Modelling tools and methods for decision support on contagious animal diseases
Transcript of Modelling tools and methods for decision support on contagious animal diseases
Monique Mourits and Helmut Saatkamp
Modelling tools and methods for decision
support on contagious animal diseases
Monique Mourits and Helmut Saatkamp
Wageningen University – Business Economics
Aim of presentation
Aim:
� To present a short overview of some modelling tools and
methods for decision support in contagious livestock
diseases
� Emphasis: � Emphasis:
� Usefulness, possibilities
� Limitations
� NOT: complete methodological overview
Regulations contagious livestock diseases
� Aim/goal of decision makers: minimization of disutility
(discomfort, impact, …)
� by considering the impact on various stakeholder groups
affected (whole society): producers, agri-business,
consumers/citizens, region/country
� Decision making criteria:
� Until 1980s: veterinary
� Late 1980s-early 1990s: financial-economic
� Late 1990s: animal welfare, socio-ethical
� Late 1990s-early 2000s: human health
Regulations contagious livestock diseases
� Choice of strategies; how?
� Sets of measures (discrete alternatives)
� Prevention, monitoring and control
� Satisfies the goals/aims/preferences
� Within given constraints: legal, resources, …� Within given constraints: legal, resources, …
� Complex decision making problem:
� Need for decision support
Models for decision support
All models are wrong, but some are useful
(Zeger, S.L. (1991) Statistical reasoning in epidemiology. American Journal of
Epidemiology 134: 1062-1066)
Usefulness criterion: improvement of decision making with
use of models (compared to without use):
� Insight, understanding (qualitative)� Insight, understanding (qualitative)
� Model output (quantitative)
Models: a tool to support decision making by providing
insight and useful information, as an input for improved
reasoning, aiming at taking better decisions
Models for decision support
Models:
� Conceptual models:
� Qualitative relationships of a system
� Very important, usually underestimated
� Use: before AND after mathematical modelling� Use: before AND after mathematical modelling
� Mathematical models
� Quantitative relationships, equations and data
� Simulation (what-if), particularly if empirical research is
not possible or very expensive
Models for decision support
‘Pure’ epidemiological modeling of livestock diseases:
� Aimed at understanding the disease dynamics
� ‘Part’ of the ‘real world’, scientifically oriented, detailed
Economic modeling of livestock diseases:Economic modeling of livestock diseases:
� Aimed at decision support (discrete alternatives)
� Holistic, comprehensive, integration of ‘all required’
knowledge (data problems)
� Bridge between mono-disciplinary approaches
Two examples
Two model studies performed to evaluate;
� The probability of introduction of CSFV and the
impact of preventive measuresimpact of preventive measures
� The impact of various FMD control strategies
Models for decision support 1:
introduction/prevention
Aim of the model:
� Estimation of probability of introduction of CSF in The
Netherlands
Analysis of the main risk factors for introduction of CSF in � Analysis of the main risk factors for introduction of CSF in
The Netherlands
� Analysis of the impact of prevention measures
Models for decision support 1:
introduction/prevention
Methodology:
� Pathway diagram for introduction of CSF
� Excel/@Risk model
Data: statistical, epidemiological, expert opinion� Data: statistical, epidemiological, expert opinion
� Scenario tree analysis
� Sensitivity analysis
Models for decision support 1:
introduction/prevention Pathway diagram:
Pathways
for virus
introduction
into a region
Effect of
preventive
measures
Endogenous pathwaysExogenous pathways
Import of batch of
domestic animals
At least 1
animal infected/
contaminated
Illegal import
of live animals
At least 1
animal infected/
contaminated
Import of
wild animals
At least 1
animal infected/
contaminated
Virus not
detected by
screening/testing
Returning
livestock trucks
Virus contaminated
livestock truck
Livestock truck
not disinfected
properly
Wildlife
Infected aerosols
Air currents LaboratoriesIllegal import of
animal products
Virus contaminated
animal products
Virus not
inactivated during
processing/maturing
Import of animal
products for
human consumption
Humans carrying
virus
Tourists Professional
people
Harbours,
airportsImport of
genetic material
Virus
contaminated
material
Virus sources
Virus not
detected by
screening/testing
Virus contaminated
kitchen offal
At least 1
animal infected
Illegal import of
genetic material
Import of
manure
Virus contaminated
manure
Virus not
inactivated during
transport/treatment
Birds, pets,
arthropods
and rodents
Virus contaminated
animalsInfection or
contamination
with virus
Virus
contaminated
material
Genetic material
not detected by
custom controls
Animals not
detected by
custom controls
Virus contaminated
animal products
Products not
detected by
custom controlsmeasures
Destination
for life
Destination
for slaughter
Swill
Virus not inactivated
by heat treatment
Infective dose
PRIMARY
OUTBREAK !
screening/testingproperly
Direct or indirect contact with
susceptible domestic animals
processing/maturing
Products fed as swill
screening/testing
Main routes of
virus transfer
to susceptible
domestic animals
transport/treatmentcustom controls custom controls custom controls
Models for decision support 1:
introduction/preventionMatrix for target country:
� Pathways (n)
� Countries of origin (m)
Split-up of n * m scenario trees
Each scenario tree: contribution to PCSF
Totalization of all contributions: total PCSF
Models for decision support 1:
introduction/prevention
0.03
0.035
0.04
0.045
Pro
ba
bility
of C
SF
V in
tro
du
cti
on
in
to th
e N
eth
erl
an
ds
Types of result : Probability of CSFV introduction into NL per
epidemic and per year from various countries of origin
0
0.005
0.01
0.015
0.02
0.025
Ger
man
y
Franc
e
Italy
Bel
gium
Luxe
mbo
urg
Uni
ted
Kin
gdom
Irela
nd
Den
mar
k
Gre
ece
Spa
in
Por
tuga
l
Aus
tria
Finla
nd
Sw
eden
Country of origin
Pro
ba
bility
of C
SF
V in
tro
du
cti
on
in
to th
e N
eth
erl
an
ds
Per epidemic
Per year
Models for decision support 1:
introduction/prevention Types of result: Relative contribution of pathways to the
annual probability of CSFV introduction in the NL
Pork products
1.3%Piglets
6.6%
Breeding pigs
Livestock trucks
64.8%
Breeding pigs
17.6%
Fattening pigs
9.7%
Models for decision support 1:
introduction/prevention
Main conclusions:
� careful use of absolute values (estimated Pcsf)
� indication importance of countries of origin and pathways
� priority setting strategic preventive measures
� additional tactical preventive measures� additional tactical preventive measures
Useful in Benefit/Cost-studies on prevention measures;
how to achieve considerable risk reduction at reasonable
costs?
Models for decision support 2: impact/control
Aim:
� Analysis of the social –economic consequences of FMD
outbreaks under various disease control strategies
� What-if simulation
� Uncertainty� Uncertainty
Models for decision support 2: impact/control
Methodology:
� Integrated epidemiological-economic modeling
� Epidemiological modeling (InterSpread approach):
� Spatial: exact farm location and characteristics
� Stochastic (Monte Carlo): disease spread and disease � Stochastic (Monte Carlo): disease spread and disease
control mechanisms
� Dynamic: parameter values can change with time
Models for decision support 2: impact/control
� Output: descriptive statistics of the simulated epidemic:
� # farms infected/detected
� # farms (pre-emptively) culled
� # farms vaccinated
� Duration (days)
� Size of Surveillance zone
� Output: input for economic analysis
Models for decision support 2: impact/control
� Economic analysis:
� Direct and direct consequential costs (accounting of
the epizootic): Excel, accounting
� Indirect consequential costs (price effects):
• Supply and demand of animals and products• Supply and demand of animals and products
• Partial-equilibrium modeling: producers and
consumer surpluses
• Mathematical programming
What to expect……….
when a FMD epidemic starts at a dairy farm surrounded by
240 farms in a radius of 10 km (SPLA)
1923 farms in a radius of 10 km (DPLA)1923 farms in a radius of 10 km (DPLA)
and controlled by EU compulsorymeasures?
Basic results
SPLA (A) DPLA (B)
50% 95% 50% 95%
# infected farms 3 38 1583 23621# preventively culled 8 49 679# preventively culled 8 49 679
1053# farms in MCZ (10km) 335 4406 12895 58165
Epidemic length (days) 38 105 375 end
Alternative strategies DPLA – 95% results
EU Prev-1 Vacc-2 Vacc-4
Farms
depopulated 24.674 2.425 178 169
vaccinated 0 0 1.210 2.519
in MCZ 58.165 10.484 8.478 7.808in MCZ 58.165 10.484 8.478 7.808
Animals (*1000)
culled 6.211 463 43 40
vaccinated 0 0 284 541
Length (days) > 400 200 84 75
Alternative strategies DPLA – 95% results
(in mln €)
EU Prev-1 Vacc-2 Vacc-4
‘to live’ ‘to kill’ ‘to live’ ‘to kill’
Control >4.620 421 94 174 80 235
Business Business interruption > 2.540 200 166 121 171 111
Market losses > 499 511 623 467 636 477
Total >7.659 1.132 883 762 886 824
Models for decision support 2: impact/control
Main conclusions: cons
� Model requires a lot of specific input
� Contact structure
� Farm location, type and numbers per animal type (area specific??)� Farm location, type and numbers per animal type (area specific??)
� Epidemiological data scarce; use of expert opinions
� Sensitivity analyses a must (ranking alternatives)
Models for decision support 2: impact/control
Main conclusions: pros
� Simulation realistic:
� farm data + location, spread mechanisms, control
measures, etc.measures, etc.
� Very flexible to include uncertainty
� Helpful to identify knowledge gaps
� Helpful for decision support (in peace time)
Models for decision support 2: impact/control
Main conclusions: pros
� Helpful for decision support (in peace time);
� Control-driven; capacity problems, definition of risk areas.....
� Allocation of resources; trade off possibilities between prevention,
monitoring and control
� Evaluation of epidemic disease risk financing instruments
� Evaluation of economic instruments to reduce market disruptions
� Evaluation harmonization of cross border regulations
� ....
Models for decision support
� Probability of introduction of CSF and impact of preventive measures:
De Vos et al. (2004) Risk Analysis 24: 237-253
� Effectiveness of movement-prevention regulations to reduce the
spread of foot-and-mouth disease in The Netherlands: Velthuis and
Mourits (2007) Preventive Veterinary Medicine 82 (3-4): 262-281
� Consequences of CSF and control measures: Mangen et al. (2003)
European Review of Agricultural Economics 30(2): 125-154
� Multi Criteria Decision Making to evaluate control strategies of
contagious animal diseases: Mourits et al.(2010) Preventive
Veterinary Medicine 96 (3-4): 201-210.