key: cord-0997289-kvgxat6s
authors: Dashtbali, Mohammadali; Mirzaie, Mehdi
title: A compartmental model that predicts the effect of social distancing and vaccination on controlling COVID-19
date: 2021-04-14
journal: Sci Rep
DOI: 10.1038/s41598-021-86873-0
sha: a4e3ba6d2943f5f5cad649d3bf66f6d38ff86544
doc_id: 997289
cord_uid: kvgxat6s

The understanding of the interaction between disease dynamics and human behavior is an important and essential point to control infectious. Disease outbreak can be influenced by social distancing and vaccination. In this study, we introduce two compartmental models to derive the epidemic curve and analyze the individual’s behavior in spreading and controlling the COVID-19 epidemic. The first model includes Susceptible, Exposed, Infectious, Hospitalized, Recovered and Death compartments and in the second model, we added a new compartment namely, semi-susceptible individuals that are assumed to be more immune than the susceptible. A comparison of the two models shows that the second model provides a better fit to the daily infected cases from Egypt, Belgium, Japan, Nigeria, Italy, and Germany released by WHO. Finally, we added a vaccinated term to the model to predict how vaccination could control the epidemic. The model was applied on the record data from WHO.

reduce and control the spread of an epidemic was studied by Reluga 14 . According to optimal control and differential game approaches, the impact of social distancing in the spread of infectious diseases on the network was studied in 2020 15 .

Vaccination policy is related to the population health adopted by the government. Eradicating the disease or the immunity of the population is the aim of vaccination. Individuals' deciding about vaccination is involved considering the risk of morbidity from vaccination and the probability of getting infected [16] [17] [18] [19] . On the other hand, each person's decisions in vaccination are influenced by the decisions of others in social distancing 20, 21 . In this study, two models have been studied. The first model consists of seven compartments including the Susceptible (S), Exposed (E), Infected (I), Hospitalized (H), Recovered (R), and Death (D) individuals, called SEIHRD. After getting infected, an individual enters the hospitalized compartment and at the end of this state either recovers or dies. The model also considers social distancing and we assume that number of dead people among the total number of infected individuals could affect the behavior of people in social distancing. Some researches indicate that some people are less likely to become infected. In the second model, we added a compartment, Semi-susceptible (M) and assume that these individuals are half as susceptible to COVID-19. The new model is called SMEIHRDV model. The results of the two models were compared using WHO data from Japan, Italy, Belgium, Germany, Nigeria and Egypt. The second model was fitted better than the first one. Finally, the SMEIHRDV model was simulated for the different coverages of vaccination to predict how vaccination can control the epidemic.

The main purpose of this study is to develop a compartmental model to fit the released data by WHO and could predict the behavior of epidemic. This model can be used to investigate how vaccination coverage will control the epidemic. Social distancing is one of the best strategies limiting contact rate and reducing transmission rates. A differential game theory model also introduced to investigate the role of individual's investment in social distancing for controlling the epidemic. The organization of this paper is as follows: in "SEIHRDV model" we present the SEIHRDV model. In "SMEIHRDV model", the SEIHRDV model is expanded to the SMEIHRDV model. The game theory of social distancing is considered in "The game theory of social distancing". Next, results and discussion are presented in "Results and discussion".

The most commonly implemented dynamical model in infectious disease is the SIR model, consists of three compartments including susceptible, infected and recovered individuals represented by S, I and R, respectively. In order to represent the number of susceptible, infected and recovered individuals with respect to time, these variables are considered as function of t (time): S(t), I(t) and R(t). The dynamics of SIR model is as follows:

Letting total population size denoted by N , is constant and N = S(t) + I(t) + R(t) . Several compartmental models have been extended from the SIR model. According to the reported characteristics of the COVID-19 disease, we develop a mathematical model consists of seven compartments:

1. Susceptible individuals denoted by S who are not infected by the disease pathogen, 2. Exposed individuals denoted by E. The persons in the incubation period after being infected by the virus.

They have no visible clinical signs but infectious. 3. Infected individuals denoted by I. After the incubation period, it is the first compartment of the infectious period, where the person has finished the incubation period, may infect other people and starts developing clinical signs. After this period, people needed to be hospitalized (in a hospital or at home), therefore the next state is: 4. Hospitalized individuals denoted by H. After the hospitalization period, people can be, either recovered or removed. Therefore, we have two other compartments: 5. Recovered individuals denoted by R , and 6. Death individuals denoted by D. In order to predict the effect of vaccination, after releasing the vaccine, we considered the vaccinated compartment. So: 7. Vaccinated individuals denoted by V. The suggested dynamics of this model (called SEIHRDV) is as follows: 

where N is the number of people in the population.

The diagram for this model is depicted in Fig. 1 .

Social distancing is a non-pharmaceutical strategy to reduce and control the spread of an epidemic 14 . It could decrease the transmission rate between susceptible and infected individuals and affects the dynamics of an epidemic. In this study, we consider behavior of individuals by simple function of the available information. In fact we assume that susceptible individuals change their contact rate based on the number of confirmed death per day in a country. Also, in order to investigate the effectiveness of social distancing during time, we use function σ(t) as it was discussed by Reluga 14 . Therefore, the system of Eq. (2) is modified as follows:

(2) where c(t) is, the ratio of the number of daily died COVID-19 patients at time t to the number of daily diagnosed COVID-19 patients at time t . In order to consider individual heterogeneities such as differences in social activity, mixing patterns including household, spatial aspects, and travels, we used a parameter, namely p, in the model which is calibrated to fit the model to data. The selected values of p −1 for different countries are as follows: Egypt: 0.47, Germany: 0.50, Belgium: 0.54, Japan: 0.55, Italy: 0.56, Nigeria: 0.57.

Elderly and sick people are most susceptible to severe forms of COVID-19. There have been studies to discover why some individuals are more susceptible than the others 22 . For example, some studies show that people carrying certain variants of the ACE2 gene would be protected against the COVID-19 infection 23, 24 . Therefore we divided susceptible individuals into two compartments, the susceptible compartment denoted by S and the semi-susceptible denoted by M. In compartment S, people are more susceptible than M. The new model is called the SMEIHRDV model. A diagram summarizing the main structure of our model is presented in Fig. 2 . β 1 is the transmission rate between susceptible and infected individuals and β 2 is the transmission rate between semi-susceptible and infected individuals. The remaining parameters are the same as the SEIHRDV model. The dynamics of this model is as follows: 

Social distancing as a strategy limiting contact rates and reducing transmission rate has costs in terms of money, time and convenience. Reluga 14 introduced a population game where the payoff of individuals is characterized by the individual's behavioral strategy and the average behavioral strategy adopted by the population. Let c S the aggregate daily investment adopted by the susceptible population and c S the specific susceptible individual's investment. The impact of social distancing is defined by a function σ(c S ) which is the relative risk of disease at daily investment c S in social distancing. Here, we consider a differential game for a susceptible individual choosing his or her best investment strategy in social distancing relative to the aggregate investment strategy of susceptible population as whole and a differential game for a semi-susceptible individual choosing his or her best investment strategy in social distancing relative to the aggregate investment strategy of semi-susceptible population as whole. The modified version of the equations with social distancing according to differential game approach is as follows:

c I and c S (t) the aggregate investment in social distancing by the susceptible people. 4. c 2 (t) = c M (t) c I and c M (t) the aggregate investment in social distancing by semi-susceptible individuals. 5. c I the cost of infection. 6. σ is the relative risk and σ(c 1 (t)) = 1 p(1+c 1 (t)) , σ(c 2 (t)) = 1 p(1+c 2 (t)) .

where N is the number of people in the population and the parameters are described in Table 1 . Note that if the investment in social distancing exceeds the cost of infection, then social distancing is not economical, and so 0 ≤ c 1 (t) ≤ 1 and 0 ≤ c 2 (t) ≤ 1. The dimensionless version of the equations is obtained as follows (let www.nature.com/scientificreports/ After this, the hat-notation will be dropped and the dimensionless parameters will be used. The expected present values are calculated (the expected present value is the average value one expects after accounting for the probabilities of all future events, and discounting future costs relative to immediate costs) because of the exact time spent and the precise payoff are not predicted and so we use the Markov process. For each state, the expected present values denoted by V and justify in the following equation 14 :

where h is the discount rate. We will take h = 0 as discassed by Reluga 25 . Als o, v(t; c 1 (t), c 2 (t)) = (−c 1 (t), −c 2 (t), −1 + (c 1 (t)+c 2 (t)) 2 , −1 + (c 1 (t)+c 2 (t)) 2 , 0, 0, 0, 0) and Therefore, we have: Now the purpose is finding c 1 * (t) and c 2 * (t) , so it is used the maximum principle:

Infectious diseases are a threat to the health of the population. One of the easiest ways for individuals to reduce the risk of infection during an epidemic is to decrease their rate of contact with infected individuals. However, the value of these actions depends on how the epidemic progresses. Few cases analyses indicate how changes in behavior will change the epidemic wave. The policies included large-scale quarantine, strict controls on travel, and extensive monitoring of suspected cases, and as a rule, social distancing was applied by individuals and respectively. These parameters were estimated using WHO data for each country. In order to reduce and control the spread of COVID-19, social distancing is a non-pharmaceutical strategy that was considered in the model and the system (2) modified to the system (3). This system solved numerically by using the estimated parameters in Table 2 . Social distancing is an attitude of behavior and changing the behaviors can decrease contact rates and that makes to reduce the transmission of infectious and control the diseases. The number of confirmed death per day published by several media can affect the behavior of individuals in social distancing. Therefore, here we approximate the social distancing effect as the ratio of the number of daily died at time t to the number of accumulated patients at time t. In the second model, we assume that some part of population have lower susceptibility to infection. In general, factors such as age, smoking, obesity and genetic factors can affect disease susceptibility 26 . Therefore, we added a new compartment called semi-susceptible denoted by M, and assume that their infection rate to COVID-19 is half the infection rate of susceptible people to it. The new model is SMEIHRDV. In this new model, for each country, M 0 = 15 100 N and the rates of changes in states are, β 1 (S to E), β 2 = 1 2 β 1 (M to E) and the remaining parameters are the same as the SEIHRDV model. Social distancing was considered in this model and the system (4) was solved numerically, by using the estimated parameters in Table 2 . The confirmed and predicted infected cases of the SMEIHRDV and SEIHRDV models is shown in Fig. 3 , for Japan (from 22th January to 23th June 2020), Italy (from 31th January to 23th June 2020), Belgium (from 4th February to 23th June 2020), Germany (from 27th January to 23th June 2020), Nigeria (from 28th February to 23th June 2020) and Egypt (from 14th February to 23th June 2020). The blue dots denote the reported infected cases and the black (or less width) and blue lines represent the model simulation results, respectively. The comparison of the results of two models shows that the SMEIHRDV model provides better estimation with the reported WHO data. Therefore, the semisusceptible compartment improves the result of the model in the prediction of infected individuals.

In the following, the game theory of social distancing was presented and in order to obtain the best investment strategy in social distancing, we used the dimensionless version of the dynamic the system (5) and a differential game for a susceptible individual choosing his or her best investment strategy in social distancing relative to the aggregate investment strategy of susceptible population as whole and a differential game for a semi-susceptible individual choosing his or her best investment strategy in social distancing relative to the aggregate investment strategy of semi-susceptible population as whole. Since the exact time spent and the precise payoff are not predicted, so we used the Markov process. The system (6) that is a dimensionless version of the SMEIHRDV model, was used to determine the transition-rate matrix. This model was simulated to investigate the effect of vaccination and the investment in social distancing to drop the epidemic peak. The aggregate investment strategy of susceptible individuals denoted by c 1 (t) and the aggregate investment strategy of semi-susceptible denoted by c 2 (t) . Here, it was used the maximum principle to find the best strategy of the investment in social distancing for a susceptible individual ( c 1 * ) and for a semi-susceptible individual ( c 2 * ), and they were obtained in (12) and (13) . Finally, the SMEIHRDV model was simulated for when the vaccine was released for Coronavirus. We investigate the effect of the behavior of individuals in vaccination in the epidemic process. The different vaccine coverage were examined in the epidemic process for the infected cases in Egypt (see Fig. 4 ) and Germany (see Fig. 5 ). The number of total infected individuals ( at the epidemic peak) in Egypt decreased from 540 to 200 as the vaccination coverage changes from 0.2 to 0.6 (see Fig. 4 ). Also, in Germany, the number of infected cases at the epidemic peak is approximately 320 and it has reduced to 120 by the increasing vaccine coverage from 0.2 to Table 2 . Parameters estimated from the observed data in Germany (DE), Italy (IT), Belguim (BE), Egypt (Eg), Nigeria (NG) and Japan (JP) inferred from the World Health Organization. www.nature.com/scientificreports/ 0.6 (see Fig. 5 ). In Fig. 6 , the effect of the best investment strategy in social distancing was studied in COVID-19 progress in Germany; the forecast of the epidemic growing tendency for infected cases through the SMEIHRDV model with the best investment strategy in social distancing. The results show that the best investment strategy in social distancing reduces the epidemic peak for infected cases and also by increasing the vaccine coverage, the epidemic peak for infected cases in Germany decreases. The number of infected cases at the epidemic peak is approximately 200 and it has reduced to 55 by the increasing vaccine coverage from 0.2 to 0.6.

Fitting the models with real data. The number of infected individuals predicted by the "SEIHRDV" and the "SMEIHRDV" models for Japan (from 22th January to 23th June 2020), Italy (from 31th January to 23th June 2020), Belgium (from 4th February to 23th June 2020), Germany (from 27th January to 23th June 2020), Nigeria (from 28th February to 23th June 2020) and Egypt (from 14th February to 23th June 2020). The blue dots denote the reported infected cases and the black (or Less width) and blue lines present the results of the "SEIHRDV" and the "SMEIHRDV" models, respectively. The deSolve package version 1.28 27 The results show that the number of infected cases at the epidemic peak is approximately 320 in Germany and it has reduced to 120 by the increasing vaccine coverage from 0.2 to 0.6. N = 80, 000, 000. Figure 6 . Simulation, the forecast of the epidemic process. The forecast of the epidemic growing tendency for infected cases through the SMEIHRDV model with the best investment strategy in social distancing. The results show that the best investment strategy in social distancing reduces the epidemic peak for infected cases and also by increasing the vaccine coverage, the epidemic peak for infected cases decreases. The number of infected cases at the epidemic peak is approximately 200 in Germany and it has reduced to 55 by the increasing vaccine coverage from 0.2 to 0.6. N = 80, 000, 000.

Emergence of a novel swine-origin influenza A (H1N1) virus in humans

Threshold dynamics of a stochastic model with vertical transmission and vaccination

Coronavirus disease 2019 (COVID-19): A literature review

Coronavirus disease 2019-COVID-19

A contribution to the mathematical theory of epidemics

A quantitative and qualitative analysis of the COVID-19 pandemic model

Data-based analysis, modelling and forecasting of the COVID-19 outbreak

Mathematical modelling of COVID-19 transmission and mitigation strategies in the population of Ontario, Canada

Epidemiological models are important tools for guiding COVID-19 interventions

Forecasting the spread of COVID-19 in Kuwait using compartmental and logistic regression models

Evaluation of the secondary transmission pattern and epidemic prediction of COVID-19 in the four metropolitan areas of China

Covid-19 spread: Reproduction of data and prediction using a SIR model on Euclidean network

The effect of human mobility and control measures on the COVID-19 epidemic in China

Game theory of social distancing in response to an epidemic

Optimal control and differential game solutions for social distancing in response to epidemics of infectious diseases on networks

The pertussis vaccine controversy in Great Britain

Measles outbreaks in a population with declining vaccine uptake

Simple rules with complex dynamics

Group interest versus self-interest in smallpox vaccination policy

Theory of Games and Economic Behavior

Evolution and the Theory of Games

The preventive strategy for pandemics in the elderly is to collect in advance samples and data to counteract chronic inflammation (inflammaging)

COVID-19 infection and circulating ACE2 levels: protective role in women and children

ACE2: The molecular doorway to SARS-CoV-2

The discounted reproductive number for epidemiology

Age-dependent effects in the transmission and control of COVID-19 epidemics

Solving differential equations in R: Package deSolve

R Core Team R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing

This research was supported by Research Core: "Bio-Mathematics with computational approach" of Tarbiat Modares University, with Grant number "IG-39706". The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

M.M conceived and designed the study. M.D developed the model and performed the analysis. Both authors contributed to writing the manuscript and participated in the discussion of the results and contributed to the revision of the final manuscript.

The authors declare no competing interests.

Correspondence and requests for materials should be addressed to M.M.Reprints and permissions information is available at www.nature.com/reprints.Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.