AN EPIDEMIC MODEL
WITH DENSITY DEPENDENT
PARAMETERS AND VACCINATION
Q.J.A.KHANandB.S.BHATT
Department
ofMathematics andComputingCollegeof Science Sultan QaboosUniversity P O Box36, Postal Code 123 AI-Khod
Muscat,SULTANATEOF OMAN
(Received July29,1993andin revisedformDecember28, 1994)
ABSTRACT. Amodel originally suggested by Greenhalgh 12] and later modifiedbythat same author [13,14] is considered under the assumption that the transmission coefficient is inversely proportional to the total population size Thepurpose ofthisstudy istosee theeffect ofthis density
dependenttransmission coefficient on thestabilitycriteriafortheequilibrium ofthemodel equations. It
isfoundthatGreenhalgh’s resultsarestillvalid
KEY WORDS AND PHRASES. Epidemic model, density-dependent death rate and transmission
coefficient,vaccination,stability
1992AMSSUBJECTCLASSIFICATION CODES. 92B05
1. INTRODUCTION.
Thispaperuses arelatively simpledeterministic mathematical modeltodescribe infectious diseases like
measles, rubella, chickenpoxandmumps Thebookby Bailey [4]describes much of thebackground in the area ofepidemicmodelsupto 1975. Weareinterested inlookingat amodelwherethe parameters
which describe the transmission of the disease and the death rate ofindividuals are dependent onthe
number ofindividuals inthe population. We also take into account the fact that infected individuals suffer a higher death rate than other individuals becausethey have the disease Some models for a
populationwith adensitydependentdeathratehave beenstudiedbyNisbetandGumey [20] Anderson
[1] has analyzed anepidemic model with birth and death in which infected individuals suffer ahigher
death ratethanother individuals. Heconsiders the deathrate of individualsto bedensity independent
Dietz and Schenzle [10] and Mollison [19] consideredthe transmission coefficient to bedependenton
population density Anderson etal [2]have discussed models for rabies withadensitydependentdeath rate McLeanand Anderson 17,18]also incorporatedthisfeatureindiscussingamodel ofmeasles Gao
and Hethcote 11 studiedSIRS/SISmodelswith restrictedpopulation growth by logistic equation dueto
density dependence inboth birth anddeathrates They also considereddensitydependent transmission coefficient (inversely proportional to the total population size) They obtained four equilibria and discussed theirstability results
Busenberg and Van den Driesche [8], Diekmann and Kretzschmav [9], Greenhalgh [12-14], May and Anderson[3],Pugliese[21,22]andTuljapurkarandMerd,th-John[23]
Greenhalgh [12] described a mathematical model for a disease where the death rate is a monotonicallyincreasingfunctionof the number of individuals in the population and infected individuals sufferahigher deathratethanother individuals Thesameauthor in[13,14]studied the modified model witha class ofndividualswho areincubatingthe disease and vaccination ofsusceptibleindividuals In
our modelwehave considereddensity dependenceinthe transmission coefficient(inversely proportional
to the total population) together with the vaccination ofsusceptible individuals Considering density dependenceof the transm,ssionparameter, wecan relax theassumption thatthenumber ofcontacts per
unit timeper susceptibleindividual increaseslinearlywiththe populationsize 2. MATHEMATICALMODEL
Weexaminea modelmwhich anindividual startsasa susceptible, catchesthe diseaseand afterashort infectiousperiodbecomespermanentlyimmunetoit Weassumetheindividualswho aresusceptibleare vaccinated at aconstant per capitalratec The spread of thedisease ismodeled bya setof differential equations which describe the transfer of individuals between these classes The system of ordinary
differentialequationswhich describethespread ofthe disease is asfollows
dx b
rN xy cx
f(N)x
(2 la)
dt N
dy b
xy-f(N)y-
(u
+
c)y (2 lb)dt N
dz
uy
+
czf(N)z
(2
lc)dt dN
rN-
f(N)N-
cy (2 ld)dt
with suitable initialconditions,where oneof the equationsisredundantsince
x(t)
+
y(t)+
z(t)
N(t)
x(t)
represents thepopulation(ordensity)ofthe susceptible classattimet,y(t) represents thepopulation of theinfectedclassattimet,
and
z(t)
represents thepopulation of theimmuneclassattimet, e.those individualswho have hadthedisease,have recoveredandarepermanentlyimmune
In ourmodel, risthe birth rate, isthe transmissioncoefficient,
f(N)
isthedensitydependentdeath rate taken as a continuous, strictly monotonic increasing function of N (Greenhalgh [12-14],
considered
-
constant), cistheadditional deathratesufferedbyinfectedindividuals, andv is therate atwhich infected individualsbecome immune, sothatv-1
istheaverageinfectiousperiodintheabsenceof adeathrate The probabilitythatasusceptibleindividualmeetsand becomes infectedbyan
(At)
+
o(At)
The per capitarateof vaccination ofsusceptibleinfectious individual in
[t,
+
At]
isindividuals isc, sothatin asmalltime interval
[t,
+
At]
the number of susceptible individualswho are vaccinated iscxAt+
o(xt)
Thisterm cxmustbe subtracted from the equation(2. la)correspondingto the fractionof susceptible individuals whoarevaccinatedand addedtoequation(2 c)correspondingto new immune individuals. We are interested inperforming anequilibrium and stability analysis of this model Thestability analysishelpsustodeterminethelong-termbehaviorof thesystem, e whetherthe diseasepersists3. EQUILIBRIUM ANALYSIS.
The first stepis to examinethe possible equilibrium solutions of these equations Firstofall, we shallsuppose that
f(o)
-imf(N)
>
r, so that if the populationsize islarge enoughthe deathrate exceeds the birthrate Settingallofthe time derivativetozeroinsystem(2 1),wededuce thefollowingTHEOREM Let
,
,
and denote theequilibrium numbersofsusceptible,infectedandimmune individualsrespectively Let Ndenote thetotal numberofindividualsatequilibriumThere are threepossible equilibria
(i) When there is no disease present because thepopulation hasdiedout
===N-0.
(31)(ii) If
f(c)
> r> f(0),
thepopulationhasreachedanequilibriumlevelbut the dsease hasdied outf(N)N
cN-
=0, "2= (32)c
+
f(N)
c+f(N)
and
r=
f(N).
(33)(iii) The disease is present and the equilibrium values of susceptible, infected and immune
individuals,are
(u
+
+
f
(N) )N
(r
f
(N)
)N
b Y= a
.{ub(r
f(1"))
+
ca(.
+
a+
f())}
"
abf
()
(34)Population valueNsatisfiesthe equation
b
(f(N)
+
c)(u
+
a+
f(N))
....
(3 5)a
f(N)
Thisequilibriumexists ifand onlyif b
_>
PROOF. Thistheorem is proved alongsimilar lines tothe corresponding results for therelated modelsinGreenhalgh
[12-14]
Setting thetimederivatives to zero insystem(21),wededuce thatr
b’
c
rf(/)
0 (36a)b’-
f(/’)-
.
(u
+
a)
0 (3 6b)+ c
f()
0 (3 6c)r
f(’)/"
a
0. (3 6d)From equation (3 6b) either
-
0 or fVl+f"+)r NOW 0 implies from equation (3 6d)that N-0 or
r--f(N)
Henceifr:/=f(N)
then =y=z=N=0Ifr=f(N)
then it isstraightforwardtoshow thatequations
(36a)
and(36c)
yieldf(N)N
cN=
=o,
=
c
+ f
(N)
c+
f
(N)
anequilibriumsolutionfor any value of If f()+("+)thenfrom equation(3 6a)wehave
r
c
f(r)
r./"
(c
+
f(/))
Y b’
f(N)
+
u+
a b(r
f
(N)
)NEquatingtheseexpressionsfor
,
we get anequation forNwhich canbe reducedto b(f(N)
+
c)(u
+
a+
f(N))
c
f(l’-
+
(.
+
a,r)f()
Thustheequilibriumvalues5,
,
2 andNmustsausfythevaluesgivenin(i), (ii)and(iii) of thetheorem The first equilibrium is always possible The second equilibrium is well defined if and only iff(0) <
r< f(c)
The thirdequilibriumexists ifandonlyif(r
+
c)(u
+
c+
r)b.
,swell definedusingthefollowinglemma
LEMMA. Theequation b_ f(9)+iu+c,-r)fl)-"r has a unique positive root
N,
ifand onlyiff() #,
where+
istheunique positiverootoft
0HereThat valueof Nsatisfiesr
_>
f(N,)ensuring ")spositiveifand onlyif(r
+
c)(
+
a+
r)b>
PROOF. Consider theequation
b
(:(N)
+
)(,
+
,
+
:(N))
cf()2
+
(,
4-r)f(//)
r Withthetransformationf(N)
Thenb
(
+
)(.
+
+
)
+(.+-)-(+c)(++)
Consider g() +(.+_)_.,.. Here g(’) has roots c, c rootsof
Q()
O,whereQ()
2
+
(.
+
r)
The asymptotes arethe
Q(()
hastworealroots(_ and(+
givenby1
(a
+
t,-r)
-4- 1_,
+,
[(
+.
)
+
4.]
The function g() is negative for 0
<
<
+
and monotonically decreasing for>
+.
Hence the equationb
has aunique positiveroot(if andonly if
f(o) > (+
Otherwise, iff(c)
<(.
this equationhas nopositiverootfor( Since
f(oo) >
(+
denote the uniquerootby(andletNbethe correspondingvalue of N For the solution to be feasible we requiref()
_<
r or<
r as __b is constant and g(() isismonotonically increasingin(andzeroat([seethegraph ofg(()] Hencer
_>
(ifandonlyifg()
_>
0.This condition isequivalentto
b>
(r
+
c)(u
+
c+
r) Thiscompletestheproof ofthe lemmab
4. STABILITY
Stability Analysis of Equilibrium (i).
By
Liapunov’s(Jordanand Smith15])
indirectmethodwedeterminethe stabilitybehaviorof the system of differential equations(2 1)
which describe the spread of the disease at the equilibrium2 N 0 Considerthe Liapunov function
L
z+
y+
zwhichleadstoL’=(r-f(N))N-ay<(r-f(O))N-ay<O
when N>0 andr<f(0).
(41)We conclude that the zero solution of
(2 1)
is globally asymptotic stable (GAS) for r<f(0)
sinceL
<
0and unstable forr>
f(0).
Thereforeitfollows that thezerosolutionoftheoriginalsystem(2 1)isGASforr
< f(0)
andunstable forr> f(0)
Analysis of Equilibrium(ii).
Theequilibrium valuesare
f(N)N
cN=
=0,=
c
+
f
(N)
c+
f
(N)
Consider a smallperturbation aboutthisequilibriumlevel
X’--+Xl,
Y=+Yl,
Z+’+Z
and N N+n.
Substitutingthese into the differential equationswhichdescribethespread ofthedisease, and usingthe
approximation
f(
+
n
f()
+
nf’()
+
o(n)
we get thestabilitymatrixAaswhichgivesthe characteristicequationas
(f(N)
+
A)(c
+
f(N)
+
A)(f()
+
u+
a) A(I’().
+
A)
O. (4 3)From equations (32), (3 3) and (4.3) we get the equilibrium (ii) to be locally stable for small perturbationsif
Ro
<
1and locally unstableifRo
>
IwhereRo
(4 4)(,:
+
,,.)(,,-
+
,,,
+
)Local Stability ofEquilibrium (iii).
Usingthe sameprocedureasfor equilibrium(ii),weget thestabilitymatrix6’asfollows
---f(9)
-4
0N
0 0
C N
u
f(gr)
0 -a 0
,.
+
f’
(9)’
-j,
)y-f’
r
f(.r)
f’(/)2
(45)
The determinantofC-
AI
isIC-
All
(f(Fr)
+
A)IDI,
where----f(N)-a D=
0
-A -1
)Y-o r-
f()
f(l)ll
,k(4.6)
The correspondingcharacteristicequation for
D
is(4.7)
where
(4.8)
TheRouth-Hurwitz(May
16])
stabilitycriteria forthethirdordersystemis(i) a >0,
a2>0
and as>0;Nowwe willshowthe positivity of allthe constantsappearingin(4 8)
Let
(4 10)
Rewriting equation(4 8)as
al 7dl -I--a2 731 _[_
ft
(.//")732
a3 Wl -[-f’()w2
(4 ll)
Since
f(N)
is monotonicincreasinginN, f’(N) >
0 Hencetoshow(i)of equation(4 9)we willshowthatul,u2, Vl, v2, Wlandw2areall positive
Using equation
(3
5)we getU
--(r
+f(N)
+c,
wherer
> f(N)
and from equation(3 4)b>
a,thereforeu>
0Thensinceu2 N
>
0,it mustbe thatal>
0Now731 canberewritten as
{
b(r
f())
+
f()
+
c}
(r
f(l’))(u
+
a+
f())
731(7"-
f
(N)
b
+
-(r-
f(N))(u
+
a+
f(N)).
Nowvl
>
0ifb
b(r
f(N))
(f(l’)
+
c)
(u
+
a+
f(.l’))
+
-(u
+
a+
f(l’))
>
O.Withthehelpof equation
(3
4),theaboveinequality reducestothefollowingform---(r-f(.))-b__{o
f(’)(f(l)(u
f(/r))+u-l-a)-r(f()’)
+
+
b
(u
+
a+
f(N))
+
-(u
+
a+
f(l’))
>
O.(u
+
a+
f())2
(b
a)
bra ora(f()
+
u+
a)
>
O. (412)
Hence
(u
+
a+
f(N))(r- f(N))
Since <
/,
from equaUon(6)rf()
<
a,so thatr-
f(N)
<a+u+ f(N).
(4 4)
(4 5)
Inequalities (4 14)and(4 15)combinedtoshow that theinequality(4 13)istruei.e
v
>
0Now
_,
>
0 fb
+
9f(-/-
5
+
9
>
0,or
(b
a)+
Nf(N)
+
cN>
0Fromequation(34),b
>
a,thereforetheabove inequalityistrueandhencea.2>
0Using
equauon
(35)thetermsinsidethebracket can be written asThus
w
0Noww2
>
0ifthe sumof alltermsinsidethebracketispositive. Using equation(3
5),thetermsinsidethebracket become
a
--(r
f(lr))
(f(/r)
+
c)
+
--(u
+
a+
f())
(.
+
a+
f(/))
(4.16)
Wehavealready shown above that thesumofalltermsof(416)
ispositiveHence,a3
>
0Nowwehavetoshowthata a2 a3
>
0.alag. a3
(UlVl
Wl)
+
f/()(UlV2
+
U2Vlw2)
+
f"2(l)(u2v2)
where
w
0 and ul, u2,v,v2,w2all are positive i.e. to show aa2as >
0 we will show that5. SUMMARYANDCONCLUSIONS
Inthispaperwehave studieda simplemathematical epidemicmodel with vaccination There are three possible equilibrium situations which arise Equilibrium wthpopulation extinct will beglobally asymptotically stable ifr
_< f(0)
and will be unstable ifr> f(0)
Equilibriumwithsteady populationb,
<
andlocallyunstableandno diseasepresentwillbelocally stabletosmallperturbationsif
if..+(
>
1 Localstability oftheequilibriumwith disease present is examinedanalyticallyandisfound stable In this paper we have extended the work ofGreenhalgh [12-14] It is unrealistic to consider the transmissionparameteraconstantbecausethis assumes contactperunit timeper susceptible
individual increaseslinearlywiththepopulationsize Thisassumptioncanberelaxed bytaking adensity
dependent transmission parameter Here we consider the transmission coefficient as inversely
proportional to the total number of individuals in the population Gao and Hethcote
[11]
have alsoconsidered density dependence in transmission coefficient similar to ours but they have taken density
dependent restrictedgrowthduetoadecreasingbirthrateandanincreasing deathrateasthepopulation
size increasestowardsits carrying capacity They have obtained four equilibria The first equilibrium
when disease fades out and population size tendsto zeroand showed that it is always a saddle The
second equilibriumwhere thepopulationsize isupthecarrying capacityisfoundtobeGASunder certain
threshold conditions The third equilibrium occurs when thebirthrate is density independent and the deathrate isdensitydependent and isLASwhen certain conditions are met The fourthequilibriumis
achieved when thebirthrateisdensitydependentand the deathrate isdensity independent andit isalso LAS under differentthreshold conditions TheSIRSmodels reduceto SIS models when theimmunity
lossrateis zero All the stabilityresults ofSIRS models holdgoodfor SIS models providedthere is some inflow into thesusceptibleclass Wecanmake our model more realisticbythe introduction of a class ofindividualswhoare incubating thediseaseand taking intoaccount the fact thatimmunity may
onlyprovide temporary protection
REFERENCES
[1 ANDERSON, RM, The persistence ofdirect lifecycle infectious diseases withinpopulation of hosts, Lectures on Mathematics in the Life Sciences, Vol 12, providence, RI American
Mathemattcal Society(1979), 1-67
[2] ANDERSON, R M, JACKSON, H C,
MAY,
R M and SMITH, AM,
Population dynamics offox rabies inEurope, Nature289(1981),765-771
[3] ANDERSON,
R.M andMAY,
RM,
Infectious
Dtseasesof
Humans."Dynamtcs
andControl,OxfordUniversity Press(1991
[4]
BAILEY,
N.TJ,
The Mathematical Theoryof Infecttous
Diseasesand ttsApphcatlons, 2nded,GriffinandCo London
(1975).
[5] BRAUER,
F.,
Models for the spreadof universally fatal diseases,J. Math. BtoL 28(1990), 451-462[6] BRAUER, F, Models for the spread of universally fatal diseases II In Differential equation
models inbiology, epidemiology and ecology, S BusenbergandM Martellieds, Proceedings of the International Conference in Claremont, Jan 1990,Lect. NotesBtomath. 92, Springer-Verlag
(1991)
[7]
BUSENBERG, S andHADELER,
KP, Demographyand epidemics, Math. Btosct. 101 (1990),63-74
[8] BUSENBERG, S and VANDEN DRIESSCHE,P,Analysis ofadisease transmission model with
varying population size,J.Math. Biol. 29(1990),257-270
[9] DIEKMAN,
O and KRETSCHMANV,M,
Patterns in the effects of infectious diseases onGeometry, Centre of Mathematics and Computer Science, Stichting Mathematical Center,
Amsterdam 990)
[10] DIETZ, K andSCHENZLE,D,Proportionate mixing forage-dependentinfectiontransmission,J.
Math. Biol. 22(1985), 117-120
[1 !] GAO, L Q and HETHCOTE, HW, Disease transmission models with density-dependent
demographics, .I.Math Biol. 30(1992),717-731
[12] GREENHALGH, D, An epidemicmodel withadensity-dependent deathrate,
IMA
J.Math.Appi.Med. Biol. 7(1990), 1-26
[13] GREENHALGH, D, Vaccination m density-dependent epidemic models, Bull. Math. Btol. 54 (1992),733-758
[14] GREENHALGH,
D,
Some results foran SEIR epidemic model withdensity-dependence inthedeath rate,IMA.I.Math.Appi.Med. Biol.9(1992),67-106
[15] JORDAN, D W and SMITH, P, Nonhnear Ordinary
Dtfferenttal
Equatton,
Clarendon press, Oxford(1987)16] MAY, R M,StabthtyandComplexitymModel
Ecosystems,
PrincetonUniversity Press(1987) 17] McLEAN, A R andANDERSON,R.M,Measles indeveloping countries,Part Epidemiologicalparametersandpatterns,Epidemial.
Infection
100(1988), 111-133[18]
McLEAN, A R andANDERSON, R M, Measlesindeveloping countries,PartII: The predicted impactof massvaccination,Epidemtal.Infectton
100(1988),
419-442.19] MOLLISON, D., Sensitivity analysis of simple epidemic models, In Populatton
Dynamtcs
of
Wddhfe, C PJ
Bacon,
ed.(1985),223-234[20] NISBET, R.M and GURNEY, W.S.C., ModelhngFluctuatingPopulattons, NewYork. Wiley
(1982)
[21 PUGILESE,A,Populationmodels for diseases with norecovery,J. math. Biol.28(1990),65-82.
[22]
PUGILESE, A.,
An SEI epidemic model with varyingpopulation size, In: Differential equationmodels inbiology, epidemiology and ecology, S Busenbergand M Martellieds, Proceedings of
the International Conference in Claremont, Jan. 1990,Lect. NotesBiomath. 92, Springer-Verlag (1991)