Some properties of the KermackMcKendrick epidemic model with fractional derivative and nonlinear incidence
 Emile Franc Doungmo Goufo^{1}Email author,
 Riëtte Maritz^{1} and
 Justin Munganga^{1}
https://doi.org/10.1186/168718472014278
© Doungmo Goufo et al.; licensee Springer. 2014
Received: 3 June 2014
Accepted: 17 October 2014
Published: 30 October 2014
Abstract
KermackMcKendrick epidemic model is considered as the basis from which many other compartmental models were developed. But the development of fractional calculus applied to mathematical epidemiology is still ongoing and relatively recent. We provide, in this article, some interesting and useful properties of the KermackMcKendrick epidemic model with nonlinear incidence and fractional derivative order in the sense of Caputo. In the process, we used the generalized mean value theorem (Odibat and Shawagfeh in Appl. Math. Comput. 186:286293, 2007) extended to fractional calculus to conclude some of the properties. A model of the KermackMcKendrick with zero immunity is also investigated, where we study the existence of equilibrium points in terms of the nonlinear incidence function. We also establish the condition for the disease free equilibrium to be asymptotically stable and provide the expression of the basic reproduction number. Finally, numerical simulations of the monotonic behavior of the infection are provided for different values of the fractional derivative order α ($0\le \alpha <1$). Comparing to the model with firstorder derivative, there is a similar evolution for close values of α. The results obtained may help to analyze more complex fractional epidemic models.
Keywords
1 Introduction and important facts
Considered as one of the first compartmental models, KermackMcKendrick epidemic model was developed in the late 1920s with the pioneering work of Kermack and McKendrick [1, 2]. The model is described as the SIR model for the spread of disease, which consists of a system of three ordinary differential equations characterizing the changes in the number of susceptible (S), infected (I), and recovered (R) individuals in a given population. The model is a good one for many infectious diseases, despite its simplicity. Ever since, numerous and more complex compartmental mathematical models have been developed. For instance in biology, modeling is particularly useful in studying organs like the lungs, heart, intestinal edema and cancer, etc. Almost all these models take their source on KermackMcKendrick’s model and serve to help gain insights into the transmission and control mechanisms of diseases like HIV, TB, malaria and their interactions with others. Then most of the works done on modeling the dynamics of epidemiological diseases have been limited only to models based on (a system of) classical firstorder differential equations.
However, there is a growing interest in applying fractional calculus to mathematical epidemiology since it has turned out recently that many phenomena in different fields, including sciences, engineering, and technology, can be described very successfully by the models using fractionalorder differential equations. As an example, the concept of a fractional Laplacian operator in the theory of Lévy flights [3] is a typical application of fractional derivatives. It leads to the theory of sub and superdiffusion, which is well applicable in reactiondiffusion systems. In the field of mathematical epidemiology, especially the superdiffusion case is of interest since it describes a more realistic spreading than normal diffusion on regular lattices. Moreover, it has recently been noted [4] that some epidemic models based on dynamics with firstorder derivative were unable to reproduce the statistical data collected in a real outbreak of some disease with enough degree of accuracy. Another example is the application of halforder derivatives and integrals that proved to be more useful and reliable for the formulation of certain electrochemical problems [5–7] than the classical models with derivative of order one. Accordingly, a number of works [8–14] have successfully generalized, in various ways, classical derivatives of integer order to derivatives of fractional order.
Therefore, following the same idea as above, we analyze and provide some interesting properties of the KermackMcKendrick epidemic model with nonlinear incidence and fractional derivative order less than one. Therefore, we generalize the classical KermackMcKendrick model with the standard mass balance incidence.
2 Model construction
In this model, a population of size $N(t)$ is divided into different classes, disjoint and based on their disease status. At time t, $S=S(t)$ is the part of population representing individuals susceptible to a disease, $I=I(t)$ is the part of population representing infectious individuals, $R=R(t)$ is the part representing individuals that recovered from the disease. We will need the following hypotheses about the transmission process of the infectious disease and its host population.

The transfer rates from a compartment to another is supposed to be proportional to the population size of the compartment. For example, the transfer rate from S to I, the (nonlinear) incidence rate can be written as $\beta g(I)S(t)$, where β is some rate constant and the function g characterizing the nonlinearity is assumed to be at least ${C}^{3}(0,{N}_{0}]$ with $g(0)=0$ and $g(I)>0$ for $0<I\le {N}_{0}$, with ${N}_{0}=N(0)$ the initial total population. Note that the classical mass balance incidence has $g(I)\equiv I$ and β is called the transmission coefficient.

The ensemble of individuals in the host population is well mixed and homogeneous so that the law of mass action holds: the number of contacts between hosts from different compartments depends only on the number of hosts in each compartment. For instance, the recovery rate, that is, the number of individuals recovering from the disease per unit time, can be expressed as μI, with μ the recovering rate.

The transmission mode is supposed to be horizontal and happens through direct contact between hosts.

There is no latent period and all the infected individual hosts become infectious following an infection.

There is no reinfection or loss of immunity and then no transfer from the compartment R back to S.
3 Properties of the model
Property i: total population
and therefore, the model (1) assumes that the total population is constant all the time.
Property ii: wellposedness
If the initial condition is nonnegative then the corresponding solution $(S(t),I(t),R(t))$ of the fractional model (1) is nonnegative all the time $t>0$.
 (a)On the coordinate plane IR, we have $S=0$ and${D}_{t}^{\alpha}{S}_{{}_{S=0}}=0.$
 (b)The same way, on the coordinate plane SR, we have $I=0$ and${D}_{t}^{\alpha}{I}_{{}_{I=0}}=\beta g(0)S=0.$
 (c)Finally, on the coordinate plane SI, we have $R=0$ and${D}_{t}^{\alpha}{R}_{{}_{R=0}}=\gamma I\ge 0.$
To conclude the property, we need the generalized mean value theorem proved in [16] and stated as follows.
where $0\le \tau <t$.
Thus, considering the interval $[0,{t}_{2}]$ for any ${t}_{2}>0$, this theorem implies that the function $U:[0,{t}_{2}]\to {\mathbb{R}}^{+}$ is nonincreasing on $[0,{t}_{2}]$ if ${D}_{t}^{\alpha}U(t)\le 0$ for all $t\in (0,{t}_{2})$ and nondecreasing on $[0,{t}_{2}]$ if ${D}_{t}^{\alpha}U(t)\ge 0$ for all $t\in (0,{t}_{2})$. Then, from the points (a)(c) above, the vector field on each coordinate plane is either tangent to the coordinate plane or and points to the interior of ${\mathbb{R}}_{+}^{3}$.
Property iii: extension of solutions on $(\mathrm{\infty},+\mathrm{\infty})$
The fundamental theory of differential equations (with integer or noninteger derivative order) tells us that their bounded solutions can be extended for all time $t\in \mathbb{R}$. From Properties i and ii we have at least ${N}_{0}\ge S\ge 0$, ${N}_{0}\ge I\ge 0$, and ${N}_{0}\ge R\ge 0$, meaning that solutions $(S(t);I(t);R(t))$ are bounded in their maximal interval of existence. Thus, solutions to the fractional model (1) exist for $t\in (\mathrm{\infty},+\mathrm{\infty})$.
Property iv: asymptotic behavior of solutions
Thus, from Theorem 3.1, $S(t)$ is nonincreasing (decreasing) on its domain of existence and since $0\le S(t)$ for all $t\in (\mathrm{\infty},+\mathrm{\infty})$, we conclude that ${S}_{\mathrm{\infty}}$ exists and $0\le {lim}_{t\to \mathrm{\infty}}S(t)={S}_{\mathrm{\infty}}$.
and from Theorem 3.1, $R(t)$ is nondecreasing (increasing) on its domain of existence and since ${N}_{0}\ge R(t)$ for all $t\in (\mathrm{\infty},+\mathrm{\infty})$, we conclude that ${R}_{\mathrm{\infty}}$ exists and ${N}_{0}\ge {lim}_{t\to \mathrm{\infty}}S(t)={R}_{\mathrm{\infty}}$.
Finally, the existence of ${I}_{\mathrm{\infty}}\phantom{\rule{0.25em}{0ex}}(={N}_{0}{S}_{\mathrm{\infty}}{R}_{\mathrm{\infty}})\ge 0$ is therefore obvious since ${S}_{\mathrm{\infty}}$ and ${R}_{\mathrm{\infty}}$ exist.
4 KermackMcKendrick fractional model for disease with zero immunity and nonlinear incidence
4.1 Existence of equilibrium points when $0\le \alpha <1$
Recall [17] that $\frac{b}{d}$ is the maximum possible value of $\overline{I}$ and in the classical mass action incidence, $\frac{b\beta}{d(\gamma +d)}$ is called the contact reproduction number. Note that the number of solutions of (9) depends on the nonlinear incidence function $g(I)$, ${lim}_{I\to 0}\frac{g(I)}{I}\equiv h(0)$ and the sign of ${h}^{\prime}(I)$.
Theorem 4.1 The model (6) has no endemic equilibrium point if $h(0)\le \frac{d(\gamma +d)}{b\beta}$ and ${h}^{\prime}(I)<0$ for all $I\in (0,\frac{b}{d})$ and has one endemic equilibrium point if $h(0)>\frac{d(\gamma +d)}{b\beta}$ and ${h}^{\prime}(I)<0$ for all $I\in (0,\frac{b}{d})$.
Proof The proof of this theorem follows the same approach as [[18], Theorem 3.1], where we can substitute H by $\frac{b}{d}$ and σ by $\frac{b\beta}{d(\gamma +d)}$. □
4.2 Stability of the DFE point ${x}^{o}$ when $0\le \alpha <1$
It is obvious that ${\lambda}_{1}$ satisfies (11) and ${\lambda}_{2}$ satisfies (11) if $\frac{\beta {g}^{\prime}(0)b}{d(\gamma +d)}<1$, which concludes the proof. □
For the fractional model (6), the quantity $\frac{\beta {g}^{\prime}(0)b}{d(\gamma +d)}$ is usually referred to as the basic reproduction number denoted by ${\mathcal{R}}_{0}$ and defined as the number of secondary cases that one case will produce in a completely susceptible population. The biological interpretation of ${\mathcal{R}}_{0}$ is that if ${\mathcal{R}}_{0}$ exceeds one, then the epidemic will occur, but if it is less than one, then the infection dies out.
5 Numerical simulations
If $\frac{\beta {g}^{\prime}(0)b}{d(\gamma +d)}=\frac{\beta b}{d(\gamma +d)}<1$ then, according to Theorem 4.2, the disease free equilibrium of (6) is asymptotically stable.
If $\frac{\beta b}{d(\gamma +d)}>1$ then $\frac{d(\gamma +d)}{\beta b}<1=h(0)$ and ${h}^{\prime}(I)=\frac{d}{b}<0$. Hence, according to Theorem 4.1, there is one endemic equilibrium point.
6 Conclusion and possible future analysis
Making use of the mean value theorem extended to fractional calculus [16], we have provided some remarkable and useful properties of a fractional KermackMcKendrick epidemic model with nonlinear incidence function. We have shown that the existence of equilibrium points and the stability of the disease free equilibrium both depend on the nonlinear incidence function and we provide the expression of the basic reproduction number for the fractional model. The results obtained here are seen as the extension of the previous works with the inclusion of the derivative of order less than unity and a nonlinear incidence. Numerical simulations show the monotonic behavior of the infection $I(t)$ and comparing to the dynamics with firstorder derivative, we note a similar evolution for close values of α ($0\le \alpha <1$). This may therefore help analyze the stability of the endemic equilibrium point for the model and also help to investigate more complex fractional epidemic models.
Declarations
Authors’ Affiliations
References
 Kermack WO, McKendrick AG: A contribution to mathematical theory of epidemics. Proc. R. Soc. Lond. A 1927, 115: 700721. 10.1098/rspa.1927.0118View ArticleGoogle Scholar
 Kermack WO, McKendrick AG: Contributions to the mathematical theory of epidemics. Bull. Math. Biol. 1991, 53: 3355.Google Scholar
 Brockmann D, Hufnagel L: Front propagation in reactionsuperdiffusion dynamics: taming Lévy flights with fluctuations. Phys. Rev. Lett. 2007., 98: Article ID 178301Google Scholar
 Pooseh S, Rodrigues HS, Torres DFM: Fractional derivatives in dengue epidemics. In Proceedings of the International Conference on Numerical Analysis and Applied Mathematics. Edited by: Simos TE, Psihoyios G, Tsitouras C, Anastassi Z. Am. Inst. of Phys., Melville; 2011:739742.Google Scholar
 Oldham KB, Spanier J: The Fractional Calculus. Academic Press, New York; 1974.Google Scholar
 Miller KS, Ross B: An Introduction to the Fractional Calculus and Fractional Differential Equations. Wiley, New York; 1993.Google Scholar
 Podlubny I: Fractional Differential Equations. Academic Press, New York; 1999.Google Scholar
 Kilbas AA, Srivastava HM, Trujillo JJ: Theory and Applications of Fractional Differential Equations. Elsevier, Amsterdam; 2006.Google Scholar
 Boto, JP: Review on fractional derivatives. CMAF manuscript (2009)Google Scholar
 Samko SG, Kilbas AA, Marichev OI: Fractional Integrals and Derivatives. CRC Press, Boca Raton; 1993.Google Scholar
 Baleanu D, Diethelm K, Scalas E, Trujillo JJ: Fractional Calculus: Models and Numerical Methods. World Scientific, Singapore; 2012.Google Scholar
 Atangana A, Botha JF: A generalized groundwater flow equation using the concept of variable order derivative. Bound. Value Probl. 2013., 2013: Article ID 53 10.1186/16872770201353Google Scholar
 Diethelm K: The Analysis of Fractional Differential Equations. Springer, Berlin; 2010.View ArticleGoogle Scholar
 Doungmo Goufo EF, Maritz R, Mugisha S: Existence results for a Marchaud fractional, nonlocal and randomly position structured fragmentation model. Math. Probl. Eng. 2014., 2014: Article ID 361234 10.1155/2014/361234Google Scholar
 Caputo M: Linear models of dissipation whose Q is almost frequency independent. J. R. Aust. Hist. Soc. 1967, 13(2):529539.Google Scholar
 Odibat ZM, Shawagfeh NT: Generalized Taylor’s formula. Appl. Math. Comput. 2007, 186: 286293. 10.1016/j.amc.2006.07.102MathSciNetView ArticleGoogle Scholar
 Liu WM, Hethcote HW, Levin SA: Dynamical behavior of epidemiological models with nonlinear incidence rates. J. Math. Biol. 1987, 25: 359380. 10.1007/BF00277162MathSciNetView ArticleGoogle Scholar
 Hethcote HW, van den Driessche P: Some epidemiological models with nonlinear incidence. J. Math. Biol. 1991, 29: 271287. 10.1007/BF00160539MathSciNetView ArticleGoogle Scholar
 Matignon D: Stability results for fractional differential equations with applications to control processing. 2. Computational Engineering in Systems Applications 1996, 963968. (Lille, France)Google Scholar
 Diethelm K, Ford NJ, Freed AD, Luchko Y: Algorithms for the fractional calculus: a selection of numerical methods. Comput. Methods Appl. Mech. Eng. 2005, 194: 743773. 10.1016/j.cma.2004.06.006MathSciNetView ArticleGoogle Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.