Loading [MathJax]/jax/output/SVG/jax.js

On a model of COVID-19 dynamics

  • Received: 01 June 2020 Revised: 01 September 2020 Published: 19 October 2020
  • Primary:37B25, 92D30;Secondary:92D25

  • A model of COVID-19 in an interconnected network of communities is studied. This model considers the dynamics of susceptible, asymptomatic and symptomatic individuals, deceased but not yet buried people, as well as the dynamics of the virus or pathogen at connected nodes or communities. People can move between communities carrying the virus to any node in the region of n communities (or patches). This model considers both virus direct (person to person) and indirect (contaminated environment to person) transmissions. Using either matrix and graph-theoretic methods and some combinatorial identities, appropriate Lyapunov functions are constructed to study global stability properties of both the disease-free and the endemic equilibrium of the corresponding system of 5n differential equations.

    Citation: Jorge Rebaza. On a model of COVID-19 dynamics[J]. Electronic Research Archive, 2021, 29(2): 2129-2140. doi: 10.3934/era.2020108

    Related Papers:

    [1] Jorge Rebaza . On a model of COVID-19 dynamics. Electronic Research Archive, 2021, 29(2): 2129-2140. doi: 10.3934/era.2020108
    [2] Gaohui Fan, Ning Li . Application and analysis of a model with environmental transmission in a periodic environment. Electronic Research Archive, 2023, 31(9): 5815-5844. doi: 10.3934/era.2023296
    [3] Zimeng Lv, Jiahong Zeng, Yuting Ding, Xinyu Liu . Stability analysis of time-delayed SAIR model for duration of vaccine in the context of temporary immunity for COVID-19 situation. Electronic Research Archive, 2023, 31(2): 1004-1030. doi: 10.3934/era.2023050
    [4] Comfort Ohajunwa, Carmen Caiseda, Padmanabhan Seshaiyer . Computational modeling, analysis and simulation for lockdown dynamics of COVID-19 and domestic violence. Electronic Research Archive, 2022, 30(7): 2446-2464. doi: 10.3934/era.2022125
    [5] Hao Nong, Yitan Guan, Yuanying Jiang . Identifying the volatility spillover risks between crude oil prices and China's clean energy market. Electronic Research Archive, 2022, 30(12): 4593-4618. doi: 10.3934/era.2022233
    [6] Yazao Yang, Haodong Tang, Tangzheng Weng . Changes in public travel willingness in the post-COVID-19 era: Evidence from social network data. Electronic Research Archive, 2023, 31(7): 3688-3703. doi: 10.3934/era.2023187
    [7] Khongorzul Dashdondov, Mi-Hye Kim, Mi-Hwa Song . Deep autoencoders and multivariate analysis for enhanced hypertension detection during the COVID-19 era. Electronic Research Archive, 2024, 32(5): 3202-3229. doi: 10.3934/era.2024147
    [8] S. Suganya, V. Parthiban, R Kavikumar, Oh-Min Kwon . Transmission dynamics and stability of fractional order derivative model for COVID-19 epidemic with optimal control analysis. Electronic Research Archive, 2025, 33(4): 2172-2194. doi: 10.3934/era.2025095
    [9] Zhiliang Li, Lijun Pei, Guangcai Duan, Shuaiyin Chen . A non-autonomous time-delayed SIR model for COVID-19 epidemics prediction in China during the transmission of Omicron variant. Electronic Research Archive, 2024, 32(3): 2203-2228. doi: 10.3934/era.2024100
    [10] Chengtian Ouyang, Huichuang Wu, Jiaying Shen, Yangyang Zheng, Rui Li, Yilin Yao, Lin Zhang . IEDO-net: Optimized Resnet50 for the classification of COVID-19. Electronic Research Archive, 2023, 31(12): 7578-7601. doi: 10.3934/era.2023383
  • A model of COVID-19 in an interconnected network of communities is studied. This model considers the dynamics of susceptible, asymptomatic and symptomatic individuals, deceased but not yet buried people, as well as the dynamics of the virus or pathogen at connected nodes or communities. People can move between communities carrying the virus to any node in the region of n communities (or patches). This model considers both virus direct (person to person) and indirect (contaminated environment to person) transmissions. Using either matrix and graph-theoretic methods and some combinatorial identities, appropriate Lyapunov functions are constructed to study global stability properties of both the disease-free and the endemic equilibrium of the corresponding system of 5n differential equations.



    Since the announcement of the novel virus SARS-CoV-2 affecting the population of Wuhan, China in December 2019, the disease caused by it, COVID-19, has extended from a small cluster in that region in China to almost every corner of the world. With globalization, countries are more interconnected than ever (with direct flights from Wuhan to the U.S. included), so human mobility is a factor that should be included in the mathematical modeling of this disease. Most current models of COVID-19 consider only one community or node (no human mobility), only direct or indirect transmission and do not include pathogen dynamics [10, 13, 18, 26]. These and other models represent in part an extension of previous research on other epidemics including cholera, as well as Zika and Ebola virus diseases [2, 6, 9, 12, 14, 17, 19, 21, 23].

    One of the main goals in this type of studies is to establish conditions under which the disease is expected to persist or die out, and to study possible control measures. To this end, it is important to study conditions for the existence and local or global stability of the steady states in the model.

    The model considered here includes human mobility across n nodes (e.g. cities or countries) connected by a human mobility network (that is, people can move from node to node carrying the disease or being exposed to it), and both direct and indirect transmissions are considered simultaneously. In addition, the dynamics of concentration of the pathogen itself at different communities is also included in this model. Disease transmission can occur through person-to-person contact or through exposure to the virus in a local or visited community. The rate of exposure, contact and transmissibility vary with the parameters of the model. In practice, this is related to mitigation measures including travel restrictions, partial or complete isolation, social distancing, mask wearing, hand washing, and environment disinfection strategies. In this kind of network connectivity models, the condition that one or more of the individual basic reproduction numbers R(i)0,i=1,,n is greater than one, is neither sufficient nor necessary for an outbreak in the metacommunity to occur, and the disease can start even if the basic reproduction number of each community is less than one [9].

    As with similar mathematical models, conditions under which the disease is expected to spread or die out are related to the stability properties of the disease-free equilibrium point, and to the value of the basic reproduction number. Here we prove that for the corresponding system of 5n differential equations, if the basic reproduction number R0 satisfies R01, then the disease-free equilibrium point is globally stable, and otherwise it is unstable, in which case an endemic equilibrium exists and is globally asymptotically stable. These stability properties are proved using matrix and graph-theoretic methods introduced in [16, 20].

    Researchers are still learning about the dynamics of COVID-19, which has been affecting regions and countries around the world, especially China, USA, Europe and Latin America. In this theoretical model, we assume that people can move with certain freedom between the n nodes in the metacommunity, possibly carrying the disease and spreading it by direct person-to-person contact or by shedding the virus in the environment on a given node or community. Indirect transmission is then possible by exposure to contaminated surfaces at any given node. As noted in [4, 10, 27], both asymptomatic and symptomatic carriers are able to spread the disease, although at different rates and probably at different timings of pathogen shedding. Therefore, we split infected people into asymptomatic and symptomatic variables, with distinct transmission probabilities, as well as distinct transition and shedding rates in this model. It is important to stress that in this work, asymptomatic refers to people carrying the virus but who never develop symptoms, while symptomatic refers to people also carrying the virus but showing mild to severe symptoms. It is also assumed that individuals leave their original community or node i with constant probabilities mS and mI for susceptibles and infectious (symptomatic and asymptomatic) respectively, reach their target node j with a probability Qij and come back to node i. It is natural to assume that the graph Γ(Q) is strongly connected, meaning that people can move from any node to any other node in the network. Then, the matrix Q describing the connections between communities is stochastic, with Qii=0 for each i=1,,n. An explicit definition of Q following a gravity model (relative attraction of communities - related to population size and distance between communities) is given in [9].

    We let Si,Ai,Ii,Di and Vi represent the population densities of susceptibles, asymptomatic, symptomatic, disease-induced deceased but not yet buried, and virus concentration respectively at the i-th community, with i=1,...,n. With the virus currently moving its epicenter from the US and Europe to Latin America and probably later to Africa, it is important to consider the increased possibility of virus transmission by infected deceased people in the days before their burial (see e.g. [3, 8]). With all parameters being positive, and all variables nonnegative, the corresponding system of 5n differential equations is:

    Si=Λi[(1ms)αip(Ei)+msnj=1Qijαjp(Ej)]Si[(1ms)βig(Vi)+msnj=1Qijβjg(Vj)]SiμSiAi=θ[(1ms)αip(Ei)+msnj=1Qijαjp(Ej)]Si+θ[(1ms)βig(Vi)+msnj=1Qijβjg(Vj)]SiϕiAiIi=(1θ)[(1ms)αip(Ei)+msnj=1Qijαjp(Ej)]Si+(1θ)[(1ms)βig(Vi)+msnj=1Qijβjg(Vj)]Si˜ϕiIiDi=˜ξiIiμDDiVi=pi[(1mI)(λAi+Ii)+˜λDi+nj=1mIQji(λAj+Ij)]μVVi, (1)

    where Λi is the constant recruitment into the i-th community, and the function p(Ei)=p(Ai,Ii,Di)=τAi+Ii+˜τDiHi is the probability of direct transmission; τ and ˜τ represent the transmissibility of asymptomatic and dead but not yet buried respectively to susceptible people; Hi is a constant bound on the total population in the i-th community; αi is the rate of direct transmission in node i, g(Vi)=ViK+Vi (where K is the half-saturation constant) is the probability of indirect transmission; that is, the probability of becoming infected due to exposure to the virus in node i; βi is the rate of exposure to virus-contaminated environment; μ is the natural human death rate; ϕi=γi+ξi+μ represents the rates of recovery, disease-induced and natural death respectively of asymptomatic people in node i; similarly for ˜ϕi=˜γi+˜ξi+μ with respect to symptomatic people; θ represents the fraction of asymptomatic infections; pi is the rate at which virus is released by infected individuals into the environment (in the air or on surfaces) in node i, with λ and ˜λ representing the different contributions of asymptomatic and deceased but not yet buried humans to virus concentration as compared to symptomatic ones; μV is the rate of death of the virus (not anymore viable or capable of infection); 1/μD is the mean time from COVID-19 death to burial. For some appropriate estimates of the values of some of these parameters, see [10, 13, 26].

    Thus, according to model (1), susceptible individuals can get infected via direct or indirect transmission in any given community (including their own) where infected people or contaminated surfaces are present. If infected, they move to asymptomatic or symptomatic stage, from which they either recover (most of them) or die of natural or induced-disease causes. Infected individuals are able to shed the virus in their own community and in communities they visit.

    Denoting Ni=Si+Ai+Ii, for i=1,,n, we get NiΛiμNi, and using a differential inequality in Section I.6 in [11], one can show that NiΛiμ=:Hi. Similarly, if we let Mi=maxAi,Ii[0,Hi](1mI)(λAi+Ii)+nj=1mIQji(λAj+Ij), we get ViMiμVVi, and therefore we get ViMiμV. In the equation for Di in (1), we get Di˜ξiHiμDDi and hence Di˜ξiHiμD=:Li. Thus, the feasible region for model (1) is:

    Γ={(S,A,I,D,V)R5n+:NiHi,DiLi,ViMiμV,i=1,n}. (2)

    Observe that if we write system (1) as zi=Fi(z), with i=1,,5n, then zi=0 clearly implies Fi(z)0, and hence the nonnegative cone (zi0,i=1,,5n) is invariant. This means the feasible region Γ in (2) is compact and (positively) invariant.

    System (1) has a unique disease-free equilibrium point (DFE):

    E0=(S0i,A0i,I0i,D0i,V0i)=(Λiμ,0,0,0,0)R5n+. (3)

    The system also has an endemic equilibrium point, which will be discussed in Section 4.

    In this section, we establish conditions under which the disease will persist or die out. In addition to the roles played by the problem parameters and by the density or concentration of each population (people and virus), the mobility of infected people throughout the network also plays a key role. We start by studying the stability properties of the disease-free equilibrium (3).

    To study global stability of the DFE, we can write system (1) as a compartmental model, by splitting the variables into two compartments: a disease compartment xR4n and a nondisease compartment yRn:

    x=[A1,,An,I1,In,D1,,Dn,V1,,Vn]T,andy=[S1,,Sn].T

    For i=1,,n, denote

    Ri=(1ms)αip(Ei)+msnj=1Qijαjp(Ej),Ti=(1ms)βig(Vi)+msnj=1Qijβjg(Vj),

    and hi=pi[(1mI)(λAi+Ii)+˜τDi+nj=1mIQji(λAj+Ij)]. Then, we let

    F(x,y)=[θR1S1+θT1S1θRnSn+θTnSn(1θ)R1S1+(1θ)T1S1(1θ)RnSn+(1θ)TnSn˜ξ1I1˜ξnInh1hn],V(x,y)=[ϕ1A1ϕnAn˜ϕ1I1˜ϕnIn μDD1μDDnμVV1μVVn].

    In general, one denotes with each entry Fi the rate secondary infections increase the i-th disease compartment, and with Vi the rate disease progression, death and recovery decrease in the i-th compartment [25]. Also note that all the conditions in [20]: Fi(0,y)=0,Vi(0,y)=0,Fi(x,y)0,Vi(x,y)0 when xi=0, and Vi(x,y)0, for i=1,...,4n hold.

    With this notation, model (1) can be split into disease/nondisease systems:

    x=F(x,y)V(x,y),y=G(x,y). (4)

    Denote y0=(Λ1μ,,Λnμ). Then, following [20, 24], we let

    F=[Fixj(0,y0)],andV=[Vixj(0,y0)],1i,j4n. (5)

    Thus, the matrices F and V have the following form

    F=[θC1θC2θC3θC4(1θ)C1(1θ)C2(1θ)C3(1θ)C40˜D00C5C6˜τI0],
    V=[diag(ϕ1,,ϕn)diag(˜ϕ1,,˜ϕn)μDIμBI],

    where each Ci(i=1,,6) is a positive, square matrix of order n, I represents the n×n identity matrix, and ˜D=diag(˜ξ1,,˜ξn). Thus, we have F0, V is invertible, and also V10.

    The basic reproduction number R0 of (1), defined as the expected number of secondary cases produced by an infected individual in a completely susceptible population [7, 24], is the spectral radius ρ of the Next Generation Matrix (NGM): R0=ρ(FV1). On the other hand, one can readily see that the matrix V1F is nonnegative and irreducible, guaranteeing that R0 is a positive eigenvalue of V1F (and of FV1) and that there exists a positive left eigenvector ω of V1F corresponding to R0 [1]. This eigenvector ω will be used in the definition of a Lyapunov function below, which in turn is used to prove the following stability result.

    Theorem 3.1. If R01, then the disease-free equilibrium point E0 (3) is globally asymptotically stable (GAS) in Γ.

    Proof. Following the matrix-theoretic method in [20], we set

    f(x,y):=(FV)xF(x,y)+V(x,y).

    Thus, f(x,y) is the vector

    [θR1(S01S1)+θ[(1mS)β1V1(S01KS1K+V1)+mSnj=1Q1jβjVj(S01KS1K+Vj)]θRn(S0nSn)+θ[(1mS)βnVn(S0nKSnK+Vn)+mSnj=1QnjβjVj(S0nKSnK+Vj)](1θ)R1(S01S1)+(1θ)[(1mS)β1V1(S01KS1K+V1)+mSnj=1Q1jβjVj(S01KS1K+Vj)](1θ)Rn(S0nSn)+(1θ)[(1mS)βnVn(S0nKSnK+Vn)+mSnj=1QnjβjVj(S0nKSnK+Vj)]00].

    Observe that f(x,y)0. Then, by applying Theorem 2.1 in [20], we see that

    D=ωTV1x

    is a Lyapunov function for (1) in Γ, when R01. To prove global stability, first consider R0<1. One can verify that the n-dimensional disease-free system associated to (1) has a unique equilibrium point y0=(Λ1μ,,Λnμ), and that this equilibrium point is GAS in Rn+. One can also see that f(0,y)=0 in Γ. Then, applying the first part of Theorem 2.2 in [20], we conclude that if\, R0<1, the DFE (3) is GAS in Γ.

    Now let R0=1. Observe that

    D=ωTV1x=ωTV1(FV)xωTV1f(x,y)=(R01)ωTxωTV1f(x,y).

    Then, D=0 implies f(x,y)=0, as ω is strictly positive and all diagonal entries of V1 are also strictly positive. Then, we get Si=S0i and Vi=0=V0i, for i=1,,n – see notation in (3). One can show that the remaining 3n-dimensional system in Ai,Ii and Di has the unique equilibrium (A0i,I0i,D0i)=(0,0,0)R3n, and that this equilibrium is GAS in R3n+. But we already have Si=S0i and Vi=V0i; thus, the largest and only invariant set where D=0 is the DFE. Using LaSalle's invariance principle, one concludes that the DFE (3) is GAS in Γ, when R0=1.

    Existence of an Endemic Equilibrium. In Theorem 3.1, we proved that if R01, then the disease-free equilibrium point (DFE) is globally asymptotically stable in the given feasible region Γ. When this condition is reversed; that is, when we instead have R0>1, the DFE loses stability (see e.g. Theorem 2 in [24], or Theorem 1 in [25]). As noted in the proof of Theorem 2.2 in [20], this instability implies uniform persistence of (1), and by Theorem D.3 in [22], the invariance of Γ and the uniform persistence imply the existence of an endemic equilibrium (EE). We will denote this endemic equilibrium point of (1) as

    E:=(Si,Ai,Ii,Di,Vi)R5n+. (6)

    In Theorem 4.2 below, we show that the condition R0>1 also implies that the EE is globally asymptotically stable.

    To study global stability of the EE (6), we use a graph-theoretic method as presented in [20]. First, we summarize some basic terminology and results from [16, 20] on directed graphs, including a technique for the construction of a Lyapunov function.

    Given a weighted digraph Γ(A) with p vertices, the p×p weight matrix A is defined with aij>0 equal to the weight of arc (j,i) if it exists, and aij=0 otherwise. The Laplacian L of Γ(A) is defined as

    lij={aij,ij,kiaik,i=j.

    Let ci be the cofactor of lii. If Γ(A) is strongly connected, then ci>0, for all i=1,,p. The following combinatorial identities are useful in finding explicit expressions for ci:

    If aij>0 and the out-degree of vertex j satisfies d+(j)=1, for some i,j, then

    ciaij=pk=1cjajk. (7)

    If aij>0 and the in-degree of vertex i satisfies d(i)=1, for some i,j, then

    ciaij=pk=1ckaki. (8)

    The following theorem provides a graph-theoretic technique to construct a Lyapunov function D.

    Theorem 4.1. [16, 20] For a given open set ERm, and a function f:ERm, consider the system

    ˙z=f(z), (9)

    and assume that

    (i) There exist functions Qi:ER, Gij:ER, and constants aij0 such that

    Qi=Qi|(9)pj=1aijGij(z),withzE,i=1,,p,

    (ii) Each directed cycle C of Γ(A) satisfies

    (s,r)S(C)Grs(z)0,zE,

    where S(C) denotes the set of all arcs in C.

    Then, there exist constants ci0,i=1,,p (as defined above), such that the function

    Q(z)=pi=1ciQi(z)

    satisfies Q|(9)0; that is, Q(z) is a Lyapunov function for (9).

    With these tools at hand, we can give a result on global stability of the endemic equilibrium of system (1). In the theorem below, we use the functions p(Ej) and g(Vj) as defined in (1). We also use the following notations:

    bij={(1mS)αi,j=imSQijαj,ji,cij={(1mS)βi,j=imSQijβj,ji,

    and hi as in Section 3.

    Theorem 4.2. If R0>1, then, the EE of system (1) is unique and globally asymptotically stable (GAS) in int (Γ).

    Proof. For i=1,,n, define

    Qi=SiSiSiln(SiSi)+AiAiAiln(AiAi)+IiIiIiln(IiIi).

    Then, taking derivatives, and solving for Λi,ϕi and ˜ϕi at the equilibrium equations in (1), we have

    Qi=(1SiSi)Si+(1AiAi)Ai+(1IiIi)Ii=(1SiSi)[nj=1bij(Sip(Ej)Sip(Ej))+nj=1cij(Sig(Vj)Sig(Vj))μ(SiSi)]+(1AiAi)[θnj=1bij(Sip(Ej)Sip(Ej)AiAi)+θnj=1cij(Sig(Vj)Sig(Vj)AiAi)]+(1IiIi)[(1θ)nj=1bij(Sip(Ej)Sip(Ej)IiIi)+(1θ)nj=1cij(Sig(Vj)Sig(Vj)IiIi)]nj=1bijSip(Ej)[1Sip(Ej)Sip(Ej)SiSi+p(Ej)p(Ej)+θ (Sip(Ej)Sip(Ej)AiAiSip(Ej)Sip(Ej)AiAi+1)+(1θ)(Sip(Ej)Sip(Ej)IiIiSip(Ej)Sip(Ej)IiIi+1)]+nj=1cijSig(Vj)[1Sig(Vj)Sig(Vj)SiSi+g(Vj)g(Vj)+θ (Sig(Vj)Sig(Vj)AiAiSig(Vj)Sig(Vj)AiAi+1)+(1θ)(Sig(Vj)Sig(Vj)IiIiSig(Vj)Sig(Vj)IiIi+1)]=nj=1bijSip(Ej)[2SiSi+p(Ej)p(Ej)(θAiAi+(1θ)IiIi)Sip(Ej)Sip(Ej)(θAiAi+(1θ)IiIi)]+nj=1cijSig(Vj)[2SiSi+g(Vj)g(Vj)(θAiAi+(1θ)IiIi)Sig(Vj)Sig(Vj)(θAiAi+(1θ)IiIi)]nj=1bijSip(Ej)[lnSiSi+ln(Sip(Ej)Sip(Ej)AiIiθAiIi+(1θ)IiAi)+p(Ej)p(Ej)θIiAi+(1θ)AiIiAiIi]+nj=1cijSig(Vj)[lnSiSi+ln(Sig(Vj)Sig(Vj)AiIiθAiIi+(1θ)IiAi)+g(Vj)g(Vj)θIiAi+(1θ)AiIiAiIi]nj=1bijSip(Ej)[ln(θIiAi+(1θ)AiIiAiIi)+p(Ej)p(Ej)θIjAj+(1θ)AjIjAjIj1ln(θIjAj+(1θ)AjIjAjIj)+p(Ej)p(Ej)θIiAi+(1θ)AiIiAiIi]+nj=1cijSig(Vj)[ln(θIiAi+(1θ)AiIiAiIi)+g(Vj)g(Vj)VjVj1lnVjVj+g(Vj)g(Vj)θIiAi+(1θ)AiIiAiIi]=nj=1bijSip(Ej)[(p(Ej)p(Ej)1)(1p(Ej)p(Ej)θIjAj+(1θ)AjIjAjIj)+ln(θIiAi+(1θ)AiIiAiIi)ln(θIjAj+(1θ)AjIjAjIj)θIiAi+(1θ)AiIiAiIi+θIjAj+(1θ)AjIjAjIj]+nj=1cijSig(Bj)[(g(Vj)g(Vj)1)(1g(Vj)Vjg(Vj)Vj)+ln(θIiAi+(1θ)AiIiAiIi)lnVjVjθIiAi+(1θ)AiIiAiIi+VjVj]nj=1bijSip(Ej)[ln(θIiAi+(1θ)AiIiAiIi)ln(θIjAj+(1θ)AjIjAjIj)θIiAi+(1θ)AiIiAiIi+θIjAj+(1θ)AjIjAjIj]+nj=1cijSig(Vj)[ln(θIiAi+(1θ)AiIiAiIi)lnVjVjθIiAi+(1θ)AiIiAiIi+VjVj]=:2nj=1aijGij,aij={bijSip(Ej),1jncijSig(Vj)n+1j2n,

    With corresponding functions Gij within both sets of brackets above.Similarly, for i=1,,n, defining:

    Qn+i=ViViVilnViVi,

    we get:

    Qn+i=hi[1ViVi+hihihihiViVi]hi[ln(hihi(θAiAi+(1θ)IiIi))ln(θAiAi+(1θ)IiIi)+lnViViViVi+hihi]hi[hi(θIiAi+(1θ)AiIi)hiAiIi1lnθIiAi+(1θ)AiIiAiIi+lnViViViVi+hihi]=hi[(hihi1)(1hi(θIiAi+(1θ)AiIi)hiAiIi)+θIiAi+(1θ)AiIiAiIilnθIiAi+(1θ)AiIiAiIiViVi+lnViVi]hi[θIiAi+(1θ)AiIiAiIilnθIiAi+(1θ)AiIiAiIiViVi+lnViVi]=:an+i,iGn+i,i.

    With the coefficients aij>0 found above, we can construct the digraph Γ(A) shown in Figure 1 so that along any cycle C, we get (s,r)S(C)Grs=0. More specifically, Gij+Gji=Gn+i,i+Gi,n+i=Gij+Gn+i,i+Gj,n+i=0.

    Then, we can let

    Q:=ni=1ciQi+ni=1cn+iQn+i,

    for some constants ci0,i=1,,2n. Also observe that the in-degree of node n+i satisfies d(n+i)=1, so that from (8) we get cn+i=nj=1cjaj,n+i/an+i,i. On the other hand, note that Gi,n+j+Gn+j,j=Gij. Then, following the same approach used in [20], we get

    Qni=1cinj=1aijGij+ni=1cinj=1ai,n+jGi,n+j+ni=1nj=1cjaj,n+iGn+i,i=ni=1nj=1ci(aij+ai,n+j)[ln(θIiAi+(1θ)AiIiAiIi)ln(θIjAj+(1θ)AjIjAjIj)θIiAi+(1θ)AiIiAiIi+θIjAj+(1θ)AjIjAjIj].

    One can let ˜aij=aij+ai,n+j, and consider the corresponding digraph Γ(˜A). Then, with the new cofactors ˜ci and ˜cn+i=nj=1˜cjaj,n+ian+i,i,i=1,,n, a Lyapunov function for (1) is given by ˜Q=ni=1˜ciQi+ni=1˜cn+iQn+i. Then, applying (8), one can verify that

    ˜Qni=1nj=1˜ci˜aij[ln(θIiAi+(1θ)AiIiAiIi)ln(θIjAj+(1θ)AjIjAjIj)θIiAi+(1θ)AiIiAiIi+θIjAj+(1θ)AjIjAjIj]=0.

    In addition, one can show that the endemic equilibrium (Si,Ai,Ii,Di,Vi) is the largest (and only) invariant set for which ˜Q=0. Then, using LaSalle's invariance principle we conclude that such equilibrium is globally asymptotically stable in\, int(Γ), and thus unique.

    Figure 1.  The digraph Γ(A).
    Figure 2.  Simulation for n=2. Populations in one community of 200K people..

    A general theoretical model of COVID-19 dynamics has been proposed. The model considers the connection and interaction between people from n different communities. The disease can spread through direct (person-to-person) or indirect (humans exposed to contaminated environment) transmission. While most infection transmission might happen within a given community, as people move from their original community i to community j, they can either get infected through contact with (asymptomatic or symptomatic) infected people or contaminated surfaces in node j, or they can pass the disease to susceptible people and shed pathogens in the environment in node j, and go back to their node i. The possibility of people getting infected when handling disease-induced dead bodies, and the dynamics of pathogen concentration are also included in the model. A simple simulation of a 2-node system is shown above, assuming 200K and 400K people in the communities respectively.

    For the particular case of the U. S., after about five months of the first case reported, the number of asymptomatic people seems to be much larger of that of symptomatic, and they seem to be the group most responsible for recent spikes in the number of infections and spreading of the disease. This is probably explained in part because of the inconsistent protocols about reopening the country, with strategies varying very much from state to state and from county to county, ignoring the role that connectivity and human mobility across counties and states play in the persistence of the disease.

    Using matrix and graph-theoretic techniques from [16, 20], it was shown that when the basic reproduction number R0 satisfies R01, the disease-free equilibrium point is globally stable, whereas for R0>1 it loses stability and an endemic equilibrium exists; it was then proved that such endemic equilibrium is unique and globally stable in the interior of a feasible region Γ.

    Model (1) includes the following cases: If all αi=0, then only indirect transmission is possible; if all βi=0, then only direct transmission is possible; if θ=τ=λ=0 and all ϕi=0, then asymptomatic and symptomatic individuals are merged into simply infected individuals; if mS=mI=0, then transmission is only possible within a given patch; finally, if n=1, then the model reduces to one isolated community.

    This work builds on previous research on modeling and analyzing mathematical models of COVID-19 and other disease epidemics, including those in [2, 5, 9, 8, 10, 13, 15, 16, 21, 26]. Some extensions and generalizations of this model are possible for future research, such as splitting infectious people into tested (or detected) and non-tested (not detected), isolated and not isolated, adding the exposed stage into the dynamics, and possibly also splitting symptomatic into those with mild and those with severe symptoms.

    I would like to thank the anonymous referees for their valuable feedback and suggestions, which led to the improvement of this article.



    [1] A. Berman and R. J. Plemmons, Nonnegative Matrices in the Mathematical Sciences, Classics in Applied Mathematics, 9, SIAM, Philadelphia, PA, 1994. doi: 10.1137/1.9781611971262
    [2] Global stability analysis of a general model of Zika virus. Nonauton. Dyn. Syst. (2019) 6: 18-34.
    [3] M. Calmon, Considerations of coronavirus (COVID-19) impact and the management of the dead in Brazil, Forensic Science Internat: Reports, 2020. In press. doi: 10.1016/j.fsir.2020.100110
    [4] Center for Disease Control and Prevention (CDC), Coronavirus Disease (COVID-19), 2020. Available from: https://www.cdc.gov/coronavirus/2019-ncov/index.html.
    [5] Dynamics of a networked connectivity model of epidemics. Discrete Contin. Dyn. Syst. Ser. B (2016) 21: 3379-3390.
    [6] R. Cui, Asymptotic profiles of the endemic equilibrium of a reaction-diffusion-advection SIS epidemic model with saturated incidence rate, Discrete Contin. Dyn. Syst. Ser. B, (2020). doi: 10.3934/dcdsb.2020217
    [7] On the definition and the computation of the basic reproduction number ratio R0 in models for infectious diseases in heterogeneous populations. J. Math. Biol. (1990) 28: 365-382.
    [8] Spread and dynamics of the COVID-19 epidemic in Italy: Effects of emergency containment measures. Proceed. Nat. Acad. of Scienc. (2020) 117: 10484-10491.
    [9] Generalized reproduction numbers and the prediction of patterns in waterborne disease. Proceed. Nat. Acad. of Scienc. (2012) 109: 19703-19708.
    [10] Modelling the COVID-19 epidemic and implementation of population-wide interventions in Italy. Nature Medicine (2020) 26: 855-860.
    [11] J. K. Hale, Ordinary Differential Equations, Second edition. Robert E. Krieger Publishing Co., Inc., Huntington, N.Y., 1980.
    [12] On the dynamics of a class of multi-group models for vector-borne diseases. J. Math. Anal. Appl. (2016) 441: 723-743.
    [13] B. Ivorra, M. R. Ferrández, M. Vela-Pérez and A. M. Ramos, Mathematical modeling of the spread of the coronavirus disease 2019 (COVID-19) taking into account the undetected infections. The case of China, Commun. Nonlinear Sci. Numer. Simul., 88 (2020), 105303, 21 pp. doi: 10.1016/j.cnsns.2020.105303
    [14] Global dynamics of a SEIR model with varying total population size. Math. Biosci. (1999) 160: 191-213.
    [15] Global stability of an epidemic model in a patchy environment. Can. Appl. Math. Q. (2009) 17: 175-187.
    [16] Global stability problems for coupled systems of differential equation on networks. J. Differential Equations (2010) 248: 1-20.
    [17] S. Li and S. Guo, Permanence and extinction of a stochastic SIS epidemic model with three independent Brownian motions, Discrete Contin. Dyn. Syst. Ser. B, (2020). doi: 10.3934/dcdsb.2020201
    [18] M. Mandal, S. Jana, S. K. Nandi, A. Khatua, S. Adak and T. K. Kar, A model based study on the dynamics of COVID-19: Prediction and control, Chaos Solitons Fractals, 136 (2020), 109889, 12 pp. doi: 10.1016/j.chaos.2020.109889
    [19] Asymptotic behaviour of the solutions to a virus dynamics model with diffusion. Discrete Contin. Dyn. Syst. Ser. B (2018) 23: 525-541.
    [20] Global stability of infectious disease models using Lyapunov functions. SIAM J. Appl. Math. (2013) 73: 1513-1532.
    [21] Modelling and control of cholera on networks with a common water source. J. Biol. Dyn. (2015) 9: 90-103.
    [22] (1995) The Theory of the Chemostat: Dynamics of Microbial Competition. Cambridge: Cambridge University Press.
    [23] J. P. Tian, S. Liao and J. Wang, Analyzing the infection dynamics and control strategies of cholera, Discrete Contin. Dyn. Syst., (2013), 747–757. doi: 10.3934/proc.2013.2013.747
    [24] Reproduction numbers and sub-threshold endemic equilibria for compartmental models of disease transmission. Math. Biosci. (2002) 180: 29-48.
    [25] Further notes on the basic reproduction number. Lecture Notes in Mathematics (2008) 1945: 159-178.
    [26] A mathematical model for the novel coronavirus epidemic in Wuhan, China. Math. Biosci. Eng. (2020) 17: 2708-2724.
    [27] Identifying airborne transmission as the dominant route for the spread of COVID-19. Proc. of National Acad. of Sciences (2020) 117: 14857-14863.
  • This article has been cited by:

    1. Josef Botlík, T. Kliestik, Covid 19 as a factor damping globalization trends - structural analysis of local extrema at the level of EU countries, 2021, 129, 2261-2424, 01004, 10.1051/shsconf/202112901004
    2. Comfort Ohajunwa, Carmen Caiseda, Padmanabhan Seshaiyer, Computational modeling, analysis and simulation for lockdown dynamics of COVID-19 and domestic violence, 2022, 30, 2688-1594, 2446, 10.3934/era.2022125
  • Reader Comments
  • © 2021 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(2957) PDF downloads(472) Cited by(2)

Figures and Tables

Figures(2)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog