A MATHEMATICAL DESCRIPTION OF COVID-19

CHENLIN GU, WEI JIANG, YIYANG YU, AND TIANYUAN ZHAO

Abstract. This is a short report about the mathematical description of the diffusion of
COVID-19. By talking about its incubation period, rate of contamination and mortality we
explain why COVID-19 is big challenge for humans. We also suggest some possible actions
to stop its diffusion.

1. Modeling
1.1. Dynamic. The COVID-19 has spread since December 2019 and WHO has raised global
COVID-19 risk to highest level Feb 28 2020. Readers can consult [5] for a general introduc-
tion and [1] for its newest evolution report. Here we give a direct mathematical description
about its diffusion to illustrate its incredible speed of spread, under the condition no specific
treatment and vaccine exist for the moment.
The diffusion of COVID-19 under certain measure can be described as a variant birth-death
process, as we study the early stage of the epidemic: We consider three types of population
• Ic : R+ → Z the accumulative confirmed patients.
• Ia : R+ → Z the active undetected patients.
• Iq : R+ → Z the unconfirmed patients under quarantine.
The total accumulative infections I(t) = Ic (t) + Ia (t) + Iq (t). The diffusion can be
described as following:
(1) Infection: We start from patient zero, every patient before the confirmation/quarantine
has rate β to infect a healthy one independently, i.e. the new infection comes as a
Poisson process of parameter β. Thus, if there is no action and let it evolve freely,
E[I(t)] = I(0) exp(βt).
(2) Incubation period: Before showing symptom, a patient has an incubation period,
which is a random variable T > 0. From the statistic in the early cases [6], T is very
close to a Gamma distribution of density p(t) (see fig. 1)
bn n−1 −bt
(1.1) p(t) = t e .
Γ(n)
and [9] use these data to identify the approximated parameters n = 5 and b = 23 ,
which implies E[T ] = n/b = 7.5.
(3) Confirmation: After the incubation period, the symptom is detected and the patient
will be confirmed with a probability θ ∈ [0, 1]. In the ideal case, this parameter could
be 1. But as the symptom of COVID-19 is close to that of flu, and the limit of test,
this number is smaller than 1.
(4) Separation and quarantine: A confirmed case is separated, and those who have
contacted the case will be put in quarantine with probability α ∈ [0, 1], because
1
 Figure 1. A picture from [9] to illustrate the distribution of incubation period
between the infection and symptom.

the infection may occur in some public space (train, airport, flight, supermarket,
restaurants etc) and it requires efforts to track all.
1.2. Reproductive number. As we have seen from the description, it is the class Ia which
has a high possibility to infect the healthy people, while this class is sometimes hard to be
identified during the incubation period: Since the infection sometimes occurs in public space,
they do not know how and where they are infected; before the symptom they even do not
know whether they are infected; but they can infect the others from the report [6]. Therefore,
main task to stop the epidemic is to transform all the Ia to Ic or Iq , since it is the stopping
time that no new infection occurs in the sense person-to-person infection.

Figure 2. The transition between three states of the infection.

2
 In the literature of medicine and epidemic, another quantity known as reproductive num-
ber R0 is also used to measure the risk of the epidemic. It is the expectation of infection from
one patient. Generally speaking, the number of the n-th generations of infections is (R0 )n .
We have three distinguished phases:
• Supercritical case: R0 > 1, then the epidemic has an exponential increment.
• Critical case: R0 = 1, then the epidemic exists for long time.
• Subcritical case: R0 < 1, then after a period, the epidemic will be well controlled.
In our model, without the measure of quarantine and suppose θ = 1, then
(1.2) R0 = βE[T ].
The basic reproduction number R0 was estimated to 2.6 with uncertainty range 1.5 to 3.5 in
[4]. Deduce from this and also eq.1.1 E[T ] = 7.5 and eq.1.2, we obtain β ' 0.34, When the
quarantine and the rate of confirmation are taken into consideration, asymptotically we have

e0 ' β(1 − α)E[T /θ] = β(1 − α) E[T ].

R
θ
Once again, we deduce that the the long incubation period and the high rate of contami-
nation is the key point in the spread of the virus. Mathematically, to control the diffusion is
to make R e0 smaller than 1, the possible efforts are
• Reduce β by avoiding activities and contacts, and do more protection.
• Increase α by using all methods to find out the trace of infection and contacts.
• Make θ to 1 as close as possible.
We add more remarks on these parameters: If we suppose θ = 1 and keep β unchanged, then
θ
αc ' 1 − ' 0.6.
βE[T ]
Numerically, it requires that α > 0.6 to prevent the spread of the COVID-19, which is a
threshold comparatively high, but still possible.
For the public, to reduce the contact and activities, thus the parameter β is a practical
and efficient way to protect themselves and also the others. If we keep the other parameters
unchanged and reduce β by 41 , then R e0 is also reduced by 1 . This requires only, for example,
4
one day or two days working from distance during a week.
Finally, we have to highlight that R
e0 is also very sensitive with respect to θ. It is especially
in the case that too many patients in hospital and the medical resources are saturated. We
see, once θ goes down to 0.5, even we improve α to 0.8,

e0 ' β(1 − α) E[T ] = 0.34 × 0.2 × 7.5 = 1.02 > 1.

R
θ 0.5

Variable Value (Range) Definition

T [0, 30] Incubation period from infection to symptom.
θ [0,1] The rate of confirmation once detected.
β 0.34 The rate of contamination.
α [0,1] The probability to be under quarantine.

3
1.3. Mortality. WHO had said last week that the mortality rate of COVID-19 can differ,
ranging from 0.7% to up to 4%, depending on the quality of the health-care system where it is
treated. The most recent report from WHO at Mars 5, 2020 estimates it to be 3.4% globally
[2]. A little earlier, [11] summarizes of 72314 cases from the Chinese Center for Disease Control
and Prevention: the case-fatality rate is 2.3% (1023 of 44672 confirmed cases), and 14.8%
in patients aged over 80 years (208 of 1408), 8.0% in patients aged 70-79 years (312 of 3918)
and 49.0% in critical cases (1023 of 2087). Thus, the death rate is 20 times higher than that
of seasonal flu (0.1%) and close to that of 1918 influenza pandemic (estimated 2.3% ∼ 3%
[8]). Especially, when the hospital is saturated, the mortality will climb. Thus it also implies
the importance to avoid the explosion in large scale.

2. Simulation
In the following, we do some simulations to back up our model and suggestion. Figure 4
is the statistics from the data now in France and in China. In fig. 5 and fig. 6, we study
the future evolution in France under different measures by 100 Monte-Carlo simulations.
By comparing the available statistics, we suppose the evolution starts from 8 unconfirmed
infections from 14/02/2020, which results in 163 infections at 01/03/2020 by average of 100
trajectories. We simplify a little the model by supposing that after quarantine, the infection
will be confirmed immediately, so Iq is also classified as Ic and we only study (Ia , Ic ). The
Representative parameters are chosen as following:
• Group A - Weak measure (β, θ, α) = (0.34, 0.3, 0.2).
• Group B - Normal measure (β, θ, α) = (0.34, 0.5, 0.4).
• Group C - Strong measure (β, θ, α) = (0.34, 0.7, 0.6).
• Group D - Very strong measure (β, θ, α) = (0.34, 0.9, 0.8).
• Group E - Normal measure with reduction of activities (β, θ, α) = (0.25, 0.5, 0.4).
We see from the simulation, in the case Group A when the measure is too weak, in a month
the confirmed infections will explode to 12400 with nearly 14000 undetected cases - then the
whole medical system will go down and be a catastrophe; it is in Group D with very strong
measure we see that the trend is controlled; however, as a comparison in Group E, which has
the same θ, α as Group B as a normal measure, but with a reduction of β by 41 , the diffusion of
epidemic is controlled as better as Group D. As a consequence, this may suggest that working
in distance and “distanciation sociale” as a possible answer to COVID-19 or the moment.

3. More discussion and references

3.1. Related works. There are active researches and publications since the outbreak of
COVID-19 in January 2020, in clinical aspect as well as in epidemiological one. A series of
reports, from alarming estimates of actual infection number based on air travels, to estimated
of basic transmission rate [4], as well as case fatality ratio estimates, could be found online1.
[10] gives several scenarios to estimate actual and forecast infections numbers and exported
cases in several major Chinese cities; the estimates was however before the severe social
distancing policies took effect in China. [7] reports a relatively complete data description of
reported confirmed cases in China as of February 2020. [3] is an attempt to model effectiveness
of isolation and contacts tracing.
1https://www.imperial.ac.uk/mrc-global-infectious-disease-analysis/
news--wuhan-coronavirus/
4
 South Korea 4,335

Italy 1,696

Iran 1,501

Diamond Princess 705

Japan 274

Germany 150

France 130

Spain 120

Singapore 108

Hong Kong 100

United States 88

Kuwait 56

Bahrain 47

Thailand 43

Taiwan 40

United Kingdom 40

Australia 30

Malaysia 29

Canada 24

Switzerland 24
0 1,000 2,000 3,000 4,000
Number of confirmed cases at 2020-03-02 09:15

Figure 3. Number of confirmed infections until 02/03/2020. The data comes

from BNO news.

Acknowledgement. We would thank Xin Lv (Orsay), Tunan Zhu (ENS), Honghao Li (In-
stitute Curie), Waner Chen, Mingxing Du, Keming Zhang (Diderot), Prof. Wenbin Chen
(Fudan) for helpful discussion.
We should thanks everyone who are fighting against COVID-19, specially we present our
deep respect to Doctor Wenliang Li and Prof. Nanshan Zhong.

References
[1] World health organization homepage.
5
 Figure 4. On the left it is the statistics by now in France and on the right
is the one in China, where it takes about one month and efforts to stop the
momentum of spread. A big jump in the graph comes from a change of criteria
at the middle of February that many pending cases are confirmed. The data
comes from the report of WHO.

[2] N. H.-D. Berkeley Lovelace Jr. Who says coronavirus death rate is 3.4% globally, higher than previously
thought.
[3] J. Hellewell, S. Abbott, A. Gimma, N. I. Bosse, C. I. Jarvis, T. W. Russell, J. D. Munday, A. J. Kucharski,
W. J. Edmunds, S. Funk, et al. Feasibility of controlling 2019-ncov outbreaks by isolation of cases and
contacts. medRxiv, 2020.
[4] N. Imai, A. Cori, I. Dorigatti, M. Baguelin, C. A. Donnelly, S. Riley, and N. M. Ferguson. Report 3:
Transmissibility of 2019-nCoV.
[5] J. W. Knvul Sheikh, Derek Watkins and M. GröndahlUpdated. How bad will the coronavirus outbreak
get? here are 6 key factors, 2020.
[6] Q. Li, X. Guan, P. Wu, X. Wang, L. Zhou, Y. Tong, R. Ren, K. S. Leung, E. H. Lau, J. Y. Wong,
X. Xing, N. Xiang, Y. Wu, C. Li, Q. Chen, D. Li, T. Liu, J. Zhao, M. Liu, W. Tu, C. Chen, L. Jin,
R. Yang, Q. Wang, S. Zhou, R. Wang, H. Liu, Y. Luo, Y. Liu, G. Shao, H. Li, Z. Tao, Y. Yang, Z. Deng,
B. Liu, Z. Ma, Y. Zhang, G. Shi, T. T. Lam, J. T. Wu, G. F. Gao, B. J. Cowling, B. Yang, G. M. Leung,
and Z. Feng. Early transmission dynamics in wuhan, china, of novel coronavirus–infected pneumonia.
New England Journal of Medicine, 0(0):null, 0.
[7] C. P. E. R. E. Novel et al. The epidemiological characteristics of an outbreak of 2019 novel coronavirus
diseases (covid-19) in china. Zhonghua liu xing bing xue za zhi= Zhonghua liuxingbingxue zazhi, 41(2):145,
2020.
[8] W. H. Organization et al. Pandemic influenza preparedness and response: a WHO guidance document.
Geneva: World Health Organization, 2009.
[9] N. Shao, J. Cheng, and W. Chen. The reproductive number R0 of COVID-19 based on estimate of a
statistical time delay dynamical system. medRxiv, 2020.
[10] J. T. Wu, K. Leung, and G. M. Leung. Nowcasting and forecasting the potential domestic and inter-
national spread of the 2019-ncov outbreak originating in wuhan, china: a modelling study. The Lancet,
2020.
[11] Z. Wu and J. M. McGoogan. Characteristics of and Important Lessons From the Coronavirus Disease
2019 (COVID-19) Outbreak in China: Summary of a Report of 72314 Cases From the Chinese Center
for Disease Control and Prevention. JAMA, 02 2020.

(Chenlin GU) DMA, Ecole Normale Supérieure, PSL University, Paris, France
Email address: [email protected]
6
 Figure 5. 100 traces of Monte-Carlo simulation under 4 different measures,
from 14/02/2020 to the end of Mars. The parameters are indicated on the title
and the blue curves, red curves represent the evolution of confirmed cases and
undetected cases. In the title “AVG” is the average number of confirmed case
at Day 50 (about the end of Mars) and “POT” is that of undetected cases.

(Tianyuan ZHAO) Ecole polytechnique, Paris, France

Email address: [email protected]

(Yiyang YU) LPSM, Université Paris Diderot, Paris, France

Email address: [email protected]

(Wei JIANG) Inria XPOP and CMAP, Ecole Polytechnique, France

Email address: [email protected]

7
Figure 6. A simulation of the evolution under the measure of reduction of
activities. Compared to the Group B in fig. 5 where θ = 0.5, α = 0.4 is a
normal measure, but the reduction of activities by 41 stops the diffusion.