Centro Nacional de Desarrollo e Investigación en Tecnologías Libres (CENDITEL)

Revista Electrónica Conocimiento Libre y Licenciamiento (CLIC) Mérida – Venezuela

ISSN: 2244-7423

Analysis of an SEIR-SEI four-strain epidemic

dengue model with primary and secondary
infections
received: 29/05/2014
accepted: 04/06/2014
pages: 3 – 7

Raúl Isea
Fundación Instituto de Estudios Avanzados
Hoyo de la Puerta
Baruta, República Bolivariana de Venezuela
risea@[Link]
Abstract— We propose an SEIR model for the populations The classic example is the SIR model which indicates that
and an SEI model for the vector to describe the transmission there are three significant populations to be examined. They are
dynamics of a four-strain model with both primary and the population that is susceptible to a given disease (S), the
secondary dengue infections. In order to accomplish this, we population that is infected with the disease (I) and the
propose and obtain an analytic solution of a system of 47 coupled population that recovers from the disease (R). Aguilar et. al. [9]
differential equations. This would be the most complete epidemic described an extension to numerically resolve dengue
model proposed to describe the dengue epidemic. epidemics with four strains, employing a SIR model.
Keywords: Dengue; SEIR; SEI; epidemic model In the case of viruses, the mathematical model must include
the incubation or latency period which occurs just before
infection. In this case, the model is called the SEIR model
I. INTRODUCTION where E represents the population that is exposed to the
Dengue is a viral disease, found in tropical and sub-tropical disease. In the case of dengue, the exposure time is
regions of the planet where it is estimated that between 2500 approximately 8-9 days before manifestation of the disease
and 3000 million people are at risk of contracting the disease once it is transmitted by an infected mosquito [10].
[1]. In fact, since 1998 the World Health Organization (WHO)
Our model proposes an analytic solution of a system of 47
has listed dengue as the tenth leading cause of death among all
differential equation that describes the dynamics of dengue
infectious diseases that are prevalent in the world [2].
transmission with four strains. In addition we take into account
Dengue is transmitted primarily by the bite of infected primary and secondary infections employing the SEIR model
female mosquitoes Aedes aegypti, but it also has been for the populations and the SEI model for vector; it is an
associated with other species such as Ae. albopictus, Ae. extension of the model of Janreung and Chinviriyasit published
polynesiensis, and Ae. scutellaris [3]. Dengue has four strains in 2014 [11]. We believe this to be the most complete analytical
denoted by Dengue I-IV [1,2]. The most common form is the analysis of the transmission of dengue.
classic dengue or dengue fever that can often be caused by one
strain. In addition, severe dengue hemorrhagic fever formerly II. MATHEMATICAL MODEL
associated with a secondary infection is caused by antibody-
dependent enhancement processes [1,2]. The model is initially based on the recent model proposed
by Janreung and Chinviriyasit [11] who resolved a system with
The first case of dengue ocurred in Australia in 1954 and 17 differential equations. In this model (Fig. 1), the host
similar outbreaks were observed in the Philippines and population (N) is subsequently subdivided into multiple
subsequently spread to Vietnam, Thailand and other Asian populations based on the following assumptions:
countries [4]. The first epidemic of dengue in the Americas
occurred in Cuba in 1981 which was caused by an Asian strain -The model assumes a homogeneous mixture of the
of Dengue serotype-II [5]. populations of both humans (host) and vectors (mosquito) so
that each mosquito bite is as likely to transmit the virus to
Due to the lack of effective drugs and vaccines against humans regardless of the type of the virus.
dengue fever there has been a lack of effective programs to
help control the disease, and for this reason, a wide range of
mathematical models to describe and characterize the
dynamics of dengue transmission has been developed [6-8].

3
 Centro Nacional de Desarrollo e Investigación en Tecnologías Libres (CENDITEL)
Revista Electrónica Conocimiento Libre y Licenciamiento (CLIC) Mérida – Venezuela
ISSN: 2244-7423

- It was also considered that the mosquitoes that were

exposed to the four virus types at different times, once infected,
cannot recover as has been already established in the scientific
literature [9,11].
Finally, we model the system with a set of 47 ordinary
differential equations which are written as:
4
dS β
=μ ( N−S )−∑ S I vi (1)
dt i=1 N
dEi β
=S I vi −( σ + μ ) Ei (2)
dt N
d Ii
=σ Ei−( γ + μ ) I i (3)
dt
4
d Si β
=γ I i −S i ∑ I vj −μ S i
dt N j=1 (4)
i≠j

d Eij β
=Si I vj −( σ +μ ) E ij (5)
Fig. 1. The compartmental model of host-vector population employed in our dt N
model (see the text for the nomenclature).
d I ij
-The total human population at time t is divided into 38 =σ E ij −( γ + μ ) I ij (6)
subpopulations. This means that the susceptible population will dt
be exposed to an infection by one strain denoted by E i., where 4
dR
i= 1 to 4 is the index to indicate the four strains. Subsequently, =γ ∑ I ij −μR (7)
the populations will become infected (denoted Ii). The dt j=1
i≠ j
secondary infection occurs when an individual is reinfected
with another strain, and in this case, the population that is d Sv β 4 4
exposed to the second infection will be denoted as Eij. The
infected population will be identified as Iij (the second index
represents the second infection while the first index represents
dt
=M −S v v
N ( ∑ Ii + ρ ∑ I ij −μ S v
i=1 j=1
i≠ j
) (8)

the primary infection, with the condition i≠j). Finally the d E vi β 4

human population that recovered is indicated with R. The total
population (N) will be equal:
N = S+E1+E2+E3+E4+I1+I2+I3+I4+S1+S2+S3+S4+
dt N (j=1
i≠ j
)
=S v v I i+ ρ ∑ I ij −( σ v + μ v ) E vi (9)

d I vi
E12+E13+E14+E21 +E23+E24+E31+E32+E34+ =σ v E vi− ( γ v + μv ) I vi (10)
dt
E41+E42+E43+I12+I13+I14+I21+I23+I24+I31 +I32+I34+

I41+I42+I43+R where i represent each strain with i= 1 to 4, and ρ is the rate of

-The total mosquito (vector) population is denoted by M secondary infections contributing to intensity of the disease
and is divided into 9 classes. The first is S v which represents [12]. Of course, the best value of ρ is calculated according to
the number of outbreaks of epidemics observed in each
the mosquito population susceptible to carry the virus. The next geographical area.
four are the mosquitoes that are exposed to the dengue virus
and are denoted by Evi. The last four, corresponding to infected
mosquitoes, are denoted by Ivi. So

M = Sv+Ev1+Ev2+Ev3+Ev4+Iv1+Iv2+Iv3+Iv4

4
 Centro Nacional de Desarrollo e Investigación en Tecnologías Libres (CENDITEL)
Revista Electrónica Conocimiento Libre y Licenciamiento (CLIC) Mérida – Venezuela
ISSN: 2244-7423

[ ]
III. RESULTS M βv ' 4

The epidemiologically relevant bioregion that is E 'vi = I i+ ρ ∑ I 'ij (18)

C5 C6 j=1
symbolized with Ω [9,11,12] is given by j≠i
Ω={ (S,E1,E2,E3,E4,I1,I2,I3,I4,S1,S2,S3,S4,E12,E13,E14,

[ ]
4
M σv βv '
E21,E23,E24,E31,E32,E34,E41,E42,E43,I12, I13,I14, I 'vi= I i+ ρ ∑ I 'ij (19)
C4 C5C6 j =1
I21,I23,I24,I31,I32,I34,I41,I42,I43,R,Sv,Ev1,Ev2,Ev3,Ev4, j ≠i

Iv1,Iv2,Iv3, Iv4 ) ≥ 0 } In this case, the constants are:

The solution of this system of equations uses the same C1 ≡ σ + μ
methodology as explained and published by Janreug and
Chinviriyasit [11]. The solution of our system of equations

[∑ ]
4
suggests that there are two points of equilibrium that we call C 2 ≡ μN + β I 'vi
Point 1 and Point 2. i=1
Point 1: This point correspond a disease free equilibrium
analogous to that found in [11] equal to:
C3 ≡ μ+ γ ; C 4 ≡γ v + μv

[ ]
4 4
(S'=N/µ,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,S'v=M/µv,0,0,0,0,0,0,0,0)
C5 ≡ μ v N + β v ∑ I + ρ ∑ I 'ij
'
i
i=1 j=1
and, j ≠i

Point 2: this is an endemic equilibrium point equal to: C6 ≡ σ v + μv ,

(S',E'1,E'2,E'3,E'4,I'1,I'2,I'3,I'4,S'1,S'2,S'3,S'4,E'12,E'13,E'14, β ' β ' β '
E'21,E'23,E'24,E'31,E'32,E'34,E'41,E'42,E'43,I'12, D1 ≡ I v2+ I v3 + I v4 + μ
I'13,I'14,I'21,I'23,I'24,I'31,I'32,I'34,I'41,I'42,I'43,
N N N
R',S'v,E'v1,E'v2,E'v3,E'v4, I'v1,I'v2,I'v3, I'v4 ) β β ' β
D 2 ≡ I 'v1+ I v3 + I 'v4+ μ
The solution is found with the same mathematical N N N
methodology published in [11], and does contribute anything
about new. The solution found is equal to: β ' β ' β '
D3 ≡ I + I + I +μ
N v1 N v2 N v4
' μ N2
S= (11) β ' β ' β '
C2 D 4 ≡ I v1 + I v2 + I v3+ μ
'
N N N
μ β N I vi
E'i= (12)
C1 C2
The most critical value in the epidemic model is the basic
βμNσ I
' reproduction value (R0) and is the resulting higher value of the
I 'i = vi
(13) eigenvalues of the Jacobian of the 47 differential equations
C1 C2 C 3 when evaluated for each critical point (ie., Point 1 and Point 2);
'
the analysis is difficult to perform and will published in the
β γ μ σ I vi future.
S 'i= (14)
C 1 C 2 C 3 Di Finally, we will examine two cases of which the first results
' 2 '
from the consideration of a single strain and the second results
I β γ μ σ I vj from the consideration of two strains.
E = vi 2
'
ij (15)
C 1 C 2 C 3 Di Model 1. Single-strain model
In the case of a single strain i=1, so
'I 'vi β 2 γ μ σ 2 I 'vj ' ' '
I = 2
ij (16) S i=0, Eij =0, I ij =0. Therefore the solution is:
C 1 C 2 C 23 Di
MN
S 'v = (17)
C5

5
 Centro Nacional de Desarrollo e Investigación en Tecnologías Libres (CENDITEL)
Revista Electrónica Conocimiento Libre y Licenciamiento (CLIC) Mérida – Venezuela
ISSN: 2244-7423

μ N2 N β γ N μ σ I v2
2 '
S '= = S= '
μN + β I 'v β I 'v 2
( σ + μ ) ( μN + β I v1 ) ( μ+ γ ) D'1
1+
μN ' 2 '
' I v1 β γ N μ σ I v2
μN βI
' E =12
E'1 ≡ E' = v
'
( σ + μ )2 ( μN + β I 'v2 ) ( μ+γ ) D'1
( σ + μ ) ( μN + β I v )
' I 'v2 β 2 γ N μ σ I 'v1
βμNσ I ' E21=
'
I1 ≡ I =
'
'
v
( σ + μ )2 ( μN + β I 'v1 ) ( μ +γ ) D '1
( σ + μ ) ( μN + β I ) ( μ+ γ )
v
' I 'v1 β 2 γ N μ σ 2 I 'v2
MN I =
12
S 'v = ( σ + μ )2 ( μN + β I 'v2 ) ( μ+ γ )2 D'1
μv N + βv I '
M βv ' I 'v2 β 2 γ N μ σ 2 I 'v1
E'vi ≡ E 'v = I' I =
21
' ( σ + μ )2 ( μN + β I 'v1 ) ( μ+ γ )2 D '1
( μ v N + β v I ) ( σ v + μv )
MN
' ' M σv βv ' S 'v =
I vi ≡ I v = I D '3
( μv N + β v I ' ) ( σ v + μ v ) ( γ v + μ v )
M β v (I 1+ I 12)
E'v1 =
D'3 C 4
Model 2. Two-strains model.
In this case, the values i are 1 and 2 and the terms that are M β v ( I 2 +I 21 )
not zero are S',E'1,E'2,I'1,I'2,,S'1,S'2,E'12,E'21,I'12,I'21,R',S'v,E'v1,E'v2,
E'v2 =
D'3 C 4
I'v1 and I'v2 whose equations are respectively:
M σ v β v ( I 1 + I 12 )
I 'v1=
' μ N2 D'3 C 4 C6
S=
D'1 M σ v β v ( I 2+ I 21)
I 'v2=
μ β N I 'v1 D'3 C 4 C 6
E'1=
( σ + μ ) ( D '1)
In this case, the constants are:
' μ β N I 'v2
E= 2 ' D '1 ≡(I 'v1 + I 'v2 ) β+ μ N
(σ + μ )( D 1 ) '
' D 3 ≡ μv N + β v ( I 1 + I 2+ ρ( I 12 + I 21) )
μσ β N I
I '1 = v1
The constants C4 and C6 keep their original definitions.
( σ + μ ) (μ+ γ ) ( D'1 )

μ σ β N I 'v2 Numerical simulations:

'
I2 = In Table 1 lists the numerical values are listed of the
( σ + μ ) (μ+ γ ) ( D'1 ) parameters used in the simulation that were obtained from
previous publication [11]. The time series plot is shown in Fig.
β γ N 2 μ σ I 'v1 2(a) for the case of four strains for the time interval of 30 days.
S '1= The solution with two strains is shown in Fig. 2(b). The results
( σ + μ ) ( μN + β I 'v2 ) ( μ+ γ ) D'1 of both simulations are very similar.

6
 Centro Nacional de Desarrollo e Investigación en Tecnologías Libres (CENDITEL)
Revista Electrónica Conocimiento Libre y Licenciamiento (CLIC) Mérida – Venezuela
ISSN: 2244-7423

IV. CONCLUSIONS [3] L. A. Hill, J. B. Davis, G. Hapgood, P. I. Whelan, G. A. Smith, S. A.

Ritchie, R. D. Cooper, and A.F. van den Hurk, “Rapid Identification of
We have developed an SEIR-SEI model of the transmission Aedes albopictus, Aedes scutellaris, and Aedes aegypti Life Stages
dynamics of four-strains of dengue model considering primary Using Real-time Polymerase Chain Reaction Assays,” Am. J. Trop.
and secondary infections, which is an extension of the previous Med. Hyg., vol. 79(6), pp. 866–875 (2008).
model proposed in the literature but only only considering two [4] S.B. Halstead, “The XXth Century dengue pandemic: need for
strains. We found that this model has two equilibrium points: surveillance and research,” World Health Stat Q, vol. 45, pp. 292-298
(1992).
the disease free equilibrium (called Point 1) and Point 2 which
is the endemic equilibrium of the system. It is interesting to [5] G. Kourí, M.G. Guzmán, J. Bravo J, “Hemorrhagic dengue in Cuba:
history of an epidemic,” Bulletin of the Pan American Health
indicate that the model that analyzing the epidemic with four Organization, vol. 20, pp. 24-30 (1986).
strains is very similar with to the model with two strains (Fig.
[6] H. S. Rodrigues, M. T. Monteiro, D. F. Torres, “Vaccination models and
2), and therefore we conclude that it is unnecessary to perform optimal control strategies to dengue,” Math Biosci, vol. 247, pp. 1-12
analytical studies with four strains of dengue, since it is (2014).
sufficient to consider a solution with two different strains. [7] G. Chowell, R. Fuentes, A. Olea, X. Aguilera, H. Nesse, J. M. Hyman,
“The basic reproduction number R0 and effectiveness of reactive
ACKNOWLEDGMENT interventions during dengue epidemics: the 2002 dengue outbreak in
Easter Island, Chile,” Math Biosci Eng., vol. 10, pp. 1455-1474 (2013).
The author wishes to express his sincere thanks to Prof. [8] M. Canals, C. González, A. Canals, D. Figueroa, “Dinámica
Karl Lonngren and Johan Hoebeke for their unconditional help epidemiológica del dengue en Isla de Pascua,” Rev. chil. Infectol.,
and the comments concerning the manuscript. vol.29, pp. 388-394 (2012).
[9] M. Aguilar, B.W. Kooic, F. Rochaa, P. Ghaffari, N. Stollenwerk, “How
much complexity is needed to describe the fluctuations observed in
dengue hemorrhagic fever incidence data?,” Ecological Complexity,
vol. 16, pp. 31–40 (2013).
[10] M. Chan, M.A. Johansson, “The Incubation Periods of Dengue Viruses,”
PLoS ONE, vol. 7(11), pp. E50972 (2012).
[11] S. Janreung and W. Chinviriyasit, “Dengue Fever with Two Strains in
Thailand,” IJAPM., vol. 4, pp. 55-61 (2014).
[12] A. Korobeinikov, “Global Properties of SIR and SEIR Epidemic Models
with Multiple Parallel Infectious Stages,” Bull Math Biol., vol. 71, pp.
75–83 (2009).

TABLE 1. PARAMETERS OF THE SYSTEM OF DIFFERENTIAL EQUATIONS FOR THE MODEL

SEIR-SEI EPIDEMIC DENGUE MODEL
Variables Description Value
S Population of humans susceptible
S(0) = 10
of any strains
infection rate of the disease in
β 0.9
population
infection rate of the disease in
βv 1.0
vector
Si Humans susceptible with strain i S1 = 1,S2=S3=S4=0

Ei Humans exposed with i strains All values are zero

Fig. 2. The numerical solution of this model with (a) four strains, and (b) two
strains according to Table 1 values. S represent the susceptible Ii Humans infected with strain i I1 = 5, I2=I3=I4=0
population, I is equal to the sum of I1+I2+I3+I4. S1 is the humans Humans infected with strain i but
Eij All values are zero
exposed with strain 1, and finally Sv is the total population of susceptible to strain j
Humans infected with strain i and
mosquitoes susceptible to any strains. Iij All values are zero
reinfected with j strains
Humans susceptible of strain j but
Sij All values are zero
infected with strain i
REFERENCES Population of mosquitoes
Sv 1
[1] I. Kurane and T. Takasaki, “Dengue fever and dengue hemorrhagic susceptible to any strains
fever: challenges of controlling an enemy still at large,” Rev. Med. Virol. Evi Mosquitoes exposed with strain i All values are zero
Vol. 11, pp. 301-311 (2001).
[2] A. Seijo, “El dengue como problema de salud pública,” Arch. Argent Ivi Mosquitoes infected with strain i All values are zero
Pediatr., vol. 99, pp. 510-521 (2001).