Sie sind auf Seite 1von 56

Infectious disease epidemiology

&
Mathematical modeling
Hans-Peter Duerr & Martin Eichner
Institut fr Medizinische Biometrie
Universitt Tbingen

Berlin, 21. Sept. 2010

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 1 / 56

Program
Topic

Slide

Lesson 1

SARS 2002/2003: why modeling, and what is a


mathematical model? (the example of bacterial growth)

Exercise 1

Design a mathematical model yourself: the bacterial growth curve


(iterative solution of the bacterial growth curve with Excel)

11

Exercise 2

Solve the bacterial growth with Runge-Kutta software

17

Lesson 2

Deterministic models: SIR-model, theory, basic


reproduction number R0, epidemic vs. endemic case

19

Exercise 3

Simulate: epidemic according to the SIR model

29

Exercise 4

Think longterm: the influence of time and demography


(The example of measles as an endemic infection)

38

Lesson 3

Vaccination: final size of an epidemic, critical vaccination


coverage

41

Exercise 5

Predict: how many newborns to vaccinate?


(sensitivity analyses into the critical vaccination coverage)

44

InfluSim

56

AddOn

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 2 / 56

Literature selection

Infectious Disease Epidemiology: Theory and Practice. Neil Graham. Jones and Bartlett Publishers, Inc.
Mathematical Epidemiology of Infectious Diseases: Model Building, Analysis and Interpretation (Wiley
Series in Mathematical and Computational Biology), O. Diekmann and J. A. P. Heesterbeek.
Epidemic Models: Their Structure and Relation to Data (Publications of the Newton Institute). Denis Mollison
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 3 / 56

Lesson 1

Program

SARS 2002/2003:
why modeling, and what
is a mathematical model?
(the example of bacterial
growth)

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 4 / 56

Country
China
Hong Kong
Taiwan
Kanada
Singapur
Vietnam
USA
Philippinen
Deutschland

Cases Deaths Case fatality


5327
349
1755
296
665
180
251
41
238
33
63
5
33
0
14
2
9
0
8355

906

Amoy Gardens Block E

Example SARS 2003/2004

[%]
7
17
27
16
14
8
0
14
0

10.8%

For comparison Malaria: ~300 Mio. Cases / Year,


~1 Mio Deaths / Year (predominantly children)
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 5 / 56

Hotel Metropole

Amoy Gardens Block E

SARS, initial spread

CDC, taken from http://en.wikipedia.org


Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 6 / 56

Key: Basic reproduction number

Spread from Hotel Metropole

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 7 / 56

Key words: Model, Networks

Problem Global Networks

e.g. between
Chicago and New
York 25.000
Passengers per day

Passengers per day


L. Hufnagel et al. 2004, PNAS 101: 15124-9

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 8 / 56

Fig. 2. Global spread of SARS. (A) Geographical


representation of the global spreading of probable
SARS cases on May 30, 2003, as reported by the WHO
and Centers for Disease Control and Prevention. The
first cases of SARS emerged in mid-November 2002 in
Guangdong Province, China (17). The disease was
then carried to Hong Kong on the February 21, 2003,
and began spreading around the world along
international air travel routes, because tourists and the
medical doctors who treated the early cases traveled
internationally. As the disease moved out of southern
China, the first hot zones of SARS were Hong Kong,
Singapore, Hanoi (Vietnam), and Toronto (Canada), but
soon cases in Taiwan, Thailand, the U.S., Europe, and
elsewhere were reported. (B) Geographical
representation of the results of our simulations 90 days
after an initial infection in Hong Kong, The simulation
corresponds to the real SARS infection at the end of
May 2003. Because our simulations cannot describe the
infection in China, where the disease started in
November 2002, we used the WHO data for China.

L. Hufnagel et al. 2004, PNAS 101: 15124-9

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 9 / 56

Key word: Model prediction

Prediction

What is a mathematical Model?


Example bacterial groth: bacteria divide 2 times per hour.
Duration between divisions D = 0.5 hours.
Rate of division
l = 2 per hour

Logarithmic

Linear

l = 1/ D

Growth from one generation to the next:


Model:

Bi = 2 Bi -1

B(t ) = B0 2l t

Problems of this approach: - only valid for initial growth


- too simple for describing complex processes
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 10 / 56

Exercise 1:
preliminary considerations

Target: intuitive prediction


of quantitative relations

Draw into each graph the bacterial growth curve you would expect if
the culture was
started with 10000
bacteria, rather than
1 bacteria

Exercise 1

the generation
time of the bacterium
was not 0.5h but 1h
the beforementioned changes
occur simultaneously
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 11 / 56

Exercise 1 (File "00_bacterialGrowth.xls", sheet "generation time")

Exercise 1: iterative solution


of the model in Excel
Complete cells B2 to C32
in file "00_Bakterienwachstum.xls", spreadsheet "generation time"
=A2*tGen

=Bakt0

1. Parameter:
"Bakt0"

2. Parameter:
"tGen"

=A2*tGen

=C2*multFaktor

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

3. Parameter
"multFaktor"
Folie 12 / 56

Verify the preliminary considerations


of the previous slide

Exercise 1 (File "00_bacterialGrowth.xls", sheet "generation time")

Exercise 1
results

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling


Folie 13 / 56

Design models with differential equations


Example bacterial groth: bacteria divide 2 times per hour.
Duration between divisions D = 0.5 hours.
Rate of division
l = 2 per hour

l = 1/ D

"The speed by which the


total number of bacteria
B changes over time t"

Integration

"is proportional
to the individual
rate of division l
"

"and proportional
to the number of
bacteria reproducing
at time t "

dB(t ) ~
= l B(t )
dt
~
lt
B(t ) = const e
Total number of bacteria at time t

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 14 / 56

Derivative

Differential equations offer more

B(t )

1 K

"The growth rate


approaches zero when
the bacterial culture
approaches the value of
the capacity (B(t)=K)."

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Linear

Previously: the bacterial culture grows indefinitely (unrealistic in a finite world)


Now:
the culture cannot exceed a certain capacity K (realistic: test tube)

B(t ) =

Folie 15 / 56

B0 K
B0 + ( K - B0 )e

~
-l t

Summary

A mathematical Model is just a


mathematical way to describe a process.
Simple processes may be intuitively
described "by hand"
Differential equations are a useful tool to
describe complex processes.
Differential equations allow for describing
dynamic processes by means of
interpretable parameters.

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 16 / 56

Exercise 2:
preliminary considerations

Aim: intuitive prediction


of a dynamic process

Draw a qualitative
curve for the per
capita-reproduction
rate l and the number of bacteria over
time, B(t). Where is
the inflection point
of B(t)?

Number of bacteria B(t)

Fill in the numbers


for the upper and
lower bounds of
each axis into the
white boxes

Per capita-reproduction rate l

Exercise 2

Assume A) a bacterium which, under optimal conditions, reproduces 2 times per


hour (per capita-reproduction rate=2/h) and
B) a volume which can harbour at maximum 1,000,000 bacteria.

Time

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 17 / 56

Exercise 2 ("00_bacterialGrowth.txt")

Exercise 2:
First steps
Aim: first steps with modeling
software
Complete file
"00_bactGrowth.txt" with the
equations of the bacterial growth
curve (save your work), and specify
the initial value B(0).

Program here, using the


parameters listed beyond
{--- Parameters ---}

Copy the text into the program


editor of Berkeley-Madonna
Click "Run"
Make your sliders in Menu
Parameters|Define
Sliders...
Verify the preliminary
considerations of the previous slide
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 18 / 56

Lesson 2

Program

Deterministic models:
SIR-model, theory, basic
reproduction number R0,
epidemic vs. endemic
case

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 19 / 56

SIR-Model
Extensions of
the SIR-model:
SIRS
SEIR
SEIRS

Polio virus type 1

Common way of
representing a model:
Compartiments
& Transititions

Susceptible

Infectious

immune

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 20 / 56

Information needed
Durations:

latent and infectious period

Rates:

contact rate(s)

Probabilities:

P(infection | transmission)

Demography:

birth and death rate,


age structure of the population

Disease:

Proportion of inapparent infections

....
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 21 / 56

Dynamic description: Birth


Birth of (susceptible) individuals

S dS(t) / dt = m

[ S(t)+I(t)+R(t) ]

=1

R
S Proportion susceptibles

m per capita birth rate

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 22 / 56

Dynamic description: Infection


new Infections

S dS(t) / dt = m - bc I(t) S(t)


I

dI(t) / dt = bc I(t) S(t)

R
S Proportion susceptible

m Per capita birth rate

I Proportion infectious

b Contact rate

R Proportion immune
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 23 / 56

P (infection | contact)

Mass action law


The probability of encounterings between
susceptible and infectious individuals
depends on:
the contact rate b ("Temperature")
the ratio Susceptible : Infectious

1
0.8
P

S*S
2(S*I)
I*I
Sum

0.6
0.4
0.2
0

Susceptible

Infectious

0 0.2 0.4 0.6 0.8 1


Proportion susceptible

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 24 / 56

Dynamic description: Infection


new infections

S dS(t) / dt = m - bc I(t) S(t)


I

dI(t) / dt = bc I(t) S(t)

R
S Proportion susceptible

m Per capita birth rate

I Proportion infectious

b Contact rate

R Proportion immune
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 25 / 56

P (infection | contact)

Dynamic description: Loss of infection


Loss of infectiousity

S dS(t) / dt = m - bc I(t) S(t)


I

dI(t) / dt = bc I(t) S(t) - g I(t)

R dR(t) / dt = g I(t)
S Proportion susceptible

m Per capita birth rate

I Proportion infectious

b Contact rate

R Proportion immune

g rate of loss of infectiousity

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 26 / 56

P (infection | contact)

Dynamic description: Mortality


Mortality

S dS(t) / dt = m - bc I(t) S(t) - m S(t)


I

dI(t) / dt = bc I(t) S(t) - g I(t) - m I(t)

R dR(t) / dt = g I(t) - m R(t)


S Proportion susceptible

m Per capita birth rate

I Proportion infectious

b Contact rate

R Proportion immune

g rate of loss of infectiousity

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 27 / 56

P (infection | contact)

choose parameter values for b, c, g and m


choose initial values for S(0), I(0), and R(0)

First iteration (time = 0)


compute for a short time step D the changes
dS(0)/dt, dI(0)/dt and dR(0)/dt
extrapolate changes to S(0+D), I(0+D) and R(0+D)

Following iterations (time = t)


1. compute for a short time step D the changes
dS(t)/dt, dI(t)/dt and dR(t)/dt
2. extrapolate changes to S(t+D), I(t+D) and R(t+D)
t=t+D, goto 1
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 28 / 56

www.uni-tuebingen.de/modeling/Mod_Pub_Software_SIR_en.html

Initialisation

A ready-to-use software of the SIR model is available from

Numeric solution of the model

Exercise 3:
preliminary considerations

Aim: understanding the role


of the parameters in the SIR
produces qualitatively a graph like

Zeit

Draw your qualitative prediction into the graph on how the course of the
epidemic would change (higher, faster, slower, etc) if
the contact rate between
people increases? (b )

patients recover more


rapidly? (g )

Time

b and g increase at the


same time
No. of infectous people

No. of infectous people

No. of infectous people

Exercise 3

dS(t) / dt = m - bc I(t) S(t) - m S(t)


dI(t) / dt = bc I(t) S(t) - g I(t) - m I(t)
dR(t) / dt = g I(t) - m R(t)

Zahl der
Infizierten

The SIR model, defined as,

Time

Time

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 29 / 56

Exercise 3 (File "10_SIR.txt")

Exercise 3: solving differential equations numerically


Aim: quantitative
epidemiology of
infectious diseases
learning by doing
Complete file
"10_SIR.txt" with the
equations of the SIRmodel (save your work),
and specify the initial
values (INIT S, I, R).

Program here, using


the parameters listed
beyond
{--- Parameters ---}

Copy the text into the


program editor of
Berkeley-Madonna
Click "Run"
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 30 / 56

Make your sliders in Menu


Parameters|Define Sliders...

Exercise 3 (File "10_SIR.txt")

Exercise 3:

Aim: Performing a
sensitivity analysis

and verify your preliminary considerations of the previous slide


Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 31 / 56

R0: Basic reproduction number

Wichtig!

Wichtig!

Wichtig!

Average number of secondary infections


which one infectious individual would cause
in a fully susceptible population
Definition: R0 = b c D
R0>1: Infection can persist;
an endemic state ist possible
R0<1: Infection cannot persist; goes extinct
D = 1 / (g+m) average duration of the infectious period
bc

Number of (sufficiently close) contacts per unit of time

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 32 / 56

R0 for some infectious diseases

Disease
Measles
Pertussis
Mumps
Rubella
Polio
Diphteria

Average
age at
infection
[years]
5.0
4.5
7.0
10.2
10.4
10.4

R0
15.6
17.5
11.5
7.2
6.1
6.1

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Critical
vaccination
coverage
pcrit [%]
94
94
91
86
84
84
Folie 33 / 56

Epidemic
SIR Model; without births and deaths
bc = 0,5/day, g = 0,1/day, m = 0/day R0 = 5

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 34 / 56

Epidemic
SIR Model; without births and deaths
bc = 0,2/Tag, g = 0,1/Tag, m = 0/Tag R0 = 2

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 35 / 56

At the end of an Epidemic...


... susceptible individuals may remain

Susceptible

Infectious

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Resistent

Folie 36 / 56

Proportion S susceptible at the end of the epidemic

Proportion susceptible

- log (S) = R0 (1 - S)

0.8
0.6
0.4
0.2
0
1

2
3
4
Basic reproduction number R0

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 37 / 56

Exercise 4 (File "11_SIRreparameterized.txt")

Exercise 4:
endemic infection

Aim: Understanding the


influence of demography

Make sure that slider settings in file "11_SIRreparameterized.mmd"


are as follows
Parameter

Minimum

Use

Maximum

STOPTIME

100

5000

DT

0.1

DTOUT

10

iniInfected

0.0001

lifeExpectYears

50

100

durationInfected

10

20

R0

15

20

For R0=15 (Measles-like) simulate an extended period of time by increasing STOPTIME from 100 to 5000 days (=13.7 years)

What is the reason for


recurrent epidemics?

Simulate a population with a lower life expectancy (developing countries) by decreasing lifeExpectYears from 50 to 30 years.

Why is the time between


epidemics reduced?

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 38 / 56

Endemic case
SIR Modell with demography
bc = 0,5/day, g = 0,1day, m = 0,0005/day R0 = 5

Proportions

1
0.8
suszeptible
infizierte
immune

0.6
0.4
0.2
0
0

1000
2000
Time [days]

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

3000

Folie 39 / 56

Summary
Neglecting births and deaths,
we model an epidemic scenario;
after the epidemic, a proportion of susceptibles,
which depends on R0, remains

Considering births and deaths,

logKurve

we model an endemic scenario


the model variables (S,I,R, ...) approach the
endemic state (if R0 > 1);
the equilibirum prevalence depends on R0.
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 40 / 56

Lesson 3

Program

Vaccination:
final size of an epidemic,
critical vaccination coverage

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 41 / 56

Dynamic description: SIR without vaccination

Model without vaccination

S dS(t) / dt = m - bc I(t) S(t) - m S(t)


I

dI(t) / dt = bc I(t) S(t) - g I(t) - m I(t)

R dR(t) / dt = g I(t) - m R(t)


S Proportion susceptible

m Per capita birth rate

I Proportion infectious

b Contact rate

R Proportion immune

g rate of loss of infectiousity

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 42 / 56

P (infection | contact)

Dynamic description: SIR with vaccination

A proportion of newborns will be vaccinated

S dS(t) / dt = m ........
(1-p) - bc I(t) S(t) - m S(t)
I

dI(t) / dt = bc I(t) S(t) - g I(t) - m I(t)

m p + g I(t) - m R(t)
R dR(t) / dt = .......
S Proportion susceptible

m Per capita birth rate

I Proportion infectious

b Contact rate

p Proportion vaccinated

R Proportion immune

g rate of loss of infectiousity

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 43 / 56

P (infection | contact)

Exercise 5 (File "11_SIRreparameterized.txt")

Exercise 5: Vaccination

Aim: Understanding the


concept of thresholds

Implement parameter "p" for the proportion of vaccinated newborns (see


previous slide) in the equations of file "11_SIRreparameterized.txt"
file and save it as "12_SIRvaccination.txt"
Define slider for "p" in
Menu
Parameters|Define
Sliders... and choose
slider settings as shown in
the screenshot on the right

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 44 / 56

Exercise 5 (File "12_SIRvaccination.txt")

Exercise 5: Vaccination
Technical remark:

Aim: Understanding the


concept of thresholds

For purposes of better inspectation, we change in the


output window axis settings for compartment I to "Auto"
see below and ask the lecturer.

1.

2.

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 45 / 56

Aim: Understanding the


concept of thresholds

Exercise 5 (File "12_SIRvaccination.txt")

Exercise 5: Vaccination
For R0=15, increase
p up to a value when
there is no epidemic
anymore. This is the
ciritical vaccination
coverage p*. Repeat
the procedure for
R0=10, 5 and 2, and
plot your results in
the graph to the right.
What is the critical
vaccination coverage
when the basic reproduction number tends
to values of R01

p*
1
0.8
0.6
0.4
0.2
0

R0
0

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

10

Folie 46 / 56

15

Endemic equilibrium
no change of model variables in the endemic equilibrium

0 = m (1-p) - bc I(t) S(t) - m S(t)

0 = bc I(t) S(t) - g I(t) - m I(t)

0 = m p + g I(t) - m R(t)

S Proportion susceptible

m Per capita birth rate

I Proportion infectious

b Contact rate

p Proportion vaccinated

R Proportion immune

g rate of loss of infectiousity

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 47 / 56

P (infection | contact)

Endemic equilibrium
no change of model variables in the endemic equilibrium

0 = m (1-p) - bc I S - m S

0 = bc I S - g I - m I

0 =mp+gI-mR

I=...

S=...

R=...

S Proportion susceptible

m Per capita birth rate

I Proportion infectious

b Contact rate

p Proportion vaccinated

R Proportion immune

g rate of loss of infectiousity

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 48 / 56

P (infection | contact)

Endemic equilibrium
Estimate R0 from the
proportion of
susceptibles in the
endemic equilibrium

R0 = bc / (g+m)

S = (g + m) / (bc) = 1 / R0

R = 1-S-I

= (1 - 1/R0 - p) m / (g + m)

S Proportion susceptible

m Per capita birth rate

I Proportion infectious

b Contact rate

p Proportion vaccinated

R Proportion immune

g rate of loss of infectiousity

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 49 / 56

P (infection | contact)

Critical vaccination coverage


Parameters of the right hand side are known, except p

S
I

= (1 - 1/R0 - p) m / (g + m)

R
I Proportion infectious

m Per capita birth rate = death rate

Basic reproduction number:


R0 = bc / (g+m)

g rate of loss of infectiousity

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

p Proportion vaccinated

Folie 50 / 56

Critical vaccination coverage


Parameters of the right hand side are known, except p

S
I
R

0 = (1 - 1/R0 - pcrit) m / (g + m)
pcrit = 1 - 1 / R0 =1-S
To eliminate a disease, it is not necessary to
vaccinate the whole population
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 51 / 56

Critical vaccination coverage


1

Proportion vaccinated

Elimination
0.8
0.6

Persistence

0.4
0.2
0
0

10

12

14

Basic reproduction number R0


Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 52 / 56

16

18

20

Summary
The proportion of susceptibles in the endemic
equilibrium does not depend on the proportion p of
vaccinated children
Transmission stops if p pcrit
The critical vaccination coverage is

pcrit = 1 - 1/R0

The model can be used for sensitivity analyses into the


effects of different vaccination strategies:
- What is the critical vaccination coverage?
- How does vaccination impact on the prevalence and incidence of the
infection?
- what is the best vaccination strategy (e. g. ring vaccination vs. mass
vaccination)?
Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 53 / 56

Estimation of model parameters


m

Per capita birth rate = death rate


1 / m is the life expectancy

Loss-of-infection rate
1 / (g+m) is the average duration of the infectious period

R0

Basic reproduction number


1 / R0 is the endemic prevalence of susceptibles

pcrit

Critical vaccination coverage


pcrit = 1 - 1 / R0

bc

Effective contact rate


bc = R0 (g+m)

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 54 / 56

Vergleich:
deterministische vs. stochastische Modelle

Wichtig!
Wichtig!
Wichtig!

Deterministische Modelle

stochastische Modelle

werden i.d.R. durch explizite


Formeln (Differenzialgleichungen)
erstellt

werden i.d.R. durch (individuenbasierte) Simulationsprogramme


erstellt

liefern bei gleichen Anfangsbedingungen stets identische


Ergebnisse

liefern bei gleichen Anfangsbedingungen zufallsbedingt


unterschiedliche Ergebnisse

ihre Ergebnisse sind meist besser


verallgemeinerbar Zur Planung
von Interventionsmanahmen sind
deterministische Modelle oft besser
geeignet

ihre Ergebnisse sind meist


realittsnher da sie zufllige
Effekte wiedergeben knnen
(Stochastizitt)
Fr die Untersuchung von Effekten
in kleinen Populationen besser

Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 55 / 56

AddOn: InfluSim

free download: www.influsim.de


Hans-Peter Duerr, University of Tuebingen, www.uni-tuebingen.de/modeling

Folie 56 / 56

Das könnte Ihnen auch gefallen