MATerials MATemàtics

MAT 2
Volum 2013, treball no. 3, 17 pp. ISSN: 1887-1097
Publicació electrònica de divulgació del Departament de Matemàtiques
de la Universitat Autònoma de Barcelona
www.mat.uab.cat/matmat

The SIR model

and the
Foundations of Public Health
Howard (Howie) Weiss

We introduce and analyze a basic transmission

model for a directly transmitted infectious disease. The
model consists of a system of three coupled non-linear
ordinary differential equations which does not possess
an explicit formula solution. However, simple tools
from calculus allow us to extract a great deal of infor-
mation about the solutions. Along the way we illustrate
how this simple model helps to lay a theoretical foun-
dation for public health interventions and how several
cornerstones of public health required such a model to
illuminate.
Sir Ronald Ross

1 Introduction
“As a matter of fact, all epidemiology, concerned as it is with the vari-
ation of disease from time to time or from place to place, must be
considered mathematically, however many variables as implicated, if it
is to be considered scientifically at all.”
Sir Ronald Ross, MD

We introduce and analyze the most basic transmission model for a di-
rectly transmitted infectious disease caused by bacteria, viruses, or fungi.
Direct transmission occurs through individual-to-individual contact: through
a sneeze or cough, through skin-skin contact, or through exchange of body
2 The SIR model and the Foundations of Public Health

fluids. Examples in the American media this week are the H7N9 flu, whoop-
ing cough, mumps, tuberculosis, and MERS-CoV.
The SIR (susceptible-infected-removed) model, developed by Ronald Ross1 ,
William Hamer, and others in the early twentieth century [4], consists of a
system of three coupled non-linear ordinary differential equations, which does
not possess an explicit formula solution. However, simple tools from calculus
allow us to extract a great deal of information about the solutions. Along the
way we illustrate how this simple model helps to lay a theoretical foundation
for public health interventions and how several cornerstones of public health
required a similar model to discover.
The SIR disease transmission model is derived assuming several strong
assumptions. There are hundreds of papers (and some books) where the
authors extend this basic model in many directions by relaxing some as-
sumptions. The mathematical analysis quickly becomes significantly more
sophisticated and in this article we focus on one of the simplest models.
We begin with a population that can consist of humans (e.g., a school,
hospital, or city), animals (e.g., a pig farm, bat colony, or deer in a forest),
or plants (e.g., spruce forest, sod farm, or wheat field). We then partition
the population into three groups or compartments: susceptible individuals,
infected individuals, and removed individuals. We denote the sizes of these
subpopulations at time t by S(t), I(t), and R(t). There are many assump-
tions behind the model, including a large and closed population, the outbreak
is short lived; no natural births or natural deaths occur, the infection has zero
latent period (an individual becomes infectious as soon as they become in-
fected), recovering from infection confers lifetime immunity, and mass-action
mixing of individuals. Mass action mixing assumes that the rate of encounter
between susceptible and infected individuals is proportional to the product
of the population sizes. Doubling the size of either population results in
twice as many new infections per unit time. This requires that the members
of both populations are homogeneously distributed in space and thus do not
mix mostly in any smaller subgroups. Intuitively, every person will encounter
every other person per unit time with equal probability. But keep in mind
that the SIR model is deterministic and there are no probabilities2 .
Is the mass action mixing assumption reasonable? Most humans have
contacts with only a small fraction of individuals in their community, and are
more likely to have contacts with family members, neighbors, and classmates.
Children typically have many more contacts than seniors. A common and
1
Sir Ronald Ross received the second Noble Prize in Medicine and Physiology for his
discovery of the transmission of malaria by the mosquito. He was also a closet Mathe-
matician and published papers in several areas of pure and applied mathematics.
2
One can formulate a stochastic analog of the SIR model as a Markov chain.
Howard (Howie) Weiss 3

pragmatic view is that one should begin modeling with the simplest model
and latter add more complexity, if required. The next step is sometimes to use
multiple classes of susceptible and infected individuals and assume well mix-
ing between these sub-classes with different rates. Models with well-mixing
can serve as null models against the influence of more detailed mechanisms.
The well-mixing hypothesis also allows the use of ordinary differential equa-
tions (ODEs) instead of partial differential equations or agent based models,
which can be substantially more difficult to parametrize, simulate, and ana-
lyze.

2 Methods
2.1 The SIR Model
The SIR model is the following system of quadratic ODEs:

dS
= −β S I (1)
dt
dI
=βSI −νI (2)
dt
dR
= ν I, (3)
dt

where the disease transmission rate β > 0 and the recovery rate ν > 0 (or in
other words, the duration of infection D = 1/ν).
The bi-linear incidence term β S I for the number of new infected indi-
viduals per unit time corresponds to homogeneous mixing of the infected and
susceptible classes. The total population size should remain constant, and
this easily follows from the SIR system: that the sum of the left hand sides
of the three equations is the derivative of the total population size and the
sum of the right hand sides is zero. We denote the total population size by
N . Since R(t) = N − S(t) − I(t), the system can be reduced to a system of
two ODEs: (1) and (2).
Suppose that each infected individual has κ contacts (each sufficient for
transmission) per unit time and κ is independent of the population size. Then
κ S/N of these contacts are with susceptible individuals. If the fraction τ of
adequate contacts result in transmission, then each infected individual infects
κ τ S/N susceptible individuals per unit time. Thus β = b/N where b = κ τ .
The parameter τ is called the transmissibility of the infectious disease.

MATerials MATemàtics

MAT 2
Volum 2006, treball no. 1, 14 pp.
Publicació electrònica de divulgació del Departa
de la Universitat Autònoma de Barcelona
www.mat.uab.cat/matmat
4 The SIR model and the Foundations of Public Health

2.2 Analysis of the SIR Model

2.2.1 Long term limits exist
Since the right hand side of (1) is negative and the right hand side of (3)
is positive, this implies that dS/dt ≤ 0 and dR/dt ≥ 0. Since 0 ≤ S(t) ≤
S(0) ≤ N and 0 ≤ R(0) ≤ R(t) ≤ N , this implies that the limits S(∞) =
limt→∞ S(t), R(∞) = limt→∞ R(t), and thus I(∞) = limt→∞ I(t) = N −
S(∞) − R(∞) exist.

2.2.2 The disease always dies out

It is also easy to prove that the disease always dies out, I(∞) = 0 for all initial
conditions, without having a formula for I(t). If not, (3) implies that for t
sufficiently large, dR/dt > νI(∞)/2 > 0, and this implies that R(∞) = ∞,
a contradiction.

2.2.3 Epidemic threshold theorem

We define the effective reproductive number Re = (S(0)/N ) b/ν and the basic
reproductive number R0 = b/ν. If the entire population is initially suscep-
tible, i.e., S(0) = N − 1, I(0) = 1, R(0) = 0, and large (recall this is a
model assumption), then Re = ((N − 1)/N ) b/ν is approximately equal to
R0 . Henceforth, to beautify formulas involving R0 , we will assume that the
quantity (N − 1)/N is equal to 1.
We now show that Re is the threshold value or tipping point that deter-
mines whether an infectious disease will quickly die out or whether it will
invade the population and cause an epidemic.
Theorem 2.1. 1. If Re ≤ 1, then I(t) decreases monotonically to zero as
t → ∞.

2. If Re > 1, then I(t) starts increasing, reaches its maximum, and then
decreases to zero as t → ∞. We call this scenario of increasing numbers
of infected individuals an epidemic.
It follows that an infection can invade and cause an epidemic in an entirely
susceptible population if R0 > 1 or b > ν.
Proof. Equation (2) and the discussion in Section 2.2.1 imply that dI/dt =
(β S − ν) I ≤ (β S(0) − ν) I = ν (Re − 1) I ≤ 0 for Re < 1. This observation
together with I(∞) = 0 (see Section 2.2.2) proves the first statement.
Equation (2) implies (dI/dt)(0) = ν (Re − 1) I(0) > 0 for Re > 1. Thus
I(t) is increasing at t = 0. Equation (2) also implies that I(t) has only one
Howard (Howie) Weiss 5

non-zero critical point. These observations, together with I(∞) = 0 imply

the second statement.

Figure 1 contains solutions of the SIR system simulating a highly virulent

(Re = 3.5) flu epidemic in a town of 50, 000 people.

4
x 10
5
S
4.5 I
R
4

3.5

2.5

1.5

0.5

0
0 10 20 30 40 50 60 70 80 90 100

Figure 1: Solutions of SIR system of ODEs with β = 0.7/50000, ν = 1/5, S(0) =

49955, I(0) = 5, R(0) = 0

We stress that the existence of a threshold for infection is far from obvious
and was missed by many public health and infectious disease experts. The
reason is that such a threshold can not be discerned from data; it requires a
mathematical model to illuminate.
Above we observed that (dI/dt)(0) = ν (Re − 1) I(0), which implies that
the number of infected individuals initially starts growing/decreasing expo-
nentially at rate ν (Re − 1). The next section will provide strong intuition
for the exponential growth.

2.2.4 Public Health interpretation of Re

We defined the effective reproductive number

S(0) β S(0) b D κ τ S(0)

Re = = = (4)
ν νN N

MATerials MATemàtics

MAT 2
Volum 2006, treball no. 1, 14 pp.
Publicació electrònica de divulgació del Departa
de la Universitat Autònoma de Barcelona
www.mat.uab.cat/matmat
6 The SIR model and the Foundations of Public Health

which is the product of the duration of infection, the number of contacts

an infected individual has with susceptible individuals per unit time, and
the transmissibility (rate of transmission). Thus Re is the number of new
infections caused by each infected individual at the beginning of the out-
break. The parameter Re is a measure of the fitness of the pathogen. With
this interpretation, the first theorem is almost obvious: if at the beginning
each infected individual infects three susceptible individuals, and each of
these three infected individuals infects three additional susceptible individ-
uals, then of course the number of infections starts growing exponentially.
This is schematically illustrated in the Figure 2.

Figure 2: The exponential growth of infected individuals at the beginning of an epidemic

2.2.5 The theoretical foundation of public health interventions

Theorem 2.1 and (4) provide strategies for public health experts to prevent
an epidemic by reducing Re to less than one. For example, for the flu:

1. Reduce the duration of infection D with antivirals;

2. Reduce the contact rate κ by self-isolation of susceptible individuals

(request that they stay at home and skip school or work);

3. Reduce S(0) by offering flu vaccines;

4. Reduce the transmissibility τ by encouraging frequent hand washing

and, in some cultures, distributing face masks.
Howard (Howie) Weiss 7

These strategies provide a theoretical underpinning for public health in-

terventions.

2.2.6 Vaccination and heard immunity

Vaccinating susceptible individuals removes them from the susceptible class.
Even if a vaccine is 100% effective (during a year when the flu vaccine matches
the circulating strains well, the annual flu vaccine is estimated to be about
60% effective), vaccinating an entire population is very expensive, and not
everybody can take the vaccine. Some individuals, such as those with com-
promised immune systems or severe allergies, the vaccine may be worse than
the disease. We ask the question, can an epidemic be prevented by vaccinat-
ing only a fraction of the susceptible class?
It easily follows from our analysis of the SIR model that the answer
is yes, and the phenomenon is called herd immunity. Recall, to prevent an
epidemic, we require that Re ≤ 1. Let ρ denote the fraction of the susceptible
individuals who gets vaccinated (assuming that the vaccine is 100% effective).
These ρ S(0) vaccinated individuals have moved from the susceptible class
to the removed class, and thus the size of the susceptible class becomes
(1 − ρ) S(0).
To prevent an epidemic, we require that

(1 − ρ) S(0) β/ν ≤ 1.

This will occur when ρ ≥ ρc , where ρc = 1 − 1/Re denotes the critical

vaccination threshold.
Thus vaccination against a disease can be completely effective without
making everyone immune.
The existence of a herd immunity threshold is also far from obvious and
was missed by many public health experts. A significant number of experts
thought that such a threshold did not exist and thus believed that mass
vaccination programs were bound to fail. The reason it was missed is that it
can not be discerned from data; it requires again a mathematical model to
illuminate.
Assuming a completely susceptible population, to prevent a flu outbreak
with R0 = 1.3, one must vaccinate “only” 23% of the population to prevent
an epidemic. If the vaccine is only 60% effective then it is easy to verify (do
it!) that one must vaccinate at least 23%/.6 ≈ 39% of the population to
prevent an epidemic.
To prevent a smallpox epidemic with R0 = 5, one must vaccinate 80%
of the population. Based in part on this finding, along with the belief that

MATerials MATemàtics

MAT 2
Volum 2006, treball no. 1, 14 pp.
Publicació electrònica de divulgació del Departa
de la Universitat Autònoma de Barcelona
www.mat.uab.cat/matmat
8 The SIR model and the Foundations of Public Health

humans are the only natural hosts of the smallpox virus, in 1967 the World
Health Organization (WHO) mounted a successful worldwide smallpox erad-
ication program. Smallpox is one of only a very small number of human
infectious diseases that has been almost completely irradiated around the
world. Even if a malaria vaccine that is 100% effective, since R0 > 100, it
would be necessary to vaccinate 99% of the population to prevent epidemics.

2.2.7 The maximum number of infected individuals

We now show that although we can not explicitly solve the SIR system of
ODEs, we can still obtain a formula solution for Imax .
Dividing (1) by (2) yields the ODE
dS −β S I
= .
dI βSI −νI
This ODE is separable, since for I > 0 it can be rewritten as
βS−ν
Z Z
dS = − dI.
βS
Hence, −I − S + ν/β log S = C. In other words, for every t ≥ 0,

I(t) + S(t) − ν/β log S(t) = I(0) + S(0) − ν/β log S(0). (5)

Imax occurs when dI/dt = 0 which from (2) yields occurs when S = ν/β.
Applying (5) yields

Imax + ν/β − ν/β log ν/β = I(0) + S(0) − ν/β log S(0)

or
Imax = I(0) + S(0) − ν/β log S(0) − ν/β + ν/β log ν/β.

An easy exercise is to show that for an initially fully susceptible popu-

lation, the maximum fraction of infected individuals is solely a function of
R0 :
1
Imax /N = 1 − (1 + log R0 ) .
R0
Expression (5) has another useful application. It states that the solutions
(S(t), I(t)) viewed in the S–I plane (orbits) are contained in the level curves
of the function F (S, I) = S + I − ν/β log S. The level curves of this function
are shown in Figure 3.
Howard (Howie) Weiss 9

Figure 3: Parametric plots of I(t) verses S(t) with β = 1.66 and ν = 0.44

2.2.8 Why do epidemics end?

Why do epidemics end? Do they end because there are no longer susceptible
individuals in the population? This question perplexed public health experts
for many years. If so, then it would be the case that S(∞) = 0. We now
show this is not true.

Proposition 2.2. The limiting number of susceptible individuals

S(∞) ≥ S(0) exp(−R0 ) > 0.

Proof. Dividing (1) by (3) yields the ODE

dS −β I S −β S
= = .
dR νI ν
This ODE is separable, since for S > 0 it can be rewritten as

−β
Z Z
1
dS = dR.
S ν

Integrating both sides yields

S(t) = S(0) exp(−β (R(t) − R(0))/ν), (6)

MATerials MATemàtics

MAT 2
Volum 2006, treball no. 1, 14 pp.
Publicació electrònica de divulgació del Departa
de la Universitat Autònoma de Barcelona
www.mat.uab.cat/matmat
10 The SIR model and the Foundations of Public Health

Since 0 ≤ R(t) − R(0) ≤ N it follows that S(t) ≥ S(0) exp(−β N/ν) and
thus
S(∞) ≥ S(0) exp(−β N/ν) = S(0) exp(−R0 ) > 0.3

As an epidemic proceeds, the number of susceptible individuals decreases

and so the rate at which new infections arise also decreases. Eventually, S(t)
drops below ν/β, and the rate at which individuals recover exceeds the rate
at which new infections occur. Thus, I(t) starts decreasing. The epidemic
ends because of the lack of new infected individuals and not because of the
lack of susceptible individuals.
This is still another fundament fact can not be discerned from data; it
requires a mathematical model to illuminate.

2.2.9 The size of an epidemic?

Equation (6) easily yields a transcendental equation for S(∞), which when
R(0) = 0, reduces to

S(∞)/N = (S(0)/N ) exp(−b (R(∞)/N )/ν)

= (S(0)/N ) exp(−b (1 − (S(∞)/N ))/ν). (7)

(recall from Section 2.2.2 that I(∞) = 0).

If the entire population is initially susceptible (7) has the following simple,
but transcendental form solely in terms of R0 :
   
S(∞) S(∞)
log = R0 −1 .
N N

We numerically solve this equation, and Figure 4 contains a plot of

S(∞)/N verses R0 .
It also follows in this case that the attack rate, the total fraction in
individuals who get infected, is 1 − S(∞)/N .
3
If simple models are pushed too far they may yield unreasonable results. If, for exam-
ple, R0 = 103 , then for almost any population the lower bound S(0) exp(−R0 ) < 1. In this
totally unrealistic case everybody becomes infected. However, for most infectious diseases
R0 ≤ 5, and for large populations (recall this is a model assumption), the conclusion is
meaningful.
Howard (Howie) Weiss 11

SH¥LN verses R0
1.0

0.8

0.6

0.4

0.2

0.0
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Figure 4: Plot of S(∞)/N verses R0 .

2.2.10 Five key epidemiological roles of the reproduction numbers

To recap, we have seen that Re plays the following five key epidemiological
roles:

1. Re is a threshold value for an epidemic: an epidemic will occur if Re > 1.

2. The initial exponential growth rate of an epidemic is (Re − 1)ν.

3. The critical vaccination threshold for herd immunity is 1 − 1/Re .

4. Assuming a fully susceptible initial population, the maximum fraction

of infected individuals satisfies Imax /N = 1 − (1/R0 ) (1 + log R0 ).

5. Assuming a fully susceptible initial population, the percentage of sus-

ceptible individuals S(∞)/N at the end of an epidemic is the root of
the transcendental equation log(S(∞)/N ) = R0 ((S(∞)/N ) − 1).

3 Testing the SIR model with data

We now present examples to illustrate the predictions of the SIR model. In
these examples the SIR model is parametrized using actual infection data,
and the I(t) output of the model is compared with the time series data. By
parametrizing, we mean that the parameters β and ν are estimated from the
data. We note that all the fundamental facts about epidemics that we have

MATerials MATemàtics

MAT 2
Volum 2006, treball no. 1, 14 pp.
Publicació electrònica de divulgació del Departa
de la Universitat Autònoma de Barcelona
www.mat.uab.cat/matmat
12 The SIR model and the Foundations of Public Health

so far obtained were discovered without parametrizing the model with any
data.
In practice, the transmission rate β, being a complicated function of con-
tact rate κ and transmissibility τ , is difficult to estimate directly and usually
requires data to estimate. However, the duration of infectiousness D = 1/ν
can usually be estimated independently of data (e.g., by measuring the virus
shedding of infected individuals over time).
The main model parametrization methods are the method of least squares,
the method of maximal likelihood, and Bayesian methods. There are other
statistical challenges when parametrizing models, for example, under-re-
porting. Not all infected individuals are officially counted. For example,
most people who have the flu suffer at home and do not visit a doctor and
have their diagnosis confirmed by a laboratory. One needs to estimate the
underreporting rate. At the beginning of the 2009 H1N1 flu pandemic the
under-reporting rate in the U.S. was estimated to be about 80 to 1 ([13]).

3.1 Dengue fever outbreaks

Dengue fever is a viral disease that is transmitted by mosquitoes. The
“Dengue triad" of fever, rash, and headache is characteristic of Dengue. The
disease commonly frequently found in the tropics and is endemic in about 100
countries. There are approximately 50 million cases of Dengue fever virus
worldwide. Global warming is causing the range of Dengue to spread.
Figure 5 shows the number of infected individuals during outbreaks in
Venezuela and Santiago de Cuba [5] along with the output of parametrized
SIR models. Figure 6 illustrates an attempt to model a Dengue outbreak in
Havana.

Figure 5: SIR model predictions of Dengue fever outbreaks in (a) Venezuela (2000) (b)
Santiago de Cuba (1997)
Howard (Howie) Weiss 13

Figure 6: SIR model prediction of 2001 Dengue fever outbreak in Havana .

3.2 Classical swine fever outbreak in the Netherlands

Classical swine fever or “hog cholera” is a highly contagious viral disease
of pigs and wild boar. Swine fever usually causes death within 15 days,
particularly in young animals. Figure 7 shows the number of infected swine
during an outbreak in the Netherlands during 1997-1998 [14] along with a
parametrized SIR model.

Figure 7: SIR model prediction of classical swine fever virus outbreak in the Netherlands

3.3 Norovirus in Brussels long-term care facility

Norovirus is a highly contagious viral disease and a major cause of gastroin-
testinal illness in closed and crowded environments, such as hospitals, nursing
homes and cruise ships. Typically, people with norovirus infection develop
diarrhea and abdominal pain and begin to vomit within 24 to 48 hours of

MATerials MATemàtics

MAT 2
Volum 2006, treball no. 1, 14 pp.
Publicació electrònica de divulgació del Departa
de la Universitat Autònoma de Barcelona
www.mat.uab.cat/matmat
14 The SIR model and the Foundations of Public Health

exposure. Norovirus symptoms may last a few days, but most people recover
completely without treatment.
Figure 8 shows the number of norovirus infections in a Brussels long-term
care facility during 2007 [16] along with a parametrized SIR model.

Figure 8: SIR model prediction of norovirus outbreak in Brussels long-term care facility
2007

3.4 How well do the models fit the data?

All the fits except Figure 6 look pretty good. But do not to be overly im-
pressed. The displayed data was used to parametrize the model and the
parametrized model may be just regenerating the data which was used to
parametrize it. More convincing would be if a parameterized model can
accurately reproduce new observed data.
Furthermore, a model can achieve an excellent fit to data for reasons
that have nothing to do with the model’s fidelity to the underlying biological
process. For an extreme example, in [12] we show that nearly any positive
function is the I(t) output of an SIR model with variable transmission rate
β(t). However, if the output of a model does not mimic the data (e.g.,
Figure 6), then the model is likely to be missing at least one important
component. In this sense, a deficient model may be more helpful than a
model that fits the data.

4 Concluding remarks
What have we learned? Mathematically, we learned how we can extract many
useful properties of solutions of nonlinear ODEs even though the system has
Howard (Howie) Weiss 15

no explicit formula solution. Biologically, we learned how the simple unpa-

rameterized SIR model yields several fundamental insights into outbreaks of
infectious diseases and their control - insights that were nearly impossible
to discern from infection data alone. Thus the analysis of the SIR model
provides a theoretical framework for public health interventions.
How else could compartment models be helpful to study disease trans-
mission? Here is a recent example from my research. The gray bars in Figure
9 represent the number of pandemic H1N1 viral isolates tested the United
States during the 2009 H1N1 flu pandemic [9]. Think of this as a proxy for
the number of infections.

Figure 9: Number of pandemic H1N1 viral isolates tested the United States from April
2009 through March 2010.

Notice the two peaks. If all the assumptions of the SIR model held for
this outbreak, Theorem 2.1 would imply there would be only one peak. Also,
the US experienced one peak every year except during the pandemic years
1918, 1958, 1968, 2009, when it experienced two or more waves. What is
the mechanism(s) that is generating the multiple peaks? Nobody knows.
Recently in [10], we used relatively simple extensions of the SIR model to
exhibit five plausible mechanisms, each of which could have generated the
two peaks during 2009, both quantitatively and qualitatively.
The first two mechanisms capture changes in virus transmissibility and
behavioral changes. The third mechanism involves population heterogeneity
where each wave spreads through one sub-population. The fourth mechanism
is virus mutation which causes delayed susceptibility of individuals. The fifth
mechanism is waning immunity. We use the models to examine the effects
of border control at the beginning of the outbreak and the timing of and
amount of available vaccinations.

MATerials MATemàtics

MAT 2
Volum 2006, treball no. 1, 14 pp.
Publicació electrònica de divulgació del Departa
de la Universitat Autònoma de Barcelona
www.mat.uab.cat/matmat
16 The SIR model and the Foundations of Public Health

We also use the models to try to understand why China, which instituted
strict border control at the onset of the outbreak, had only one peak and
the US had two peaks. The models indicate that had the US also instituted
strong border control at the onset of the outbreak, they would have also
experienced a single peak of infections. However, the models also indicate
that strong border control would not have decreased the total number of
infections.

Want to learn more about infectious disease transmission models? Some

references include [6, 7, 8, 11]. References for the mathematics (dynamical
systems or nonlinear dynamics) required to analyze these models include
[3, 15]. References [1, 2] contain introductions to stochastic transmission
models.

References
[1] Linda JS Allen. An introduction to stochastic processes with applications
to biology. Pearson Education New Jersey, 2003.

[2] Linda JS Allen. An introduction to stochastic epidemic models. In

Mathematical Epidemiology, pages 81–130. Springer, 2008.

[3] Kathleen T Alligood, Tim D Sauer, and James A Yorke. Chaos: an

introduction to dynamical systems. Springer, 1996.

[4] R.M. Anderson. Discussion: the Kermack-McKendrick epidemic thresh-

old theorem. Bulletin of mathematical biology, 53(1):1–32, 1991.

[5] José Luis Hernández Cáceres. Extracting useful information from

dengue incidence data. www.rcim.sld.cu/revista_13/articulos_
htm/caceres.htm.

[6] Herbert W Hethcote. Three basic epidemiological models. Applied Math-

ematical Ecology, 18:119–144, 1989.

[7] Herbert W Hethcote. The mathematics of infectious diseases. SIAM

review, 42(4):599–653, 2000.

[8] Herbert W Hethcote. The basic epidemiology models: Models, expres-

sions for r0, parameter estimation, and applications. Mathematical un-
derstanding of infectious disease dynamics, 16:1–61, 2009.
Howard (Howie) Weiss 17

[9] Michael A Jhung, David Swerdlow, Sonja J Olsen, Daniel Jernigan,

Matthew Biggerstaff, Laurie Kamimoto, Krista Kniss, Carrie Reed, Ali-
cia Fry, Lynnette Brammer, et al. Epidemiology of 2009 pandemic
influenza a (h1n1) in the united states. Clinical Infectious Diseases,
52(suppl 1):S13–S26, 2011.
[10] Anna Mummert, Howard Weiss, Li-Ping Long, Jose Amigó, and Xiu-
Feng Wan. A perspective on multiple waves of influenza pandemics.
PLOS ONE, 8(4), 2013.
[11] James Dickson Murray. Mathematical biology, volume 2. springer, 2002.
[12] Mark Pollicott, Hao Wang, and Howard (Howie) Weiss. Extracting the
time-dependent transmission rate from infection data via solution of
an inverse ode problem. Journal of Biological Dynamics, 6(2):509–523,
2012.
[13] Carrie Reed, Frederick J Angulo, David L Swerdlow, Marc Lipsitch,
Martin I Meltzer, Daniel Jernigan, Lyn Finelli, et al. Estimates of
the prevalence of pandemic (h1n1) 2009, united states, april–july 2009.
Emerging infectious diseases, 15(12):2004, 2009.
[14] Arjan Stegeman, Armin RW Elbers, Jan Smak, and Mart de Jong.
Quantification of the transmission of classical swine fever virus between
herds during the 1997–1998 epidemic in the netherlands. Preventive
veterinary medicine, 42(3):219–234, 1999.
[15] Steven Strogatz. Nonlinear dynamics and chaos: with applications to
physics, biology, chemistry and engineering. Perseus Books Group, 2001.
[16] J. Vanderpas, J. Louis, M. Reynders, G. Mascart, and O. Vandenberg.
Mathematical model for the control of nosocomial norovirus. Journal of
Hospital Infection, 71(3):214–222, 2009.

Mathematics Department
Georgia Institute of Technology
Atlanta, Georgia, USA
weiss@math.gatech.edu

Publicat el 1 de juliol de 2013

MATerials MATemàtics

MAT 2
Volum 2006, treball no. 1, 14 pp.
Publicació electrònica de divulgació del Departa
de la Universitat Autònoma de Barcelona
www.mat.uab.cat/matmat