2020. Each point is for one continental USA state, excluding District of
Columbia, for which no temperature data were available (NOAA, 2021). The trend line is meant
merely to illustrate the correlation discussed in the text. It results from the usual least squares
fit, using all the points in the graph. The colour-code of the other 48 continental states is shown
in section 2. Data were retrieved and calculations made as described in section 2.
There is a clear positive trend. Here (Figure 23), the four main high-smp1/pop-value
outliers are Mississippi, South Carolina, Alabama and Louisiana; whereas the three
main low-smp1/pop-value outliers are Massachusetts, Connecticut and New Jersey.
Such a trend between an excess (above-SB) mortality and mean temperature, per
state, does not exist, whatsoever, in the winter burden mortality (WB/pop) for any of the
pre-COVID-era cycle-years, 2014-2019 (data not shown).
Obesity, poverty, and climatic temperature
Next, we examine the above correlations further. Figure 24 shows that obesity (OB) and
poverty (PV) are somewhat correlated to each other.
85
Figure 24. Obesity versus poverty. Each point is for one continental USA state. The trend line
is meant merely to illustrate the correlation discussed in the text. It results from the usual least
squares fit, using all the points in the graph. The colour-code of the 49 continental states is
shown in section 2. Data were retrieved as described in section 2.
Given the above, we decided to try using the product of obesity and poverty (OB.PV) as
a variable. Figure 25 shows smp1/pop versus OB.PV, with added circle-symbol-size
(radius) determined by the July-August 2020 average daily maximum temperature
(averaged by state and over the two-month period).
86
Figure 25. smp1/pop versus the product of obesity and poverty (OB.PV), with the radius
size determined by Tmax Jul-Aug 2020. Each point is for one continental USA state,
excluding District of Columbia, for which no temperature data were available (NOAA, 2021). The
colour-code of the other 48 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
The correlation is excellent. Climatic temperature (circle size) also appears to be
correlated to OB.PV (Figure 25). Figure 26 shows the average of daily average
temperatures over the calendar-year 2020 (Tav 2020) versus OB.PV, with added circle-
symbol-size (radius) determined by the outcome smp1/pop.
87
Figure 26. Tav 2020 versus the product of obesity and poverty (OB.PV), with the radius
size determined by smp1/pop. Each point is for one continental USA state, excluding District
of Columbia, for which no temperature data were available (NOAA, 2021). The colour-code of
the other 48 continental states is shown in section 2. Data were retrieved as described in
section 2.
Figure 26 shows two things.
First, climatic temperature is correlated to the product OB.PV.
Second, a diagram of climatic temperature versus OB.PV provides a strong predictor of
whether there will be large summer mortality following an extended period of chronic
psychological stress applied to the population.
Age structure of the population
More than 60 % of COVID-assigned deaths in the USA occur in the 85+ years age
group (Kostoff et al., 2021; their Figure 1). The same is generally true of all viral
respiratory diseases in Western nations.
88
Figure 27 shows WB/pop versus percent of population consisting of 85+ year olds
(“85+/pop”), for each pre-COVID-era cycle-year, 2014-2019. The latter percentage more
than doubles across all states, from approximately 1.2 % to approximately 2.6 %.
Whereas the illustrated correlation is weak, it is persistently positive, having similar
slope magnitudes, across all cycle-years, except for cycle-year 2016 (Figure 27c) where
the nominally positive correlation (not shown) is not statistically meaningful.
Figure 27a. WB/pop versus 85+/pop for cycle-year 2014. Each point is for one continental
USA state. The trend line is meant merely to illustrate the correlation discussed in the text. It
results from the usual least squares fit, using all the points in the graph. The colour-code of the
49 continental states is shown in section 2. Data were retrieved and calculations made as
described in section 2. Outliers: Utah (bad data point in 2014), Wyoming (less populous state,
poor statistics, underestimation of SB).
89
Figure 27b. WB/pop versus 85+/pop for cycle-year 2015. Each point is for one continental
USA state. The trend line is meant merely to illustrate the correlation discussed in the text. It
results from the usual least squares fit, using all the points in the graph. The colour-code of the
49 continental states is shown in section 2. Data were retrieved and calculations made as
described in section 2. The outlier Wisconsin is due to bad data points in 2015 for this state (see
Appendix).
90
Figure 27c. WB/pop versus 85+/pop for cycle-year 2016. Each point is for one continental
USA state. The colour-code of the 49 continental states is shown in section 2. Data were
retrieved and calculations made as described in section 2.
Figure 27d. WB/pop versus 85+/pop for cycle-year 2017. Each point is for one continental
USA state. The trend line is meant merely to illustrate the correlation discussed in the text. It
results from the usual least squares fit, using all the points in the graph. The colour-code of the
49 continental states is shown in section 2. Data were retrieved and calculations made as
described in section 2. Outlier: Wyoming (less populous state, poor statistics).
91
Figure 27e. WB/pop versus 85+/pop for cycle-year 2018. Each point is for one continental
USA state. The trend line is meant merely to illustrate the correlation discussed in the text. It
results from the usual least squares fit, using all the points in the graph. The colour-code of the
49 continental states is shown in section 2. Data were retrieved and calculations made as
described in section 2. Outliers: West Virginia (underestimation of SB, overestimation of WB),
Montana (reverse).
Figure 27f. WB/pop versus 85+/pop for cycle-year 2019. Each point is for one continental
USA state. Outlier: District of Columbia (small state, poor statistics).
92
The same phenomenon (positive correlation of WB/pop with population fraction of the
age group, in the pre-COVID-era cycle-years) occurs for all the older age groups: 45-54,
55-64, 65-74, 75-84, and 85+ ages. The correlation is then negative (anti-correlation) for
35-44 years, and not discernable for younger age groups (data not shown).
This age-dependence of winter burden mortality was expected, and is well known.
Young people do not generally die of viral respiratory diseases that are prevalent in the
winter.
In the COVID-era, cvp1/pop does not have a statistically meaningful correlation with
85+/pop, as shown in Figure 28a. It might best be described as no correlation
whatsoever for states having essentially zero-magnitude cvp1/pop values, and several
randomly placed outliers above the group having near-zero values of cvp1/pop. This is
consistent with the idea that the cvp1 feature is predominantly due to the jurisdiction-
specific response to the declaration of a pandemic.
Surprisingly, however, the summer-2020 excess (above-SB) mortality (smp1/pop) has
an anti-correlation (“neg-cor”) with 85+/pop, again with significant outliers, as shown in
Figure 28b; and the fall-winter-2020-2021 mortality (cvp2/pop) has no discernable
correlation with 85+/pop, as shown in Figure 28c. Correspondingly, the WB/pop versus
85+/pop has a positive correlation for cycle-year 2020 (Figure 28d), and a uniquely
strong negative (anti-)correlation for cycle-year 2021 (Figure 28e).
93
Figure 28a. cvp1/pop versus 85+/pop. Each point is for one continental USA state. The
colour-code of the 49 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
Figure 28b. smp1/pop versus 85+/pop. Each point is for one continental USA state. The trend
line is meant merely to illustrate the correlation discussed in the text. It results from the usual
least squares fit, using all the points in the graph. The colour-code of the 49 continental states is
shown in section 2. Data were retrieved and calculations made as described in section 2.
94
Figure 28c. cvp2/pop versus 85+/pop. Each point is for one continental USA state. The
colour-code of the 49 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
Figure 28d. WB/pop versus 85+/pop for cycle-year 2020. Each point is for one continental
USA state. The trend line is meant merely to illustrate the correlation discussed in the text. It
results from the usual least squares fit, using all the points in the graph. The colour-code of the
49 continental states is shown in section 2. Data were retrieved and calculations made as
described in section 2.
95
Figure 28e. WB/pop versus 85+/pop for cycle-year 2021. Each point is for one continental
USA state. The trend line is meant merely to illustrate the correlation discussed in the text. It
results from the usual least squares fit, using all the points in the graph. The colour-code of the
49 continental states is shown in section 2. Data were retrieved and calculations made as
described in section 2.
The same types of state-wise correlations for smp1 and cvp2 occur for other age groups
also (data not shown). In summary, as follows.
• smp1/pop: pos-cor with -18/pop, neg-cor with 55-64/pop, neg-cor with 85+/pop
• cvp2/pop: pos-cor with -18/pop, neg-cor with 45-54/pop, neg-cor with 55-64/pop
Population density
The USA state-wise data offers a unique opportunity to examine the relation between
population density (“popD”) (number of inhabitants per unit surface area) and excess
(above-SB) mortality, since popD varies by more than two orders of magnitude, from
Wyoming to New Jersey.
96
Figure 29 shows WB/pop versus popD, for each pre-COVID-era cycle-year, 2014-2019.
Here (Figure 29), there is no detectable, statistically significant, correlation between
winter burden mortality (WB/pop) and popD, in any of the years studied.
Given the synchronous mortality patterns, state-to-state (Figures 10 and 11, for the pre-
COVID-era cycle-years), and given present theoretical understanding of contagious
disease transmission (Hethcote, 2000) (McCallum et al., 2001), our results (Figure 29)
impose constraints on models of the phenomenon of seasonal mortality, and strongly
suggest that the seasonal preponderance of viral respiratory diseases is not the result
of transmission and spread by person-to-person “contact”.
Figure 29a. WB/pop for cycle-year 2014 versus population density. Each point is for one
continental USA state, excluding District of Columbia, which has an extreme density. The
colour-code of the other 48 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
97
Figure 29b. WB/pop for cycle-year 2015 versus population density. Each point is for one
continental USA state, excluding District of Columbia, which has an extreme density. The
colour-code of the other 48 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
Figure 29c. WB/pop for cycle-year 2016 versus population density. Each point is for one
continental USA state, excluding District of Columbia, which has an extreme density. The
colour-code of the other 48 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
98
Figure 29d. WB/pop for cycle-year 2017 versus population density. Each point is for one
continental USA state, excluding District of Columbia, which has an extreme density. The
colour-code of the other 48 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
Figure 29e. WB/pop for cycle-year 2018 versus population density. Each point is for one
continental USA state, excluding District of Columbia, which has an extreme density. The
colour-code of the other 48 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
99
Figure 29f. WB/pop for cycle-year 2019 versus population density. Each point is for one
continental USA state, excluding District of Columbia, which has an extreme density. The
colour-code of the other 48 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
This result (Figure 29) is in contrast to correlations observed for the COVID-era, where
mortality has strong correlations and anti-correlations with popD. In the COVID-era,
cvp1/pop has a large positive correlation with popD, although the New York outlier is
significant, as shown in Figure 30a. While, on the other hand, both the summer-2020
excess (above-SB) mortality (smp1/pop) and the fall-winter-2020-2021 mortality
(cvp2/pop) have anti-correlations with popD (Figures 30b and 30c, respectively).
Correspondingly, the WB/pop versus popD has a large positive correlation for cycle-
year 2020, with New York outlier (Figure 30d), and a strong negative (anti-)correlation
for cycle-year 2021 (Figure 30e).
100
Figure 30a. cvp1/pop versus population density. Each point is for one continental USA state,
excluding District of Columbia, which has an extreme density. The colour-code of the other 48
continental states is shown in section 2. Data were retrieved and calculations made as
described in section 2.
Figure 30b. smp1/pop versus population density. Each point is for one continental USA
state, excluding District of Columbia, which has an extreme density. The colour-code of the
other 48 continental states is shown in section 2. Data were retrieved and calculations made as
described in section 2.
101
Figure 30c. cvp2/pop versus population density. Each point is for one continental USA state,
excluding District of Columbia, which has an extreme density. The colour-code of the other 48
continental states is shown in section 2. Data were retrieved and calculations made as
described in section 2.
Figure 30d. WB/pop for cycle-year 2020 versus population density. Each point is for one
continental USA state, excluding District of Columbia, which has an extreme density. The
colour-code of the other 48 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
102
Figure 30e. WB/pop for cycle-year 2021 versus population density. Each point is for one
continental USA state, excluding District of Columbia, which has an extreme density. The
colour-code of the other 48 continental states is shown in section 2. Data were retrieved and
calculations made as described in section 2.
We do not believe that a new virus causes the unprecedented correlations of mortality
with popD, in the COVID-era. Rather, we interpret the results to mean that high-
population-density states, with large urban centers would have had similar institutional
structures and policy responses, generally different from those in low-population-density
states. Also, the Southern states with large smp1 mortality due to climatic temperature,
poverty and obesity are lower population-density states.
One pair of states, New York and Florida, strikingly demonstrates that population
density in itself is not a controlling factor. Whereas these two states have essentially
identical values of popD, they have diametrically opposed values of cvp1 mortality
(Figure 30a), and, in the opposite order, of summer-2020 (smp1) mortality (Figure 30b).
Indeed, the correlations with popD in the COVID-era are an indication that the mortality
is not the result of viral respiratory diseases, and rather that the mortality is tied to
institutional, governmental, socio-economic and climatological differences.
103
All-cause mortality by week (ACM/w) by age group
The age dependencies of mortality in the pre-COVID and COVID-eras are shown more
directly than only examining state-wise correlations, by examining ACM/w itself for the
USA (no state-wise resolution is available) by age group, as follows.
We represent the ACM/w for the USA (Figure 5) by age group, for the two age groups
18-64 and 65+ ages, in Figure 32a. Here (Figure 32a), we have multiplied the ACM/w
for the 18-64 years age group by a factor sufficient to make the ACM/w equal to that for
the 65+ years age group, in the summer-2014 trough. This is equivalent to multiplying
the population of the 18-64 years age group until the deaths per week are equal to the
deaths per week in the 65+ years age group, in the summer-2014 trough. This is done
to better visualize and compare the relative seasonal changes in mortality between the
two age groups.
Figure 32a. All-cause mortality by week in the USA for the 18-64 and 65+ years age
groups (light blue and dark blue lines, respectively), from 2014 to 2021. The ACM/w for the
18-64 years age group is rescaled (multiplied), as explained in the text, to make the number of
deaths per week of both age groups equal in the summer-2014 trough, for comparison
purposes. Data are displayed from week-40 of 2013 to week-37 of 2021 for the whole
104
continental USA, including Alaska and Hawaii. Data were retrieved from CDC (CDC, 2021a), as
described in Table 1.
Figure 32a shows that, in the pre-COVID-era, the elderly group (65+ years) is always
approximately 2-3 times more susceptible to the additional challenges and stress of
winter than the younger group (18-64 years). This rule is not followed in the COVID-era.
In the COVID-era, the relative summer-2020 and summer-2021 mortalities are greater
for the younger age group than for the elderly group (Figure 32a), which is reversed
compared to known age-dependent vulnerability to dying from viral respiratory diseases.
This reversal in the COVID-era is more explicitly illustrated in Figure 32b, which shows
the difference by week of the two curves depicted in Figure 32a.
Figure 32b. Difference in all-cause mortality by week in the USA between the 65+ years
and the rescaled 18-64 years age groups, from 2014 to 2021. The ACM/w for the 18-64
years age group was rescaled (multiplied), as explained in the text, to make the number of
deaths per week of both age groups equal in the summer-2014 trough, for comparison
purposes. Data are displayed from week-40 of 2013 to week-37 of 2021 for the whole
continental USA, including Alaska and Hawaii. The dashed line emphasizes the zero. Data were
retrieved from CDC (CDC, 2021a), as described in Table 1.
105
Here (Figure 32b), we see that the younger age group (18-64 years) has moderately
more (rescaled) deaths in summer-2020, and significantly more (rescaled) deaths in
summer-2021. Two possible interpretations come to mind: either the integrated
cumulative long-term stress from the government measures takes longer to affect more
tolerant younger individuals than older individuals, or the massive vaccination campaign
administered between the two summers (Figure 31, below) has had a disproportionate
negative impact on the younger age group.
A more detailed examination of the COVID-era is possible thanks to more age-group
resolution being publicly available for that time period (CDC, 2021b), at the national
level (not state-resolved), as follows. A selection of these data is shown in Figure 33.
Figure 33a. All-cause mortality by week normalized by population for the USA for the 14
years and less (“-14 years”) age group, for each of both sexes, from 2020 to 2021. Data
are displayed from week-1 of 2020 to week-37 of 2021 for the whole continental USA, including
Alaska and Hawaii. The population used for normalization is the population of the specific age
group and sex. ACM data were retrieved from the CDC (CDC, 2021b) and population data were
retrieved from the US Census Bureau (US Census Bureau, 2021b), as described in Table 1.
106
Figure 33b. All-cause mortality by week for the USA for the 15-34 years age group, both
sexes, from 2020 to 2021. Data are displayed from week-1 of 2020 to week-37 of 2021 for the
whole continental USA, including Alaska and Hawaii. The population used for normalization is
the population of the specific age group. ACM data were retrieved from the CDC (CDC, 2021b)
and population data were retrieved from the US Census Bureau (US Census Bureau, 2021b),
as described in Table 1.
Figure 33c. All-cause mortality by week normalized by population for the USA for females
of the 15-34 years age group, from 2020 to 2021. Data are displayed from week-1 of 2020 to
week-37 of 2021 for the whole continental USA, including Alaska and Hawaii. The population