Demographic science aids in understanding the spread and fatality rates of COVID-19

Abstract

Governments around the world must rapidly mobilize and make difficult policy decisions to mitigate the COVID-19 pandemic. Because deaths have been concentrated at older ages, we highlight the important role of demography, particularly how the age structure of a population may help explain differences in fatality rates across countries and how transmission unfolds. We examine the role of age structure in deaths thus far in Italy and South Korea and illustrate how the pandemic could unfold in populations with similar population sizes but different age structures, showing a dramatically higher burden of mortality in countries with older versus younger populations. This powerful interaction of demography and current age-specific mortality for COVID-19 suggests that social distancing and other policies to slow transmission should consider both the age composition of local and national contexts as well as the social connectedness of older and younger generations. We also call for countries to provide case and fatality data disaggregated by age and sex to improve real-time targeted nowcasting.

Full text

1
Demographic science aids in understanding the spread and fatality
rates of COVID-19
Jennifer Beam Dowd*, Valentina Rotondi, Liliana Andriano, David M. Brazel, Per Block, Xuejie Ding, Yan
Liu, Melinda C. Mills*
Leverhulme Centre for Demographic Science, University of Oxford & Nuffield College, UK
*Corresponding authors email: jennifer.dowd@sociology.ox.ac.uk and melinda.mills@nuffield.ox.ac.uk
Abstract. Governments around the world must rapidly mobilize and make difficult policy decisions to
mitigate the COVID-19 pandemic. Because deaths have been concentrated at older ages, we highlight
the important role of demography, particularly how the age structure of a population may help explain
differences in fatality rates across countries and how transmission unfolds. We examine the role of age
structure in deaths thus far in Italy and South Korea and illustrate how the pandemic could unfold in
populations with similar population sizes but different age structures, showing a dramatically higher
burden of mortality in countries with older versus younger populations. This powerful interaction of
demography and current age-specific mortality for COVID-19 suggests that social distancing and other
policies to slow transmission should consider both the age composition of local and national contexts as
well as the social connectedness of older and younger generations. We also call for countries to provide
case and fatality data disaggregated by age and sex to improve real-time targeted nowcasting.
Background
Governments are rapidly mobilizing to minimize transmission of COVID-19 through social distancing and
travel restrictions to reduce fatalities and the outstripping of healthcare capacity. It is becoming clear
that the pandemic’s progression and impact may be strongly related to the demographic composition of
the population, specifically population age structure. Demographic science can provide new insights into
how the pandemic may unfold and the intensity and type of measures needed to slow it down.
Currently, COVID-19 mortality risk is highly concentrated at older ages, particularly those aged 80+. In
China, case-fatality rate (CFR) estimates range from 0.4% (40-49 years), jumping to 14.8% (80+ years).(1)
This is consistent with the data from Italy as of March 13, where the reported CFR is 10.8% for those 70-
79, 17.5% for 80-89, and 21.1% for those >90, with only six deaths under the age of 50. Thus far, only 3%
of deaths have occurred in those under aged 60 (see Table S1, Supp Info).(2)
The Importance of Age Structure
Population age structure may explain the remarkable variation in fatalities across countries and why
countries such as Italy are especially vulnerable. The deluge of critical and fatal COVID-19 cases in Italy
was unexpected given the health and wealth of the affected region. Italy is one of the oldest populations
in the world with 23.3% its population over age 65, compared to 12% in China (3). Italy is also a country
characterized by extensive intergenerational contacts which are supported by a high degree of
residential proximity between adult children and their parents (4). Even when inter-generational
families do not live together, daily contacts among non-co-resident parent-child pairs are frequent.
Many Italians also often prefer to live close to their extended family and commute to work daily.
According to the latest available data by the Italian National Institute of Statistics, this extensive
commuting affect over half of the population in the northern regions.(5) These intergenerational
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

2
interactions, co-residence, and commuting patterns may have accelerated the outbreak in Italy through
social networks that increased the proximity of elderly to initial cases (see Supp Info).
Age structure, along with early detection and treatment, also likely explains the low numbers of
fatalities in South Korea and Singapore compared to Italy. The Korean outbreak, while large, was
concentrated amongst the young recruits of the Schincheonji religious group (6), with only 3.3% of cases
falling into the very vulnerable >80 group.(7) Singapore is notable with zero deaths thus far, but have
had only one confirmed case over 80 and only 10/200 cases above age 70.(8)
COVID-19 transmission chains that begin in younger populations may have a low number of severe cases
and thus go longer undetected, (9, 10) with countries thereby slow to raise the alarm. The low case
fatality rate in England thus far (0.01%) may reflect the relatively young age structure of populations
impacted to date, including Greater London, which has a small fraction of residents over age 65
compared to more rural areas (11). COVID-19 was only detected in King County, Washington once it
reached the Life Care Centre in Kirkland, where 19 out of 22 deaths occurred, despite estimates based
on virus genetic sequences suggesting it circulated for several weeks prior (12). Once community
transmission is established, countries that have a high level of intergenerational contacts and co-
residence may see faster transmission to high-fatality age groups as seen in Italy.
In Figure 1, we use population pyramids to illustrate how population age structure interacts with high
COVID-19 mortality rates at older ages to generate large differences across populations in the number of
deaths, using existing assumptions about infection prevalence and age-specific mortality. The top panel
considers two countries currently affected, Italy and South Korea. The larger number of expected
fatalities for Italy is clearly visible in the right panel. In the bottom panel, we consider two countries yet
untouched by the pandemic who have similar population sizes but very different age distributions. In
Brazil, which has 2.0% of its population age 80+, the simulated scenario leads dramatically more deaths
(478,629) compared to Nigeria (137,489), where the fraction over 80 is only 0.2%. Figure 2 uses an
alternative visualization to depict the expected deaths by age groups in Italy, Brazil, Nigeria, UK and US,
together with the proportion of the population in different age groups. Both figures demonstrate the
stark implications of an older population age structure for higher fatalities and critical cases.
Demographic Science and COVID-19 Policy
Going forward, demographically informed projections will better predict the COVID-19 burden and
inform governments about targeted action. While population age structure is crucial for understanding
the populations at the highest risk of mortality both across and within countries, it is also vital for
understanding how much social distancing measures are required in each population to reduce the
number of most critical cases and overload on the health system—aka “flattening the curve”(13). Our
illustrations show that countries with older populations will need to take more aggressive protective
measures to stay below the threshold of critical cases that outstrip health system capacity. For these
measures to be effective, special attention should be devoted to those population groups that are more
at risk and patterns of intergenerational contact.
Consideration of population age structure also necessitates understanding the interlinkage of policy
measures and how one policy might create a domino effect of unintended consequences. While schools
may be a hub of contact and virus transmission, school closings may inadvertently bring grandparents
and children into closer contact if they become the default carers. In aged populations with close
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

3
intergenerational ties, governments need to facilitate childcare solutions that reduce contact. In a
pending decree, for instance, the Italian government will introduce a special leave for parents with
children at home from school and a voucher (around 600 euros) for babysitting.
The age structure of populations also suggest that the often squeezed “sandwich” generation of adults
who care for both the old and young are an important link for mitigating transmission. Beyond
introducing sick pay for those who need to self-isolate or care for family members, joint government
and industry emergency policy measures should seek to counter family economic crises by delaying rent
and mortgage payments for example, particularly for vulnerable and precarious workers. In the absence
of economic security measures, this crucial sandwich generation may be less able to comply with
policies that allow social distancing.
Conclusion
The rapid spread of COVID-19 has revealed the need to understand how population dynamics interact
with pandemics now and in the future. Population ageing is currently more pronounced in wealthier
countries, which mercifully may lessen the impact of this pandemic on poorer countries with weaker
health systems but younger age structures. It is plausible that poor general health status and co-
infections such as tuberculosis may still increase the danger of COVID-19 among younger cases in these
countries. Thus far, the lower than expected number of cases detected in Africa (despite extensive trade
and travel links with China), suggests that the young age structure of the continent may be protective of
severe and thus detectable cases, or it may be undetected. Beyond age structure, there are large sex
differences in mortality that need to be understood – with men at higher risk – some of which may be
accounted for by the stark differences in smoking rates by sex in Asia. Distributions of underlying co-
morbidities such as diabetes, hypertension and COPD will likewise refine risk estimates. Until these
more nuanced data are available, the concentration of mortality risk in the oldest old ages remains one
of the best tools we have to predict the burden of critical cases and thus more precise planning of
availability of hospital beds, staff and other resources.
At this moment, few countries are routinely releasing their COVID-19 data with key demographic
information such age, sex, or comorbidities. We call for the timely release of this disaggregated data to
allow researchers and governments to nowcast risk for more focused prevention and preparedness.
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

4
Figure 1. Population composition (left panel) and expected deaths in population (right panel), Italy and
Republic of Korea (top panel) and Nigeria and Brazil (bottom panel)
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

5
Figure 2. Expected deaths by total population (top panel) and proportion of total population by age
group (bottom panel), Italy, Brazil and Nigeria
Methods
Data. The data to produce Figures 1 and 2 use population data from (https://population.un.org/wpp)
with the relative risk of death taken from the most recent Italian data, last accessed March 13, 2020.(14)
SI analyses breakdown data across regions in Italy over time and geographically, with detailed data
sources listed there.
Data Analysis. For Figures 1 and 2, the total number of expected deaths by age group was derived by
multiplying the total number of people in each age group and country by an assumed population
infection rate of 0.4 and age-and sex-specific mortality rates extracted from most recent Italian data.
The male-to-female relative risk of 1.65 based on current estimates from China,(1) as Italian estimates
by sex have not consistently released.
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

6
Data analysis is conducted in R using the packages [ggplot2]. Additional analyses are shown in the SI
including: regional variation in Italy, variation in population pyramid estimates by differences in the
infection rate and relative risk by sex, and additional countries not shown in the main article.
Supplementary Information.
Author Contributions. All authors jointly devised the study. JBD & MCM drafted the manuscript in which
all authors contributed and commented. VR & LA wrote the Italian case and graphics. PB worked on
intergenerational transmission. LA and DMB generated graphics and with XD & YL, updated country
statistics.
References
1. C. P. E. R. E. Novel, The epidemiological characteristics of an outbreak of 2019 novel cornavirus
diseases (COVID-19) in China. Zhonghua liu xing bing xue za zhi= Zhonghua liuxingbingxue zazhi
41, 145 (2020).
2. I. S. di S. (ISS), Istituto Superiore di Sanità (Higher Institute of Health) official website (2020)
(March 13, 2020).
3. United Nations, World Population Prospects 2019 (2020) (March 13, 2020).
4. M. Kalmijn, C. Saraceno, A Comparative Perspective on Intergenerational Support. Eur. Soc. 10,
479–508 (2008).
5. I. N. di S. (ISTAT), Spostamenti quotindiani e nuove forme di mobilità [Daily shifts and new forms
of mobility] (2018) (March 13, 2012).
6. A. Salmon, Why are Korea’s COVID-19 death rates so low? Asia Times (2020).
7. KCDC, KCDC Press Release (2020) (March 12, 2020).
8. M. of H. S. (SMH), Updates on COVID-19 Local Situation (2020) (March 13, 2020).
9. Y. Bai, et al., Presumed Asymptomatic Carrier Transmission of COVID-19. JAMA (2020)
https:/doi.org/10.1001/jama.2020.2565.
10. T. Ganyani, et al., Estimating the generation interval for COVID-19 based on symptom onset data.
medRxiv (2020) https:/doi.org/10.1101/2020.03.05.20031815.
11. P. H. E. (PHE), COVID-19: track coronavirus cases (2020) (March 13, 2020).
12. T. Bedford, Cryptic transmission of novel coronavirus revealed by genomic epidemiology. Bedford
Lab Blog (2020) (March 12, 2020).
13. S. Roberts, Flattening the Cornavirus Curve. New York Times (2020)
https://www.nytimes.com/2020/03/11/science/coronavirus-curve-mitigation-infection.html.
14. Istituto Superiore di Sanità, Age specific mortality rates COVID-19 (2020) (March 13, 2020).
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

7
Supplementary Information
Demographic science aids in understanding the spread and fatality rates of COVID-19
1. A Case Study of Population Ageing, Intergenerational Contacts, and COVID-19 in Italy
After Japan, Italy is currently the world’s second oldest population, with 23% of the population
aged 65 years and older compared to 13.2% aged 15 and under.1 This population ageing and
population decline has been driven by very low fertility rates and growing rates of
childlessness,1 which are not compensated by immigration flows. In the ranking of current cases
and deaths from coronavirus in the world, Italy sadly occupies the second position after China.
On Friday February 21st 2020, the first case of COVID-19 in Italy was confirmed in in the
province of Lodi, in Lombardy.2 Since then, other regions in Northern Italy – including Emilia
Romagna, Piedmont, and Veneto – started to report rapid increases. Between February 24th
and March 13th, the number of cases and deaths in these regions increased exponentially
(Figure S1).
The response of the Italian government to the spread of the coronavirus across Italy has been
firm, yet not always effective. Most of the initial interventions targeted the four regions in
northern Italy that were the most affected: Lombardy, Piedmont, Veneto, and Emilia Romagna.
In particular, on February 23rd, the government issued a decree which prohibited the
movement of people outside of 10 municipalities located in Lombardy, in the province of Lodi,
and a municipality located in Veneto, in the province of Padova.3
While cases in the province of Bergamo began to increase from Feb 24th, in contrast to Lodi no
shutdowns or restrictions were imposed. A few hours later, the Veneto and Lombardy regions
issued an ordinance by which all schools of all levels and grades, including Universities, were
closed and all cultural, recreational, sporting and religious events, being them public or private,
were suspended.
Starting from March 8th, the shutdown implemented in the province of Lodi, was expanded to
the entirety of Lombardy (including Bergamo) and to fourteen provinces in Veneto, Emilia
1 Balbo N, Billari FC, Mills M. Fertility in Advanced Societies: A Review of Research. Eur J Popul / Rev Eur
Démographie 2013; 29: 1–38.
2 Lombardy is one of those Northern Italian regions that experienced a slight growth in the population in recent
years. However, this growth is mainly due to immigration flows both from within and outside the country. Indeed,
also in Lombardy the natural balance is negative and decreasing since 2012. According to Eurostat, Lombardy is
also the most populated region in Italy and one among the wealthier regions in Europe, with a GDP per capita
amounting to €37,800 in 2017 (compared to €28,400 in Italy and €29,500 in the EU)
(https://ec.europa.eu/growth/tools-databases/regional-innovation-monitor/base-profile/lombardy)
3 However, the decree did not have immediate implementation. The control plan did not start rigidly, and
checkpoints were not very effective. Law enforcement officers (at least 500 men) are deployed only on February
25 to ensure that no one enters and exits the so-called red areas creating 35 gates.
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

8
Romagna, Piedmont, and Marche. The decree, however, leaked in the late evening of March
7th, generating a massive outflow of people from these regions in the North to several regions
in the South, where the lowest number of cases are currently registered. Only two days later, a
new decree of the Prime Minister, Giuseppe Conte, extended until April 3rd identical measures
to the entire country of Italy. On March 11th, a further crackdown resulted in the closure of all
shops (except for groceries and pharmacies), pubs, and restaurants.
Figure S1. Number of cases and deaths across regions over time, March 13, 2020
As of March 13th, the number of cases that tested positive for COVID-19 in Italy amounted to
17,660, the number of deaths to 1,266 and the number of recovered to 1,439. With a total of
9,820 (i.e. 55.6% of the total) cases tested positive to COVID-19, 890 deaths, and 1,198
recovered, Lombardy is the most affected region in Italy (Figure S1).
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

9
Figure S2 illustrates the number of cases in Lombardy as of March 12th by province. The more
intense red indicates a greater number of cases, as reported in the legend.
Figure S2. Number of cases by province as of March 13, 2020
As of March 13th, the most affected province of Bergamo (2,368 cases) has largely overcome
the province of Lodi (1,133 cases) where the outbreak started and the containment measures
were introduced first, as shown in Figure S3. We note that social distancing interventions were
invoked on Feb 23rd in Lodi but until March 8th in Bergamo, providing some empirical evidence
for the potential of “flattening the curve” interventions.
As already suggested, Italy is a country characterized by extensive intergenerational contacts
which are supported by a high degree of residential proximity between adult children and their
parents. Although proximity is the highest in smaller villages and in poorer regions, the
geography of proximity across Italy is rather homogeneous, with the two richest regions —
Lombardy and Trentino-Alto Adige — showing high rates of proximity as well. Even when inter-
generational families do not live under the same roof, daily contacts among non-co-resident
parent-child pairs are frequent. While on the one hand, this geographical proximity guarantees
high rates of mutual intergenerational solidarity, both financial and in-kind, one obvious
consequence is that Italians often prefer not to move for work but to live close to their family
and commute daily to go to work. According to the latest available data by the Italian National
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

10
Institute of Statistics, in the northern regions this extensive commuting affect over half of the
population4.
Figure S3. Number of cases in the Province of Bergamo (red) and Lodi (green) as of March 13,
2020
A stylized example from social network theory can be helpful in explaining why
intergenerational interactions, co-residence, and commuting patterns might have played a role
in the spread of the COVID-19 infection to the older population in Italy. Individuals’ social
networks are generally composed of people similar in age. The population structure of contact
can be represented as age-homogeneous communities that have low contact between groups.
If the initial infections in northern Italy were younger people commuting to cities and having
plausibly international contacts, a crucial determinant of risk for the elderly is their network
distance to these younger sources, i.e. how many intermediaries need to be infected until they
are reached. Network science showed that even relatively few connections between
communities can lead to a stark reduction in average network distances; the so-called small
world phenomenon5. Such community “connecting” individuals might be those young people
around Milan that work in the city but reside in the most hard-hit villages in the surrounding
with their parents and grandparents. Thus, intergenerational co-residence may have
accelerated the outbreak by creating intercommunity connections that increase the proximity
of elderly to the initial cases, an area for further study.
4 https://www.istat.it/it/archivio/224469
5 Watts, D. J. (1999). "Networks, dynamics, and the small-world phenomenon." American Journal of Sociology
105(2): 493-527.
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

11
This stylized example may serve, once more, to show why while population age structure is
crucial for understanding the populations at the highest risk of mortality both across and within
countries, it is also important for understanding how much social distancing measures are
required in each population to reduce the number of most critical cases and overload on the
health system—aka “flattening the curve.”6 At this time of severe crisis, policy makers are called
to define containment measures which are often difficult to sustain in the long run and which
have immense repercussions in terms of socio-economic sustainability. For these measures to be
effective, a special attention should be devoted to those population groups that are more at risk
and to the strength of the connections across generations.
An interesting example in this direction comes from the Canton Ticino,7 the canton in Switzerland
bordering to Lombardy. On March the 12th, the Canton Ticino has adopted measures aimed at
contrasting the diffusion of the virus which are explicitly aimed at protecting the elderly and at-
risk populations. To this aim, the resolution 12628 “strongly discourages” people over 65 (and
those categories at risk of incurring serious complications that can endanger their lives) to “look
after children, participate in public or private events, use public transportation, except for
medical and professional needs or for the purchase of basic necessities, and attend public
exercise.” Further research to test how population age structure, intergenerational contacts, and
social distancing measures interact to best mitigate risk is needed.
6 https://www.nytimes.com/2020/03/11/science/coronavirus-curve-mitigation-infection.html
7 We thank Prof. Luca Crivelli for this suggestion.
8 https://www4.ti.ch/fileadmin/DSS/DSP/UMC/malattie_infettive/Coronavirus/1262.pdf
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

12
2. Variation in fatality rates by different population structures
Here we replicate Figure 1 in Figures S4-S6 and show the variation in fatality rates with varying
assumptions about the infection rate and relative risk differences between males and females.
Figure S4. Population composition (left panel) and expected deaths in population (right panel), Italy and
Republic of Korea (top panel) and Nigeria and Brazil (bottom panel) using Infection rate = 0.2; Relative
risk M/F = 1.65
Note: Total number of expected deaths by age group is derived by multiplying the total number of people in each
age group and country by an assumed infection rate of 0.2 and age-and sex-specific mortality rates extracted from
Italian data. The male-to-female relative risk of 1.65 based on current estimates from China.
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

13
Figure S5. Population composition (left panel) and expected deaths in population (right panel), Italy and
Republic of Korea (top panel) and Nigeria and Brazil (bottom panel) using Infection rate = 0.6; Relative
risk M/F = 1.65
Note: Total number of expected deaths by age group is derived by multiplying the total number of people in each
age group and country by an assumed infection rate of 0.6 and age-and sex-specific mortality rates extracted from
Italian data. The male-to-female relative risk of 1.65 based on current estimates from China.
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

14
Figure S6. Population composition (left panel) and expected deaths in population (right panel), Italy and
Republic of Korea (top panel) and Nigeria and Brazil (bottom panel) using Infection rate = 0.4; Relative
risk M/F = 2.4
Note: Total number of expected deaths by age group is derived by multiplying the total number of people in each
age group and country by an assumed infection rate of 0.4 and age-and sex-specific mortality rates extracted from
Italian data. The male-to-female relative risk of 2.4 based on current estimates from South Korea.
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

15
3. Demographic population pyramid projections for additional countries
Figure S7 graphs the population composition and expected deaths in the population for the
additional countries of the United States, Japan and South Africa. Japan has a relatively old
population, South Africa a younger population and the US is more evenly distributed. Based on
the age-specific mortality rates extracted from the Italian data, we project how these different
countries will experience deaths attributed to COVID-19 by age and sex.
Figure S8 depicts the population composition (left panel) and the expected deaths in the
population (right panel), this time for Italy versus South Korea (top panel) and Italy versus the
United Kingdom. Here we immediately see from the upper right panel that Korea is expected to
experience markedly fewer deaths than Italy, largely attributed to the population age structure.
The United Kingdom is as of March 13 2020 is standing out as one of the few European
countries to take not stringent actions such as closing schools or stopping large public events.9
In spite of the comparatively younger population of the UK, the bottom right panel illustrates
that the UK could face similar numbers of COVID-19 deaths as Italy. Due to age structure
differences, the UK will likely have slightly fewer deaths of those 80+ in comparison to Italy, but
in the coming weeks still likely to face considerable pressure on its healthcare system.
9 https://en.unesco.org/themes/education-emergencies/coronavirus-school-closures
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

16
Figure S7. Population composition (left panel) and expected deaths in population (right panel), United
States, Japan and South Africa
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

17
Note: Total number of expected deaths by age group is derived by multiplying the total number of people in each
age group and country by an assumed infection rate of 0.4 and age-and sex-specific mortality rates extracted from
Italian data. The male-to-female relative risk of 1.65 based on current estimates from China.
Figure S8. Population composition (left panel) and expected deaths in population (right panel), Italy and
Republic of Korea (top panel) and Italy and United Kingdom (bottom panel)
Note: Total number of expected deaths by age group is derived by multiplying the total number of people in each
age group and country by an assumed infection rate of 0.4 and age-and sex-specific mortality rates extracted from
Italian data. The male-to-female relative risk of 1.65 based on current estimates from China.
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint

18
Table S1. Age-Specific COVID-19 Case Fatality Rates from Italy as of March 13, 2020
AGE
CFR
(CASE FATALITY RATE
)
0
-
9
0.0%
10
-
19
0.0%
20
-
29
0.0%
30
-
39
0.2
%
40
-
49
0.2
%
50
-
59
0.8
%
60
-
69
2.7%
70
-
79
10.8
%
80
-
89
17.5
%
90+
21.1
%
Source: Istituto Superiore di Sanità, Age specific mortality rates COVID-19 (2020) (March 13,
2020).
. CC-BY 4.0 International licenseIt is made available under a
author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the(which was not peer-reviewed) The copyright holder for this preprint .https://doi.org/10.1101/2020.03.15.20036293doi: medRxiv preprint