Abstract
The efficacy of vaccination, incomplete treatment and disease relapse are critical challenges that must be faced to prevent and control the spread of infectious diseases. Age heterogeneity is also a crucial factor for this study. In this paper, we investigate a new age-structured SVEIR epidemic model with the nonlinear incidence rate, waning immunity, incomplete treatment and relapse. Next, the asymptotic smoothness, the uniform persistence and the existence of interior global attractor of the solution semi-flow generated by the system are given. We define the basic reproduction number \({\mathcal {R}}_0\) and prove the existence of the equilibria of the model. And we study the global asymptotic stability of the equilibria. Then the parameters of the model are estimated using tuberculosis data in China. The sensitivity analysis of \({\mathcal {R}}_0\) is derived by the Partial Rank Correlation Coefficient method. These main theoretical results are applied to analyze and predict the trend of tuberculosis prevalence in China. Finally, the optimal control problem of the model is discussed. We choose to take strengthening treatment and controlling relapse as the control parameters. The necessary condition for optimal control is established.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Since the pioneering work carried out by Kermack and McKendrick (1927, 1932, 1933), many mathematical models were put forward to study different infectious diseases. At the same time, age heterogeneity is an important factor that cannot be ignored in the mathematical modeling of infectious diseases. Age structure usually includes two categories: physiological age and infection age. The introduction of the age structure allows for the mathematical description of the disease progression process by partial differential equations(PDEs), which has given rise to the PDEs model or forms of models coupled with ordinary differential equations(ODEs) and PDEs. In 1974, Hoppensteadt developed and studied for the first time an age-dependent epidemic model with physiological and infection age (Hoppensteadt 1974). Because of the complexity of these issues, many scholars have begun modeling and studying physiological age-structured models (Ai and Wang 2024; Dai and Zhang 2023; Huang et al. 2022; Kumar and Abbas 2022; Tian and Wang 2020; Yu et al. 2023) and infection age-structure models (Duan et al. 2014; Guo et al. 2021, 2023; Wang et al. 2017; Li et al. 2017; Liu and Liu 2017; Sun et al. 2021; Wang et al. 2016; Rodrigue Yves et al. 2018), respectively.
In particular, Duan et al. (2014) considered the vaccination period and latent period as two important factors affecting disease dynamics, for this purpose established an SVEIR epidemic model with ages of vaccination and latency. They derived the basic reproduction number and proved the global stability of the disease-free equilibrium and the endemic equilibrium. Wang et al. (2017) considered the age at relapse, the vaccinated individuals can be infected and incomplete vaccination of newborns based on the literature (Duan et al. 2014). Li et al. (2017) built on the literature (Duan et al. 2014) by considering the age at relapse, the disease-induced death rate of infected individuals and the nonlinear incidence rate, promoting the linear and saturation incidence rates already available. Liu and Liu (2017) studied the global stability of an SVEIR epidemic model with age-dependent waning immunity, latency and relapse. Sharp threshold properties for global asymptotic stability of equilibria are given. Wang et al. (2016) investigated the global dynamics of an age-structured SVEIR epidemic model with media influence, incomplete vaccination of newborns and the nonlinear incidence, the analysis showed that the introduction of media-induced nonlinear incidence and age-dependent vaccination and latency had no fundamental effect on the qualitative behavior of the model. Rodrigue Yves et al. (2018) proposed an SVEIR model with age-dependent vaccination, infection and latency, considered a more significant incidence rate, where transmission by both age-mates infective individuals and infective individuals of any age. Sun et al. (2021) supported the results of the literature (Li et al. 2017) by taking into account the age of infection. Recently, Guo et al. (2021, 2023) studied two age-structured tuberculosis(TB) models, and proved the global asymptotic stability of TB-free equilibrium and the endemic equilibrium, respectively.
In addition, optimal control theory is a method of seeking control strategies that make the objective function extremely small or big, it opens up new ideas for the study of infectious diseases and the formulation of control strategies. Moreover, the study of optimal control is an important part of infectious disease modeling, which is significant in guiding public health work and disease prevention and control, can effectively control the spread of disease and reduce control costs. In 1967, Revelle et al. (1967) applied optimal control theory for the first time in the process of controlling diseases to find the optimal control strategy. The purpose is to minimize the prevalence rate of infectious diseases and the cost of control. After that, many scholars have further studied the optimal control model for multiple types of infectious diseases by using control strategies such as vaccination, treatment, isolation, etc., see (Guo et al. 2022; Li et al. 2020; Mohammed Awel et al. 2021; Wang and Nie 2023; Yang et al. 2019, 2023).
Recently, Khan and Zaman (2018) studied optimal control strategy for the SEIR epidemic model with continuous age-structured for exposed and infected classes. Using vaccination as the controlling variable, optimal control theory was applied to minimize susceptible, exposed and infected individuals to reduce the rate of infection. The objective function was established and the form of the optimal control was derived using the adjoint and sensitivity systems. For the numerical simulation of the optimal system, the forward-backward sweep method was adopted to numerically solve the state and adjoint systems. Duan et al. (2020) discussed the optimal control problem by considering enhanced vaccination and improved treatment as control strategies, the results showed that both the age of vaccination and recovery age appeared in the expressions of the optimal control variables, affecting the efficiency of the optimal control. Guo et al. (2024) developed an age-structured TB model of vaccination, treatment and relapse, and investigated the stability of model equilibria and the optimal control issue. The findings guide how to effectively utilize limited resources to reduce the spread of TB.
Although the above literature has explored the dynamic behavior and optimal control problem of epidemic models. However, to varying degrees, the impact of some factors on disease transmission has been overlooked. There are few papers that have considered all these factors together to study an epidemic model. To this end, we develop a new age-structured SVEIR model coupled with PDEs and ODEs, focusing on the following four main aspects: (1) the effects of the nonlinear incidence rate, waning immunity, incomplete treatment of disease, and relapse on the global stability of the model; (2) the global stability of the disease-free equilibrium and the endemic equilibrium. It is worth noting that the model is built with multiple return terms, and the derivative of the Volterra-type Lyapunov function concerning time t still contains various Volterra-type functions. So, some construction and computational skills are required in the proof; (3) parameter fitting using real data and (4) the problem of the existence of optimal control. It is a challenging task for us to prove the properties of the model and establish the Lyapunov function.
The basic structure of this paper is as follows: the next section investigates a new age-structured SVEIR epidemic model with the nonlinear incidence rate, waning immunity, incomplete treatment and relapse. In Sect. 3, we give the properties of the model and the basic reproduction number \({\mathcal {R}}_0\). Section 4 contains our main results on the asymptotic smoothness, the uniform persistence, the existence of a global attractor and the global stability of equilibria. In Sect. 5, the parameters were fitted using actual data to obtain the initial values of the model parameters. Furthermore, we conduct sensitivity analysis in this section. In Sect. 6, we formulate an optimal control problem, involving two types of control measures: enhancing treatment and controlling relapse. And applying the Gâteaux derivatives to obtain the necessary conditions for optimal control. A brief discussion is given in Sect. 7.
2 Model formulation
We divide the total population N(t) at the time t into five classes: the susceptible S(t), vaccinated V(t), latent E(t), infectious I(t) and recovered R(t). Let S(t) and I(t) represent the density of the susceptible individuals and infectious individuals at time t, \(v(\theta ,t)\) stands for the density of the vaccinated individuals at time t with vaccination age \(\theta \), e(a, t) represents the density of latent individuals with latency age a at time t, and r(w, t) be the density of the recovered individuals at time t with relapse age w.
We assume positive constants \(\Lambda \), \(\alpha \) and \(\mu \) to be the recruitment rate, the vaccination rate of the susceptible individuals, and the natural death rate of the population, respectively. \(\delta _1\) and \(\delta _2\) represent the disease-related mortality rate of latent and infectious individuals, \(\sigma _1\) and \(\sigma _2\) denote the treatment rate and incomplete treatment rate of the disease. Denote the infectious force function by \(\beta S(t)f(I(t))\), where \(\beta \) is the probability of infection. The vaccine-induced immunity wanes at the rate \(h(\theta )\) and the total number of vaccinated individuals at time t is \(V(t)=\int _0^\infty v(\theta ,t)\textrm{d}\theta \), thus the total number of waning immunity individuals who progress into the susceptible class alive reads \(\int _0^\infty h(\theta )v(\theta ,t)\textrm{d}\theta \).
Similarly, the vaccine efficacy is given by the function \(\sigma (\theta )\), thus the total number of vaccinated individuals who progress into the latent class alive reads \(\int _0^\infty \beta f(I(t))\sigma (\theta ) v(\theta ,t) \textrm{d}\theta \), where \(\sigma (\theta )\in [0,1]\); The removal rate from latent class is given by the function k(a), thus the total number of latent individuals who progress into the infectious class alive reads \(\int _0^\infty k(a)e(a,t)\textrm{d}a\); The relapse rate in recovered class is given by the function \(\gamma (w)\), thus the total number of recovered individuals who progress into the infectious class alive reads \(\int _0^\infty \gamma (w)r(w,t)\textrm{d}w\). Our model is described by the following diagram in Fig. 1:
From Fig. 1, we can establish the following coupled system of ODEs and PDEs:
with the initial conditions
and the boundary conditions
here \(S_0\), \(I_0 \in {\mathbb {R}}_ +\), and \(v_0(\theta )\), \(e_0(a)\), \( r_0(w) \in L_+^1\), where \(L_+^1:(0,\infty )\rightarrow {\mathbb {R}}_+\) is the space of Lebesgue integrable functions. All the parameters are positive. We make the following assumptions about the parameters of the system (1):
(A1) \(\theta \in [0,\Theta ]\), \(a \in [0,A]\) and \(w \in [0,W]\), where \(\Theta \), A and W are the maximum ages of vaccination, latency and relapse, respectively. If \(\Theta =\infty \), \(A=\infty \) and \(W=\infty \), suppose that \(v(\theta ,t)=0\), \(e(a,t)=0\) and \(r(w,t)=0\) for all sufficiently large \(\theta \), a and w, respectively.
(A2) \(h(\theta )\), \(\sigma (\theta )\), k(a), \(\gamma (w) \in L_+^1\) are bounded with the essential upper bounds \({\bar{h}}\), \( {\bar{\sigma }}\), \({\bar{k}}\), \({\bar{\gamma }}\) for \(\theta \ge 0\), \(a\ge 0\), \(w\ge 0\), and Lipschitz continuous with Lipschitz coefficients \(M_h\), \(M_\sigma \), \(M_k\), \(M_\gamma \), respectively.
(A3) There exists a positive constant \(\mu _0\in [0,\mu ]\), such that \(h(\theta )\), \(\sigma (\theta )\), k(a), \(\gamma (w)\ge \mu _0\) for \(\theta \ge 0\), \(a\ge 0\), \(w\ge 0\), respectively.
(A4) Function f(I) is nonnegative and twice differentiable for all \(I\ge 0\), \(f(I)\ge 0\) with equality if and only if \(I=0\), and \(f'(I)\ge 0\), \(f''(I)\le 0\).
Remark 1
The f(I) satisfying assumption (A4), then there are \(f'(I)I \le f(I)\le f'(0)I\), for \(I\ge 0\).
Remark 2
The f(I) satisfying assumption (A4) can be expressed by the following functions: \(f(I)=I\), \(f(I)=\frac{I}{1+\alpha I}\), where \(\alpha >0\), and \(f(I)=\Big (\beta -\frac{\beta _1I}{m+I}\Big )I\), where \(\beta >\beta _1\), etc.
3 Preliminaries
In this section, some preliminary results are presented to establish our main conclusions.
3.1 Notations
In order to simplify the later derivation, we use the following notation:
3.2 Volterra formulation
By application of Volterra formulation (Webb 1985), solving \(v(\theta ,t)\), e(a, t) and r(w, t) in system (1) along the characteristic lines \(t-\theta \) = constant, \(t-a\) = constant and \(t-w\) = constant, respectively, we get
3.3 State space and well-posedness
The state space X for system (1) is defined by
equipped with the norm
The initial conditions (2) of the system (1) can be rewritten as \(x_0=\big (S_0,v_0(\cdot ),e_0(\cdot ),I_0,r_0(\cdot )\big )\in {\textbf {X}}\). Then, from the standard theory of functional differential equations (Hale 1971), it follows that when the initial condition is satisfied, for all \(t\ge 0\), there exists a unique nonnegative solution to the system (1). Thus, we can obtain a continuous solution semi-flow \(\Phi :{\mathbb {R}}_+ \times {\textbf {X}}\rightarrow {\textbf {X}}\) associated with the system (1),
with the norm
Set
We have the following lemma.
Lemma 1
For system (1), we have
-
(i)
\(\Omega \) is a positively invariant for \(\Phi \). i.e., \(\Phi (t,x_0) \subset \Omega \), \(\forall t\ge 0\), \(x_0 \in \Omega \);
-
(ii)
\(\Phi \) is point dissipative and \(\Omega \) attracts each point in \({\textbf{X}}\);
-
(iii)
There exist constants \(\varsigma _1\), \(\varsigma _2 > 0\), such that \(\lim \limits _{t \rightarrow \infty } \mathop {\inf } S(t)\ge \varsigma _1\), \(\lim \limits _{t \rightarrow \infty }\mathop {\inf } \int _0^\infty v(\theta ,t)\textrm{d}\theta \ge \varsigma _2.\)
Proof
From (6), we have
From the second equation in the system (1), it follows that
So
Similarly
Combining (7-9) and the first and fourth equations of the system (1) yields
Further, we get
Thus \(x_0 \in {\textbf {X}}\), there are all the solutions to the system (1) have \(\Phi (t,x_0) \subset \Omega \). So, \(\Omega \) is positively invariant for \(\Phi \), which leads to \(\mathop {\lim }\limits _{t\rightarrow \infty }\sup \left\| {{\Phi }({t,x_0})} \right\| _{\textbf {X}} \le \frac{\Lambda }{\mu }\) for any \(x_0 \in {\textbf {X}}\). Therefore, \(\Phi \) is point dissipative and \(\Omega \) attracts all points in X, i.e., Lemma 1 (i) and (ii) hold.
Now prove (iii) in Lemma 1. Since \(I(t)\le \frac{\Lambda }{\mu }\) and \(f(I)\le f'(0)I\), it is easy to see from the first equation of the system (1) that
Then we have that
According to assumption (A2), using the second equation of the system (1) yields
Here we use the result that \(v(\theta ,t)|_{\theta =\infty }=0\) from assumption (A1) gives
This completes the proof. \(\square \)
The following proposition directly follows from Lemma 1.
Proposition 1
If \( x_0 \in {\textbf{X}}\) and \(\Vert x_0\Vert _{\textbf{X}} \le M\) for some constant \(M \ge \frac{\Lambda }{\mu }\), then the following statements hold true for \(t \ge 0\):
-
(i)
\( 0\le S(t)\), \(\int _0^\infty v(\theta ,t)\textrm{d}\theta \), \( \int _0^\infty e(a,t)\textrm{d}a\), I(t), \(\int _0^\infty r(w,t)\textrm{d}w\le M;\)
-
(ii)
\(v(0,t)\le \alpha M\), \(e(0,t)\le \beta f'(0)M^2+\int _0^\infty \beta f'(0) M^2 \sigma (\theta )\textrm{d}\theta +\sigma _2M\), \(r(0,t)\le \sigma _1 M.\)
3.4 Equilibria and the basic reproduction number
In this subsection, we consider the equilibria and the explicit expression of the basic reproduction number of the system (1).
Theorem 1
System (1) always exists a disease-free equilibrium \(E^0=\big (S^0,v^0(\theta ),e^0(a),I^0,r^0(w)\big )\), and has a unique endemic equilibrium \(E^*=\big (S^*,v^*(\theta ),e^*(a),I^*,r^*(w)\big )\) if and only if \({\mathcal {R}}_0>1\).
Proof
Obviously, the system (1) always exists a disease-free equilibrium \(E^0=\big (S^0,v^0(\theta ),e^0(a),I^0,r^0(w)\big )\) satisfies
To find the endemic equilibrium of the system (1), we first define the basic reproduction number (Diekmann et al. 1990) as
where \(\varpi =\mu +\delta _2+\sigma _1+\sigma _2\). \(\frac{1}{\varpi }\) is the average time of an infectious individual in infectious class on the first pass, \(\theta _2\) and \(\theta _1\) are the probability of the infectious individuals surviving from latent class and relapse individuals from the recovered class. Thus the total average time in the infectious class is calculated by
Hence, \(\frac{\beta f'(0) \theta _2 S^0}{\varpi -\sigma _1 \theta _1-\sigma _2 \theta _2}\) and \(\frac{\beta f'(0)\theta _2 \int _0^\infty {\sigma (\theta )v^0(\theta )\textrm{d}\theta }}{\varpi -\sigma _1 \theta _1-\sigma _2 \theta _2}\) represent the number of susceptible and vaccination individuals that can be infected by an infectious individual, respectively. \({\mathcal {R}}_0\) represents the average number of new infections produced by an infectious individual over the entire infectious period, it is a crucial threshold parameter for the global stability of equilibria.
Note that any endemic equilibrium \(\big (S^*,v^*(\theta ),e^*(a),I^*,r^*(w)\big )\) of system (1) satisfies
where \(n^*(\theta )=h(\theta )+\mu +\beta \sigma (\theta )f(I^*)\). From the second, third and fifth equations of the system (11), we obtain
where \(\eta ^*(\theta )={e^{-\int _0^\theta {n^*(r)\text {d}r}}}\). Substituting \(v^*(\theta )\) into the first equation of the system (11), yields
Substituting \(v^*(\theta )\) and \(S^*\) into the fourth equation of the system (11) gives
Denoting the right-hand side of (12) as \(\kappa (I)|_{I=I^*}\), the existence of the endemic equilibrium \(E^*\) depends on the existence of \(I^*\), there is
Derivation of the above equation gives
By assumption (A4) and Remark 1, there are \(f'(I)\ge 0\) and \(f(I)-f'(I)I\ge 0\), then \(\kappa '(I)\le 0\), which means that \(\kappa (I)\) monotonically decreases in \(\left( 0,\frac{\Lambda }{\mu }\right) \). Since
If \({\mathcal {R}}_0 \le 1\), then \(\kappa (0) \le 0\), so \(\kappa (I)=0\) has no positive real root. If \({\mathcal {R}}_0 >1\), then \(\kappa (0) >0\), so \(\kappa (I)=0\) has a unique positive real root \(I^*\). Thus, the system (1) has a unique endemic equilibrium \(E^*=\big (S^*, v^*(\theta ), e^*(a), I^*, r^*(w)\big )\) if and only if \({\mathcal {R}}_0>1\). This completes the proof. \(\square \)
4 Main results
In this part, we give some results on the asymptotic smoothness, the uniform persistence and the existence of a global attractor for the semi-flow generated by the system (1). Moreover, the stability of equilibria is captured.
4.1 Asymptotic smoothness of the semi-flow
Asymptotic smoothness of the semi-flow is vital to apply the invariance principle. The following lemma is useful for proofing asymptotic smoothness.
Lemma 2
(Hale and Waltman 1989) If the following two conditions hold, then the semi-flow \(\Phi (t,x_0)=\phi (t,x_0)+\varphi (t,x_0):{\mathbb {R}}_+\times {\textbf{X}}\rightarrow {\textbf{X}}\) is asymptotically smooth in \({\textbf{X}}\).
-
(i)
There exists a continuous function \(u: {\mathbb {R}}_+\times {\mathbb {R}}_+\rightarrow {\mathbb {R}}_+\) such that \(\mathop {\lim }\limits _{t\rightarrow \infty }u(t,h)= 0\) and \({\left\| {\phi (t,{x_0})} \right\| _{\textbf{X}}} \le u(t,h)\) if \({\left\| {{x_0}} \right\| _{\textbf{X}}} \le h\);
-
(ii)
For \(t\ge 0,\) \(\varphi (t,x_0)\) is completely continuous.
We can now prove one of the main results.
Theorem 2
The semi-flow \(\Phi (t,x_0)\) is asymptotically smooth.
Following Lemma 2, we will divide into two steps to verify Theorem 2.
Step 1: We need to decompose the semi-flow as \(\Phi (t,x_0)=\phi (t,x_0)+\varphi (t,x_0)\), where \(\phi (t,x_0)=\big (0, v(\cdot ,t), e(\cdot ,t), 0, r(\cdot ,t)\big )\) and \(\varphi (t,x_0)=\big (S(t), {\tilde{v}}(\cdot ,t), {\tilde{e}}(\cdot ,t), I(t), {\tilde{r}}(\cdot ,t)\big ),\) here
We need to prove the following proposition to verify that condition (i) of Lemma 2 holds.
Proposition 2
For \(h>0\), let \(u(t,h) = h{e^{ - ({\mu _0}+\mu )t}}\). Then \(\mathop {\lim }\limits _{t \rightarrow \infty } u(t,h) = 0\) and \( {\left\| {\phi (t,{x_0})} \right\| _{\textbf{X}}} \le u(t,h) \) if \( {\left\| {{x_0}} \right\| _{\textbf{X}}} \le h.\)
Proof
Let \(u(t,h)=he^{-(\mu _0+\mu )t}\). Clearly, for any \(h>0\), \(\mathop {\lim }\limits _{t\rightarrow \infty }u(t,h)=0\). For \(x_0 \in \Omega \) and \(\Vert x_0\Vert _{\textbf {X}} \le h\), it follows from (5) and (6) that
Since \(n(\theta ,t)\ge \mu _0+\mu \), \(\varepsilon (a)\ge \mu _0+\mu \) and \(u(w)\ge \mu _0+\mu \) hold for all \(\theta \), a, \(w\ge 0\), we have
This completes the proof. \(\square \)
Step 2: In order to verify (ii) of Lemma 2, we discuss the completely continuous of \(\varphi (t,{x_0})\). Proposition 1 states that S(t) and I(t) remain in the compact set [0, M], where \(M \ge \frac{\Lambda }{\mu }\). We need a notion of compactness in \(L_ + ^1\) since it is a component of infinite dimensional Banach space \({\textbf{X}}\). To ensure this, we need to confirm that \({\tilde{v}}(\theta ,t)\), \({\tilde{e}}(a,t)\) and \({\tilde{r}}(w,t)\) belong to a precompact subset of \(L_+^1\) that is independent of \(x_0\in \Omega \). So we will need the following result.
Lemma 3
(Adams and Fournier 2003) Let \(K \subset {L^p}(0,\infty )\) be a closed and bounded set where \(p\ge 1\). Then K is compact if the following conditions hold:
-
(i)
\(\mathop {\lim }\limits _{h \rightarrow \infty } \int _h^\infty {{{\left| {f(z)} \right| }^p}\textrm{d}z} = 0\) uniformly for \(f\in K\);
-
(ii)
\(\mathop {\lim }\limits _{h \rightarrow 0 } \int _0^\infty {{{\left| {f(z + h) - f(z)} \right| }^p}\textrm{d}z} = 0\) uniformly for \(f\in K \).
Proposition 3
For \(t\ge 0\), \( \varphi (t,{x_0}) \) is completely continuous.
Proof
It follows from Lemma 3 that for any closed and bounded set \(B\subset {\textbf{X}}\), we have \(\varphi (t,B)\) is compact. According to Proposition 1, S(t) and I(t) remain in the compact set [0, M], where \(M\ge \frac{\Lambda }{\mu }\) is a bound for B.
We have from (5) and (13) that
As \(0\le \eta (\theta ,t)\le e^{-(\mu _0+\mu ) \theta }\). Using (ii) of Proposition 1, we obtain that
Thus (i) in Lemma 3 is satisfied.
For sufficiently small \(h \in (0,t)\), yields
where
Recall that \(0 \le \eta (\theta ,t) \le 1\) and \(\eta (\theta ,t)\) is a non-increasing function with respect to \(\theta \). Then it follows that
From Proposition 1, we have
Combining the system (1), assumption (A2), Remark 1 and Proposition 1, it follows that
Which imply \(\left| \frac{\textrm{d}S(t)}{\textrm{d}t} \right| \) is bounded and S(t) is Lipschitz continuous on \( [0,\infty )\) with Lipschitz coefficient \( {M_S}\); Similarly, \(\left| \frac{\textrm{d}I(t)}{\textrm{d}t} \right| \) is bounded and I(t) is Lipschitz continuous on \( [0,\infty )\) with Lipschitz coefficient \( {M_I}\). We deduce
Combining the above processes leads to
So \(\mathop {\lim }\limits _{h\rightarrow 0}\int _0^\infty {\left| {{\tilde{v}}(\theta + h,t) - {\tilde{v}}(\theta ,t)} \right| \textrm{d}\theta }=0\) uniformly for any \(x_0\in B\), thus, \({\tilde{v}}(\theta ,t)\) remains in a precompact subset \({B_{{{\tilde{v}}}}}\) of \( L_+^1\). Similarly, \({\tilde{e}}(a,t)\) and \({\tilde{r}}(w,t)\) also remain in precompact subsets \(B_{{\tilde{e}}}\) and \(B_{{\tilde{r}}}\) of \(L_+^1\), respectively. Thus \(\varphi (t,B)\subseteq [0,M]\times B_{{\tilde{v}}}\times {B_{{{\tilde{e}}}}}\times [0,M]\times {B_{{{\tilde{r}}}}}\), which is compact in \({\textbf {X}}\). Thus, \(\varphi (t,{x_0})\) is completely continuous. This finishes the verification of (ii) of Lemma 2.
Now, Theorem 2 has been proven by combining Lemma 2, Proposition 2 and 3. The proof is completed. \(\square \)
4.2 Uniform persistence
This subsection will prove the uniform persistence of the system (1) when \({\mathcal {R}}_0 > 1\), which means that infected individuals are always present if infectivity initially appears.
For convenience, we introduce the Volterra-type function
We can see that g(x) is positive definite and reaches a global minimum at \(x=1\) with \(g(1)=0\).
We are now introduce some notations. Set
Clearly, \({{{\bar{a}}}}\), \({{{\bar{w}}}}>0\) due to k(a), \(\gamma (w)\in L_ + ^1\). We further define
and
Let us first establish the following theorem.
Theorem 3
For the semi-flow \({\left\{ {\Phi (t,x_0)} \right\} _{t \ge 0}}\) generated by system (1), the set \(\partial {\textbf{Y}}\) is positively invariant, that is, \(\Phi (t,\partial {\textbf{Y}}) \subset \partial {\textbf{Y}}\) for any \(t\ge 0\). Furthermore, the disease-free equilibrium \(E^0\) is globally asymptotically stable for the semi-flow \({\left\{ {\Phi (t,x_0)} \right\} _{t \ge 0}}\) restricted to \(\partial {\textbf{Y}}\).
Proof
Suppose \(\big (S_0, v_0(\cdot ), e_0(\cdot ), I_0, r_0(\cdot )\big )\in \partial {\textbf{Y}}\), then there is \(\big (e_0(\cdot ), I_0, r_0(\cdot )\big )\in \partial \tilde{{\textbf{Y}}}\), this shows that \(e_0(\cdot )=0\), \(I_0=0\) and \(r_0(\cdot )=0\). We consider the following system:
Since \(0\le \sigma (\theta ) \le 1,~ S(t)+\int _0^\infty \sigma (\theta )v(\theta ,t)\textrm{d}\theta \le S(t)+\int _0^\infty v(\theta ,t)\textrm{d}\theta \le \frac{\Lambda }{\mu }\) as \(t \rightarrow \infty \). From Remark 1, \(f(I)\le f'(0)I\) for any \(I\ge 0\). The comparison theorem of differential equations implies that for any \(t\ge 0\),
where \(\big ({{\hat{e}}}(a,t), {{\hat{I}}}(t), {{\hat{r}}}(w,t)\big )\) is the solution of the following system:
Similarly to (5), solving \({{\hat{e}}}(a,t)\) and \({{\hat{r}}}(w,t)\) of the system (18), we can obtain
Substituting (19) into the second equation of system (18), yields
The initial condition \({{\hat{I}}}(0)=0\) implies a unique solution \({{\hat{I}}}(t)=0\) for the above equation. By (19), we also get \({{\hat{e}}}(a,t)=0\) and \({{\hat{r}}}(w,t)=0\) for \(0\le a, w\le t\). With \(e(a,t)=0\), \(I(t)=0\) and \(r(w,t)=0\), this holds true for all \(t\ge 0\), this result indicates that the set \(\partial {\textbf{Y}}\) is positively invariant.
The system (1)-(3) in \(\partial {\textbf{Y}}\) becomes
Construct the following Lyapunov function:
Note that \(\mu +\alpha =\frac{1}{S^0}\left( \Lambda +\int _0^\infty h(\theta )v^0(\theta )\textrm{d}\theta \right) \). Calculation the derivative of \(P_S(t)\) along with the solutions of system (20) gives
The derivative of \(P_v(t)\) satisfies
where \(v_\theta (\theta ,t)=\frac{\partial v(\theta ,t)}{\partial \theta }\). Recall that \(\frac{\textrm{d} v^0(\theta )}{\textrm{d}\theta }=-n(\theta )v^0(\theta )\), it follows that
Using integration by parts, \(v^0(0)=\alpha S^0\) and \(v(0,t)=\alpha S(t)\), we have
Substitute (23)-(24) into (22)
Noticing \(n(\theta )=h(\theta )+\mu \) and \(\alpha =\frac{\Lambda }{S^0}-\mu +\frac{1}{S^0}\int _0^\infty h(\theta )v^0(\theta )\textrm{d}\theta \). It follows from (21) and (25) that
By (26), \(\frac{\textrm{d}P(t)}{\textrm{d}t}\le 0\), the equality sign holds if and only if \(S(t)=S^0\) and \(v(\theta ,t)=v^0(\theta )\). The disease-free equilibrium \(E^0\) is globally asymptotically stable in \(\partial {\textbf{Y}}\) according to the LaSalle invariance principle (LaSalle 1976). This completes the proof. \(\square \)
Next, we apply methods from the literature (Hale and Waltman 1989; Magal and Zhao 2005) to prove the uniform persistence of the system (1). The following theorem is introduced.
Theorem 4
Suppose that \({\mathcal {R}}_0>1\). The semi-flow \({\left\{ {\Phi (t,x_0)} \right\} _{t \ge 0}}\) is uniformly persistent in \(({\textbf{Y}},\partial {\textbf{Y}})\), that is, there exists a constant \(\iota >0\) which is independent of the initial values such that \(\mathop {\lim }\limits _{t \rightarrow \infty }\inf {\left\| {\Phi (t,x_0)} \right\| _{\textbf{X}}} \ge \iota \) for any \(x_0\in {\textbf{Y}}\). Furthermore, the semi-flow \({\left\{ {\Phi (t,x_0)} \right\} _{t \ge 0}}\) has a compact global attractor \( {\mathcal {A}}_0\) in \({\textbf{Y}}\).
Proof
Since it has been verified in Theorem 3 that \(E^0\) is globally asymptotically stable in \(\partial {\textbf{Y}}\). From Theorem 4.2 in Hale and Waltman (1989) it only needs to prove that there exist \(T>0\) and \(\iota >0\), such that \(\mathop {\lim }\limits _{t \rightarrow \infty } \inf {\left\| {\Phi (t,{x_0})} \right\| _{\textbf{X}}} \ge \iota \) for any \(t>T\), \(x_0\in {\textbf{Y}}\). It suffices to examine that \({W^s}({E^0}) \cap {\textbf{Y}} = \emptyset \), where \({W^s}({E^0}) = \{ {x_0} \in {\textbf{Y}}:\mathop {\lim }\limits _{t \rightarrow \infty } \Phi (t,{x_0}) = {E^0}\}\).
By the way of contradiction, assuming that there exists a solution \(y\in {\textbf{Y}}\) such that \(\mathop {\lim }\limits _{t \rightarrow \infty } \Phi (t,y) = E^0\). Thus, we can take a sequence \(\{y_n\} \subset {\textbf{Y}}\), where \({y_n} = \big ({S_n}(0), {v_n}(\cdot ,0), {e_n}(\cdot ,0 ), {I_n}(0), {r_n}(\cdot ,0)\big )\), such that \({\left\| {\Phi (t,y_n)-E^0} \right\| _{\textbf {X}}} < \frac{1}{n}\), for \(t \ge 0\). Here, \(\Phi (t,{y_n}) = \big ({S_n}(t), {v_n}(\cdot ,t), {e_n}( \cdot ,t ), {I_n}(t), {r_n}( \cdot ,t )\big )\).
Now, we choose a sufficiently large \(n>0\), ensures that \(S^0-\frac{1}{n}>0\) and \(v^0(\theta )-\frac{1}{n}>0\). For this chosen n, there exists \(T>0\) such that \(t>T\),
From (5), we obtain that
Combining Remark 1 and (27-28) yields
Substituting (29) into the fourth equation in the system (1) and applying the comparison principle of differential equation, we obtain \(I_n(t) \ge \varrho _n(t)\), for all \(t\ge 0\), where \(\varrho _n(t)\) is the solution to the following Volterra-type integral equation:
The next proof \(\varrho _n(t)\) is unbounded. If \(\varrho _n(0)>0\), then when \({\mathcal {R}}_0>1\), choose enough large \(n \in {\mathbb {R}}_+\), so that the following equation holds:
Namely
We can obtain that \(\varrho _n(t)\) is unbounded due to \(I_n(t)\ge \varrho _n(t)\), so \(I_n(t)\) is unbounded, which contradicts the boundedness of \(I_n(t)\) in (27), thus the assumption is not valid, \({W^s}({E^0}) \cap {\textbf{Y}} = \emptyset \) holds. Uniform persistent and existence of a global attractor \({\mathcal {A}}_0 \) for \({\{ \Phi (t,x_0)\} _{t \ge 0}}\) can be derived from Theorem 3.7 in literature (Magal and Zhao 2005). This completes the proof. \(\square \)
4.3 Stability of the disease-free equilibrium \(E^0\)
In this subsection, we give the local and global asymptotic stability of the disease-free equilibrium \(E^0\), respectively.
Theorem 5
The disease-free equilibrium \(E^0\) is locally asymptotically stable if \({\mathcal {R}}_0< 1\), and unstable if \({\mathcal {R}}_0>1\).
Proof
We first take the change of variables as follows:
Linearizing the system (1) at the disease-free equilibrium \(E^0\) leads to the following system:
Let \( {x_1}(t) = x_1^0{e^{\lambda t}}\), \({x_2}(\theta ,t) = x_2^0(\theta ){e^{\lambda t}}\), \({x_3}(a,t) = x_3^0(a){e^{\lambda t}}\), \( {x_4}(t) = x_4^0{e^{\lambda t}}\) and \({x_5}(w,t) = x_5^0(w){e^{\lambda t}}\) be the solution of the system (30), where \(x_1^0\), \(x_2^0(\theta )\), \(x_3^0(a)\), \(x_4^0\), \(x_5^0(w)\) and \(\lambda \) will be determined later. Observe that it suffices to use the third, fourth and fifth equations of the system (30) to determine the characteristic equation of the system (1) at the disease-free equilibrium \(E^0\). Then
Integrating the first equation of (31) from 0 to a, yields
Similarly, solving (33), we get
Substituting (34)-(35) into (32) reduces to the characteristic equation of system (1) at the disease-free equilibrium \(E^0\)
Let \(H(\lambda )\) denote the right-hand side of (36). It’s clear that
If \({\mathcal {R}}_0>1\), then \(H(0)>0\). Since \(H(\lambda )\) is continuously differentiable, there exists at least one positive real root, so the disease-free equilibrium \(E^0\) is unstable. If \({\mathcal {R}}_0<1\), we rewrite (36) into
Let Re\(\lambda \ge 0\), then \(|\lambda +\varpi |\ge \varpi \). Using \(\Xi \) to denote the right-hand side of (37), we have
This contradicts equation (37). Thus, if \({\mathcal {R}}_0<1\), then Re\(\lambda < 0\). Namely, the disease-free equilibrium \(E^0\) is locally asymptotically stable. This proof is complete. \(\square \)
Next, we give the global stability of the disease-free equilibrium \(E^0\).
Theorem 6
The disease-free equilibrium \(E^0\) is globally asymptotically stable if \({\mathcal {R}}_0<1\).
Proof
Set Lyapunov function as follows:
where
By \(\mu =\frac{\Lambda }{S^0}-\alpha +\frac{1}{S^0}\int _0^\infty h(\theta )v^0(\theta )\textrm{d}\theta \), similar to the process of solving (21), the derivative of \(L_1(t)\) along the system (1) is
Similar to (25), we have
The derivative of \(L_3(t)\) satisfies
Obviously
Similarly to (40), we get
Rearranging (38)-(42) reduces to
where
According to Remark 1, there is \(f(I)\le f'(0)I\) holds, which means that
Summarizing, we get
Thus, it follows from (46) that \(\frac{\textrm{d}L(t)}{\textrm{d}t}\le 0\) if \({\mathcal {R}}_0<1\). When \(S(t)=S^0\), \(v(\theta ,t)=v^0(\theta )\), \(e(a,t)=0\), \(I(t)=0\), \(r(w,t)=0\) hold at the same time there is \(\frac{\textrm{d}L(t)}{\textrm{d}t}=0\). Therefore, the single point set \(\{E^0\}\subset \Omega \) is the largest invariant subset of \(\frac{\textrm{d}L(t)}{\textrm{d}t}=0\). By the LaSalle invariance principle, when \({\mathcal {R}}_0<1\), the disease-free equilibrium \(E^0\) is globally asymptotically stable. That proves the Theorem 6. \(\square \)
4.4 Stability of the endemic equilibrium \(E^*\)
In this subsection, we prove the global stability of the endemic equilibrium \(E^*\) by constructing the suitable Lyapunov function.
We have the following lemma based on Remark 1 and Proposition A.1 in Sigdel and McCluskey (2014).
Lemma 4
Assign
Then the assumption (A4) ensure that \(G(I)\le 0\) holds for all \(I >0\).
Now, we give the following theorem.
Theorem 7
The endemic equilibrium \(E^*\) is globally asymptotically stable in \({\textbf{Y}}\) if \({\mathcal {R}}_0>1\).
Proof
Define the following Lyapunov function:
where
By \(\mu =\frac{\Lambda }{S^*}-\alpha +\frac{1}{S^*} \int _0^\infty {h(\theta )v^*(\theta )\textrm{d}\theta }-\beta f(I^*)\), similar to the process of solving (21), the derivative of \(H_1(t)\) along the system (1) is given by
Similar to (25), we have
The same reasoning leads to
By \(\varpi =\frac{1}{I^*}\big (\int _0^\infty {k(a)^*e(a)\textrm{d}a}+\int _0^\infty {\gamma (w)r^*(w)\textrm{d}w}\big )\), analogously to (21), we get
The same reasoning as (25) gives
Using \(g\left( \frac{r^*(0)I(t)}{r(0,t)I^*}\right) =g(1)=0\) and \(g\left( \frac{v^*(0)S(t)}{v(0,t)S^*}\right) =g(1)=0\). The simplification from (47–51) leads to
By Lemma 4, we know that the fifth term of (52) is negative. In summary, \(\frac{\textrm{d}H(t)}{\textrm{d}t}\le 0\). \(\frac{\textrm{d}H(t)}{\textrm{d}t}=0\) if and only if \(S(t)=S^*\), \(v(\theta ,t)=v^*(\theta )\), \(e(a,t)=e^*(a)\), \(I(t)=I^*\) and \(r(w,t)=r^*(w)\) simultaneous. Hence \(\{E^*\} \subset \Omega \) is the largest invariant subset of \(\frac{\textrm{d}H(t)}{\textrm{d}t}=0\). Thus, applying the LaSalle invariance principle, we get the endemic equilibrium \(E^*\) is globally asymptotically stable in \({\textbf{Y}}\) if \({\mathcal {R}}_0>1\). This completes the proof. \(\square \)
5 Parameters estimation and sensitivity analysis
In this section, we estimate the parameters of the system (1) using the number of new TB cases per year in China from 2007 to 2020. In addition, we conduct a sensitivity analysis and predict whether China will be able to achieve the World Health Organization’s (WHO) goal of reducing TB incidence by \(80\%\) in 2030 compared to 2015.
5.1 Parameters estimation
According to data from the National Bureau of Statistics of China(NBSC) (National Bureau of Statistics of China 2023), the average newborn population and the average life expectancy in the 2007–2020 period are 16, 289, 670 persons \(\mathrm { year^{-1}}\) and 76.34 years old, respectively, and the number of the initial susceptible population is 1, 314, 480, 000 persons. Thus, we take \(\Lambda = 16,289,670\), \(\mu = 1/76.34\) and \(S(0)=1,314,480,000\). From the data of the Chinese Center for Disease Control and Prevention (Chinese Center for Disease Control and Prevention 2023), the number of new cases of TB from 2007 to 2020 is shown in Table 1.
In the numerical simulations, we use the year as the unit to ignore the time length within a few weeks, and choose \(f(I) = I\). Next, we estimate the unknown parameters and initial values of the system (1). Set
where we assume \(h(a)=d_1e^{-d_2a},~ \sigma (a)=c_1e^{-c_2a},~ k(a)=k_1e^{-k_2a},~ \gamma (a)=\gamma _1e^{-\gamma _2a},~ v_0(a)=v(0)c_2e^{-c_2a},~ e_0(a)=e(0)k_2e^{-k_2a},~ r_0(a)=r(0)\gamma _2e^{-\gamma _2a}\).
Let \(T({\mathcal {Q}},t)\) denote the number of new TB cases in the system (1) at the \(t^{th}\) year, then \(T({\mathcal {Q}},t)\) can be written as \( T({\mathcal {Q}},t)=X(t)-X(t-1), \) where X(t) means the cumulative number of people infected with TB cases by the \(t^{th}\) year. We can derive the expression for X(t) as
We utilize \(T({\mathcal {Q}},t)\) to simulate annual new TB cases. Using the Grey Wolf Optimizer(GWO) algorithm (Mirjalili et al. 2014), which allows us to estimate the unknown parameters and initial values of the system (1) are shown in Table 2.
The simulation results are shown in Fig. 2.
From Fig. 2, we think that the simulation results fit well with the number of new TB cases in China from 2007–2020. Based on the parameters in Table 2, we get the basic reproduction number \({\mathcal {R}}_0=1.1938>1\). From the Theorem 7, when \({\mathcal {R}}_0>1\), the endemic equilibrium \(E^*\) is globally asymptotically stable in \({\textbf{Y}}\), so it is difficult to say that TB is eradicated in China.
5.2 Sensitivity analysis
To measure the reliability of the \(\beta \), \(\alpha \), \(\delta _2\), \(\sigma _1\), \(\sigma _2\) estimates obtained through the GWO algorithm, we employ the Markov Chain Monte Carlo method with the Delayed Rejection and Adaptive Metropolis algorithm (Haario et al. 2006). The convergence of the chain is confirmed by using Geweke’s Z-scores. Then, the mean, standard deviation, \(95\%\) confidence intervals and Z-score of the estimated parameters values are shown in Table 3, and the fitting result can be seen in Fig. 3.
We performed a sensitivity analysis using the Partial Rank Correlation Coefficient(PRCC) method. Fixed the parameters \(\delta _1\), \(d_1\), \(d_2\), \(c_1\), \(c_2\), \(k_1\), \(k_2\), \(\gamma _1\), \(\gamma _2\), v(0), e(0), I(0), r(0), the other parameters are shown in Table 3. We calculate the PRCC between the parameter and the basic reproduction number \({\mathcal {R}}_0\), as shown in Fig. 4.
As can be seen from Fig. 4, \(\beta \), \(\sigma _1\) and \(\alpha \) have a large effect on the basic reproduction number \({\mathcal {R}}_0\). Thus, predicting whether the WHO goal will be met by 2030 by changing the values of \(\beta \), \(\sigma _1\) and \(\alpha \). We use the parameter values in Table 1 as the baseline for comparing the control effects, while also considering the effect of implementing a mixture of control measures. First, we consider decreasing the value of parameter \(\beta \) by \(50\%\), increasing the value of \(\sigma _1\) by \(50\%\) and decreasing the value of parameter \(\beta \) by \(95\%\), increasing the value of \(\sigma _1\) by \(95\%\), see Fig. 5a; Similarly, consider changing the values of \(\beta \) and \(\alpha \), see Fig. 5b; Finally, we consider changing the values of \(\sigma _1\) and \(\alpha \), see Fig. 5c.
Based on the simulation results in Fig. 5, it is difficult for China to reach the goal set by WHO. The reason for this is that people are more aware of self-protection, wearing mask, keeping social distance, and learning more about preventive measures through various media reports. The vaccination population is mainly newborns and the vaccination rate is high. Therefore, the treatment measures are relatively good.
As a result, we should take into account the actual situation, to further optimize the prevention and control measures. For instance, strengthening the capacity of medical institutions to improve early diagnosis and treatment of TB. This involves promoting advanced diagnostic techniques and drug susceptibility testing, providing high-quality and sustained anti-tuberculosis drug treatment to ensure that patients can complete the full course of treatment. The government should increase investment in public education and medical resources in low-income areas. In summary, to eliminate TB, the entire society must work together.
6 Optimal control problem
In this section, we discuss the optimal control problem. Optimal control theory is a method for finding control strategies that minimize or maximize an objective function, which opens up new ideas in the study of infectious diseases and the development of control strategies.
First, through the sensitivity analysis, we discovered that controlling the change of \({\mathcal {R}}_0\) sensitive parameters did not have a significant effect. Therefore, we decided to use the treatment rate \(\sigma _1\) of the TB-infected population and the relapse rate \(\gamma (w)\) of the recovered individuals as control parameters. Then two control variables \(l_1(t)\) and \(l_2(w,t)\) are introduced, where \(l_1(t)\) represents the strategy of reinforcement treatment and \(l_2(w,t)\) represents the strategy of controlling relapse in recovered individuals. Since f(I) is monotonically increasing and downward concave, we choose \(f(I)=I\). By substituting this into the system (1)-(3), we obtained the corresponding optimal control problem:
The control set is defined by
where \(l_{1\textrm{max}}\) and \(l_{2\textrm{max}}\) denote the maximum value of the control variables. To consider the cost of disease control, we define the following objective function:
where \(C_1\) and \(C_2\) are positive weights, \(B_1\) and \(B_2\) are the cost of the control. It is assumed that the cost consumed by the measure is a nonlinear quadratic function. We need to find an optimal control \({\mathfrak {L}}^*\), so as to minimize \({\mathcal {J}}(\mathfrak {L^*})\), i.e., to minimize the number of latent and infectious individuals and to minimize the cost of control inputs.
To obtain the necessary conditions for the control problem, we use the Gâteaux derivative rule (Kang 2009) and introduce another control \(l_1^\epsilon (t)=l_1(t)+\epsilon j_1(t),\ l_2^\epsilon (w,t)=l_2(w,t)+\epsilon j_2(w,t)\), where \(j_1(t)\), \(j_2(w,t)\) are called variation functions and \(0<\epsilon <1\). We use \({\overline{S}},~ {\overline{v}},~ {\overline{e}},~ {\overline{I}},~ {\overline{r}}\) to denote S, v, e, I, r with respect to the Gâteaux derivative of \(\epsilon \), also know as sensitivities. Therefore, we have
To obtain the adjoint system, we introduce the adjoint variables \(\lambda _1(t),~ \lambda _2(\theta ,t), \lambda _3(a,t),~ \lambda _4(t)\) and \(\lambda _5(w,t)\). By using the initial condition \(\overline{S}(0)=0\) and the transversality condition \(\lambda _1(T)=0\), with integration by parts, the first equation of the system (55) can be expressed as
where \(\big \langle f,g \big \rangle =\int _0^T fg \textrm{d}t\). With the initial condition \({{\overline{v}}}(\theta ,0)=0\), the boundary condition \({{\overline{v}}}(0,t)=\alpha {{\overline{S}}}(t)\) and the conditions \(\lambda _2(\Theta ,t)=0\), \(\lambda _2(\theta ,T)=0\). Similarly, the second equation of the system (55) is given by
where \(\big \langle \big \langle f,g \big \rangle \big \rangle =\int _0^T \int _0^A fg \textrm{d}a\textrm{d}t\). Under the initial condition \({{\overline{e}}}(a,0)=0\), the boundary condition \({{\overline{e}}}(0,t)=\beta {{\overline{S}}}(t)I(t)+\beta S(t){{\overline{I}}}(t)+\int _0^\Theta \beta \overline{I}(t)\sigma (\theta )v(\theta ,t)\textrm{d}\theta +\int _0^\Theta \beta I(t)\sigma (\theta )\overline{v}(\theta ,t)\textrm{d}\theta +\sigma _2{{\overline{I}}}(t)\) and the conditions \(\lambda _3(A,t)=0\), \(\lambda _3(a,T)=0\). From the third equation of the system (55) yields
Using the initial condition \({{\overline{I}}}(0)=0\) and the transversality condition \(\lambda _4(T)=0\). The fourth equation of the system (55) can be reformulated as
Under the initial condition \({{\overline{r}}}(w,0)=0\), the boundary condition \({{\overline{r}}}(0,t)=\sigma _1\big (1+l_1(t)\big )\overline{I}(t)+\sigma _1j_1(t)I(t)\) and the conditions \(\lambda _5(W,t)=0\), \(\lambda _5(w,T)=0\) hold. The fifth equation of the system (55) can be expressed as
Combining (56)-(60), the adjoint system can be written as
with the transversality conditions
and the boundary conditions
The Gâteaux derivative of the objective function \({\mathcal {J}}({\mathfrak {L}})\) with respect to the control variable \({\mathfrak {L}}\) is
Theorem 8
If there is \({\mathfrak {L}}^*\in {\mathcal {L}}\) such that \({\mathcal {J}}({\mathfrak {L}})\) is minimized, then
Proof
Substituting the system (61) into the equation (62) gives
Substituting (56)-(60) into equation (63), for all \({\mathfrak {L}}\in {\mathcal {L}}\), we obtain
When \({\mathcal {J}}\) is minimized at its minimum, we get
Considering the upper and lower bounds of the control variables, lead to the optimal control \({\mathfrak {L}}^*\) with the following form:
The proof is completed. \(\square \)
7 Discussion
The efficacy of vaccination, incomplete treatment and disease relapse are critical challenges that must be faced to prevent and control the spread of infectious diseases. And age heterogeneity is also a very crucial factor. However, few papers have considered all these factors together to study an epidemic model. Thus, we formulate and analyze a new age-structured SVEIR epidemic model with the nonlinear incidence rate, waning immunity, incomplete treatment and relapse. Comparison with the model put forward in references, if ignores the vaccinated individuals being infected, incomplete treatment and the disease-related mortality rate of latent individuals, namely, \(\sigma (\theta )=0,~ \sigma _2=0,~ \delta _1=0\), the system (1) is identical to that of the model in Li et al. (2017). In addition, if the incidence is bilinear and ignores vaccinated individuals being infected and incomplete treatment, i.e., \(f(I)=I,~ \sigma (\theta )=0,~ \sigma _2=0\), then the system (1) is the same as the model in Liu and Liu (2017). So, if the above parameters are not equal to zero, it means that the model contains more return terms, which is a challenging task for us to prove the properties of the system (1) and establish the Lyapunov function.
In this paper, we define the basic reproduction number \({\mathcal {R}}_0\). The asymptotic smoothness, the uniform persistence and the existence of a global attractor for the semi-flow generated by the system (1) have been shown. In addition, study results show that the basic reproduction number \({\mathcal {R}}_0\) completely determines the global dynamics of the system (1). If \({\mathcal {R}}_0<1\), the disease goes extinct; If \({\mathcal {R}}_0>1\), the disease persists. As can be seen from the expression for the basic reproduction number \({\mathcal {R}}_0\), the age-dependent parameters \(h(\theta ),~\theta _1\) and \(\theta _2\) affect the transmission of disease. The danger of disease outbreaks may be overstated or underestimated if these values are simply taken as constants.
It is worth noticing that we use the annual data of TB in China from 2007–2020 to estimate unknown parameter values in the system (1). These main theoretical results are applied to analyze and predict the trend of TB prevalence in China. It can be concluded that China may not be able to reach the WHO goal under the current TB control measures. It is difficult to say that TB is eradicated in China. Therefore, China needs to optimize its prevention and control methods according to the actual situation. We believe that eliminating TB is possible, but will require integrated and sustained action.
Finally, we consider the sensitive parameters and the actual situation. So, strengthening treatment and controlling relapse were selected as control variables. We obtain the necessary conditions for optimal control, which show that the relapse age w appears in the expression for the optimal control variables and affects the efficiency of optimal control.
References
Adams RA, Fournier JJ (2003) Sobolev spaces. Elsevier
Ai MX, Wang WD (2024) Optimal vaccination ages for emerging infectious diseases under limited vaccine supply. J Math Biol 88(1):13
Chinese Center for Disease Control, Prevention (2023) https://www.chinacdc.cn
Dai WH, Zhang HL (2023) Dynamical analysis for an age-structured model of eating disorders. J Appl Math Comput 69(2):1887–1901
Diekmann O, Heesterbeek JAP, Metz JAJ (1990) On the definition and the computation of the basic reproduction ratio \({R}_0\) in models for infectious diseases in heterogeneous populations. J Math Biol 28(4):365–382
Duan XC, Yuan SL, Qiu ZP, Ma JL (2014) Global stability of an SVEIR epidemic model with ages of vaccination and latency. Comput Math Appl 68(3):288–308
Duan XC, Jung IH, Li XZ, Martcheva M (2020) Dynamics and optimal control of an age-structured SIRVS epidemic model. Math Methods Appl Sci 43(7):4239–4256
Guo ZK, Xiang H, Huo HF (2021) Analysis of an age-structured tuberculosis model with treatment and relapse. J Math Biol 82:1–37
Guo ZK, Huo HF, Xiang H (2022) Optimal control of TB transmission based on an age structured HIV-TB co-infection model. J Frankl Inst 359(9):4116–4137
Guo ZK, Huo HF, Xiang H, Ren QY (2023) Global dynamics of a tuberculosis model with age-dependent latency and time delays in treatment. J Math Biol 87(5):66
Guo ZK, Huo HF, Xiang H (2024) Transmission dynamics and optimal control of an age-structured tuberculosis model. J Appl Anal Comput 14(3):1434–1466
Haario H, Laine M, Mira A, Saksman E (2006) DRAM: efficient adaptive MCMC. Stat Comput 16:339–354
Hale JK (1971) Functional differential equations. Springer, Berlin
Hale JK, Waltman P (1989) Persistence in infinite-dimensional systems. SIAM J Math Anal 20(2):388–395
Hoppensteadt F (1974) An age dependent epidemic model. J Frankl Inst 297(5):325–333
Huang JC, Kang H, Lu M, Ruan SG, Zhuo WT (2022) Stability analysis of an age-structured epidemic model with vaccination and standard incidence rate. Nonlinear Anal Real World Appl 66:103525
Kang YH (2009) Identification problem of two operators for nonlinear systems in banach spaces. Nonlinear Anal Theory Methods Appl 70(3):1443–1458
Kermack WO, McKendrick AG (1927) A contribution to the mathematical theory of epidemics. Proc R Soc Lond Ser A Math Phys Sci 115(772):700–721
Kermack WO, McKendrick AG (1932) Contributions to the mathematical theory of epidemics. II. Proc R Soc Lond Ser A Math Phys Sci 138(834):55–83
Kermack WO, McKendrick AG (1933) Contributions to the mathematical theory of epidemics. III. Proc R Soc Lond Ser A Math Phys Sci 141(843):94–122
Khan A, Zaman G (2018) Optimal control strategy of SEIR endemic model with continuous age-structure in the exposed and infectious classes. Optim Control Appl Methods 39(5):1716–1727
Kumar M, Abbas S (2022) Age-structured SIR model for the spread of infectious diseases through indirect contacts. Mediterr J Math 19(1):14
LaSalle JP (1976) The stability of dynamical systems. In: Society for industrial and applied mathematics
Li XZ, Yang JY, Martcheva M (2020) Age structured epidemic modeling, vol 52. Springer Nature
Li YK, Teng ZD, Hu C, Ge Q (2017) Global stability of an epidemic model with age-dependent vaccination, latent and relapse. Chaos Solitons Fractals 105:195–207
Liu LL, Liu XN (2017) Global stability of an age-structured SVEIR epidemic model with waning immunity, latency and relapse. Int J Biomath 10(03):1750038
Magal P, Zhao XQ (2005) Global attractors and steady states for uniformly persistent dynamical systems. SIAM J Math Anal 37(1):251–275
Mirjalili S, Mirjalili SM, Lewis A (2014) Grey wolf optimizer. Adv Eng Softw 69(3):46–61
Mohammed Awel J, Numfor E, Zhao R, Lenhart S (2021) A new mathematical model studying imperfect vaccination: optimal control analysis. J Math Anal Appl 500:125132
National Bureau of Statistics of China (2023) https://www.stats.gov.cn
Revelle CS, Lynn WR, Feldmann F (1967) Mathematical models for the economic allocation of tuberculosis control activities in developing nations. Am Rev Respir Dis 96(5):893–909
Rodrigue Yves MM, Suares Clovis ON, Richard G (2018) Global dynamics of an SVEIR model with age-dependent vaccination, infection, and latency. Abstr Appl Anal 2018:8479638
Sigdel RP, McCluskey CC (2014) Global stability for an SEI model of infectious disease with immigration. Appl Math Comput 243:684–689
Sun DD, Li YK, Teng ZD, Zhang TL, Lu JJ (2021) Dynamical properties in an SVEIR epidemic model with age-dependent vaccination, latency, infection, and relapse. Math Methods Appl Sci 44(17):12810–12834
Tian X, Wang WD (2020) Dynamical analysis of age-structured pertussis model with covert infection. Math Methods Appl Sci 43(4):1631–1645
Wang SF, Nie LF (2023) Global dynamics and optimal control of multi-age structured vector disease model with vaccination, relapse and general incidence. Qual Theory Dyn Syst 22(1):24
Wang LW, Liu ZJ, Zhang XA (2016) Global dynamics for an age-structured epidemic model with media impact and incomplete vaccination. Nonlinear Ana Real World Appl 32:136–158
Wang JL, Dong X, Sun HQ (2017) Analysis of an SVEIR model with age-dependence vaccination, latency and relapse. J Nonlinear Sci Appl 10(7):3755–3776
Webb GF (1985) Theory of nonlinear age-dependent population dynamics. CRC Press
Yang JY, Modnak C, Wang J (2019) Dynamical analysis and optimal control simulation for an age-structured cholera transmission model. J Frankl Inst 356(15):8438–8467
Yang JY, Yang L, Jin Z (2023) Optimal strategies of the age-specific vaccination and antiviral treatment against influenza. Chaos Solitons Fractals 168:113199
Yu Y, Tan YS, Tang SY (2023) Stability analysis of the COVID-19 model with age structure under media effect. Comput Appl Math 42(5):204
Acknowledgements
SH M is funded by the National Natural Science Foundation of China 12161054. HF H is funded by the National Natural Science Foundation of China 12361101.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Ma, S., Tian, T. & Huo, H. Global stability and optimal control of an age-structured SVEIR epidemic model with waning immunity and relapses. J. Math. Biol. 89, 32 (2024). https://doi.org/10.1007/s00285-024-02131-7
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00285-024-02131-7