Upload
ilri
View
147
Download
6
Tags:
Embed Size (px)
Citation preview
Participatory risk assessment III - Risk modelling II -
‘Learning Event’ on risk analysis and participatory methods
CSRS, November 28, 2014
Kohei Makita Associate Professor of Veterinary Epidemiology at
Rakuno Gakuen University (OIE Joint Collaborating Centre for Food Safety) Joint Appointment Veterinary Epidemiologist at International Livestock Research Institute (ILRI)
Outline
• Risk characterization – Modeling infection/ illness
– Separation of variability from uncertainty
– Sensitivity analysis
– Scenario analysis
• Importation risk assessment
• Beyond
2
Modeling infection/ illness
• Let’s have a look at examples of
– Campylobacter risk model for ready-to-eat beef
– Staphylococcal enterotoxin in milk
3
Encountering ‘a lack of bridge’
4
Encountering ‘a lack of bridge’
• When ‘dose-response relationship’ is not available
• When you would like to connect farm hygiene status and prevalence at the entrance of abattoir
P(A) = k P(B)
Connect probabilities by solving this coefficient
5
Example Human health impact of fluoroquinolone resistant Campylobacter
attributed to the consumption of chicken(FDA)
Coefficient here
6
KRes
𝐾𝑟𝑒𝑠 =Nominal mean number of campylobacteriosis cases from chicken
Estimated amount of fluoroquinolone resistant 𝐶𝑎𝑚𝑝𝑦𝑙𝑜𝑏𝑎𝑐𝑡𝑒𝑟 contaminated chicken meat consumed
7
Example Annual incidence rate of brucellosis due to consumption of milk in
Kampala, Uganda (Makita et al., 2010)
Coefficient here
Risk of purchasing contaminated milk
Annual incidence rate based on medical records
8
Outline
• Risk characterization – Modeling infection/ illness
– Separation of variability from uncertainty
– Sensitivity analysis
– Scenario analysis
• Importation risk assessment
• Beyond
9
Motivation
• The difference of variability and uncertainty needs to be recognized
– Variability • A function of the system
• Inter-individual variability
– Uncertainty • Lack of knowledge
• How much we don’t know because of the lack of survey
Separation of variability from uncertainty
• Method 1
– Model the total uncertainty first, and change all the variability parameters into single point estimates- the means – show the difference as distributions
– For a complex risk model
• Method 2
– Estimate uncertainty distributions for a series of point estimate variability, and show as a variety of uncertainty distributions
– For a model with a key variability
Staphylococcal poisoning example
12
• Each of them are uncertainty distributions
• The variety of uncertainty distributions shows variability
• Variability in this case is the growth speed of S. aureus
13
Risk of campylobacteriosis due to consumption of roast beef in Arusha, Tanzania
Outline
• Risk characterization – Modeling infection/ illness
– Separation of variability from uncertainty
– Sensitivity analysis
– Scenario analysis
• Importation risk assessment
• Beyond
14
Motivation
• We need to know:
– How much the result is reliable
– Which factor is the most sensitive because; • Data collection of the sensitive factor may reduce uncertainty the
most
• Control of the sensitive factor can reduce incidence the most
15
Sensitivity Tornado
-0.5 0
0.5 1
1.5 2
2.5
p / 1 to 2days G13
Cont rate B24
Boiling C24
p / Day 0 F13
1960 / Cont rate B11
1960 / Cont rate B16
p / 3 to 4 days H13
1960 / Boiling C16
1960 / Boiling C11
109/291 (Arcuri 2010
Temperature D10
N0 D4
Mean of Incidence rate
16
Sensitivity analysis
Prob. SA has SE genes
Prob. farmers boil
Prob. consumers boil
Contamination, farm
Store milk 3,4 days
Contamination, centre
Consume on day 0
Prob. centres boil
Contamination, farm
Store milk 1,2 days
Temperature
Initial bacteria population
*It provides efficient control options
How to perform the sensitivity analysis
• In @Risk, click the tab ‘@Risk’
• Click ‘Advanced analysis’
• Choose ‘Advanced sensitivity analysis’
• Choose the cell to monitor
• Choose the cell to analyze sensitivity- among uncertainty parameters
• Set 1000 iterations- larger value gives @Risk high work load
• See Tornado Chart or other charts
• Record 5th, 50th and 95th percentiles of the parameters in the report table
* Of course you can work out manually in R
Expressing in a table (Campylobacterosis example)
Rank Parameters Values with 50th, 1st and
99th percentiles
Mean annual incidence
per 1000 people
1 PCont 0.04 (0.004 - 0.15) 6.65 (0.59 – 23.02)
2 MPN 0.37 (0.1 - 1.27) 6.38 (1.72 - 20.55)
3 PIll|Infected 0.22 (0.11 - 0.38) 6.37 (2.95 - 10.21)
4 QCons (g) 604 (306 - 806) 6.59 (3.47 – 8.95)
5 NCustomer 4,327 (3,707 – 5,042) 6.15 (6.15 - 6.15)
18
Outline
• Risk characterization – Modeling infection/ illness
– Separation of variability from uncertainty
– Sensitivity analysis
– Scenario analysis
• Importation risk assessment
• Beyond
19
Motivation
• So far,
– Risk was assessed
– Uncertainty was quantified and presented
– Sensitive factor was identified
• The next interest for policy makers would be to know
– How much effective the possible control options are
– How much they cost and
– How much they benefit
Planning control options
• Risk communication between risk assessor and risk manager on;
– Planning control options
– Some detailed plan of each option
– Agreement of the options to monitor change
• Risk assessor may collect necessary information on cost-benefit analysis
– Estimation of costs and benefits
• Feasibility
• Negative impacts on livelihoods of consumers and value chain actors
21
Estimate reduction of risk
• Create duplicate branches of value chain in order to monitor the change
• Change inactivation parameters or quantities passing the duplicated value chains following the control option to monitor
• Model the division of risk after control measure is taken by the original risk: reduction rate of the risk
Sources of the risk by production areas Nakasongola 2.6%
Kayunga 4.1%
Peri-urban Kampala
Urban Kampala 15.9%
Mbarara 70.3%
7.0%
23
Sources of the risk by milk sellers
Without refrigerator 1.0%
Roadside vendor 1.1%
Bicycle 7.3%
Small refrigerator
12.8%
Farm gate 15.1%
Bulk cooler 62.7%
24
Control options (90% of enforcement)
Control options Reduction Inputs Feasibility Negative impact Assessment
Not to take any option 0.0 None High Risk remains Not recommendable
Construct a boiling centre
in Mbarara 62.3
A boiling centre,
legislation, fuel
Middle-
high Price up Recommendable
Construct boiling centres
in peri-urban Kampala 75.4
Boiling centres,
legislation, fuel Middle Price up Recommendable
Enforce milk shops to boil
milk or to buy boiled milk 68.9
Legislation, fuel,
facilities, enforce Very low
Price up, many shops
cannot afford Not recommendable
Ban of farm gate milk
sales 12.3
Legislation,
enforcement Low
Alternative sales may
not boil
Single measure does
not change the risk
Ban of urban dairy
farming 14.8
Legislation,
enforcement Middle
Livelihood of urban
farmers, milk supply Not recommendable
Ban of milk sales by
traders with a bicycle in
urban areas 6.6
Legislation,
enforcement High
Livelihood of traders,
alternative transport
may not boil
Single measure does
not change the risk
Ban of roadside milk sales 0.8 Legislation,
enforcement High
Livelihood of traders,
alternative transport
may not boil
Single measure does
not change the risk
Ban of milk sales at shops
without a refrigerator 0.8
Legislation,
enforcement High
Livelihood of traders,
alternative transport
may not boil
Single measure does
not change the risk
25
Example of cost-benefit analysis Tick eradication at Kuro-Shima Island, Japan
Contents of cost Calculation JPY
Acaricides Quantity x cost per unit 15,033,640
Labor cost Total working hours x labor cost/h 6,900,962
Construction of fence Number of fence used x cost per unit 1,479,177
Other Communication, transportation, accommodation
409,348
Total cost 23,823,127
Contents of benefit Calculation JPY
Increased growth Shortened grazing period x daily cost 25,742,103
Eradication of babesiosis
Treatment cost x incidence avoided + price of dead animal x death avoided
1,687,089
Other Costs which had been spent before erad. 3,953,407
Total benefit 31,382,599
Benefit-cost ratio=31,382,599/23,823,127=1.317 Yamane I (1996) Selection of economically the most ideal animal health disease control option 2. Rinsho-Juui14(5): 41-47. 26
Risk management (CAC/GL 63-2007)
• Principle 1: Protection of human health is the primary objective
• Principle 2: Taking into account the whole food chain
• Principle 3: Follow a structured approach
• Principle 4: Process should be transparent, consistent and fully documented
• Principle 5: Ensure effective consultations with relevant interested parties
• Principle 6: Ensure effective interaction with risk assessors
• Principle 7: Take into account of risks resulting from regional differences in hazards in the food chain and regional differences in available risk management options
• Principle 8: Decisions should be subject to monitoring and review and, if necessary, revision
27
ALOP
• ALOP: Appropriate level of protection (ALOP)
• Should be decided based on a particular country’s expressed public health goals
28
Outline
• Risk characterization – Modeling infection/ illness
– Separation of variability from uncertainty
– Sensitivity analysis
– Scenario analysis
• Importation risk assessment
• Beyond
29
30
OIE importation/exportation risk assessment (plus antimicrobial resistance)
Release assessment
Exposure assessment
Consequence assessment
Risk management
Risk assessment
Hazard identification
Risk communication
31
Is the disease
/agent exotic?
National
control
program?
Notably
lower
prevalence?
Could diseases be present in the animal of origin?
Can the agent/disease persist in the import?
No
Are there ways the agent could come in contact with
susceptible animals or people?
No
No No No
No
Yes
Yes Yes Yes
Yes
Yes
Hazard
iden
tification
Release assessment
• Assessing probability of agent introduction considering the every possible route coming into the specific environment
Country A Country B
Importation
Example: false negative in quarantine test
Probability of releasing the agent into country B with either animals or commodities
Exposure assessment
• Assessing probability that animals/humans are exposed to the agent released, through every possible route
Country A Country B
Importation
Probability that animals/humans originally in country B are exposed to the agents released
Consequence assessment • Assessing probability of outbreak/ epidemic and the economic loss,
as a result of exposure of animals/humans to the agent released
Country A Country B
Importation
Economic loss due to outbreak/epidemic
Outline
• Risk characterization – Modeling infection/ illness
– Separation of variability from uncertainty
– Sensitivity analysis
– Scenario analysis
• Importation risk assessment
• Beyond
35
Beyond of this course
• How to deal with multiple pathogens?
• Intervention using participatory methods
• Contribution of environment to food chain
• Another approach of disease control
36
Advantage of participatory risk assessment
I wish you prove it more in intervention programs!
• -Speed
• -Affordability
• -Flexibility in application
• -Understanding of culture
• -Best control option
• -Potential to change behavior
37
Infectious disease epidemiology
38
• Understanding disease transmission
• Control the disease based on its nature
• Starting with drawing epidemic curve
Example from Ebola (2014 Oct 16, Nature)
Infectious disease modelling
• Basic reproduction number (R0)
– Total number of individuals directly infected by a single infected individual, when introduced to totally susceptible population
– R0<1 Infection dies out
– R0=1 Infection is maintained
– R0>1 Infection takes over
R0 as a communication tool
• Example of Ebola epidemiology
– Nature (2014 Oct 16)
40
SIR model and calculation of R0
S I
S: Susceptible
I: Infectious
R
R: Recovered
SIR model
SIR model and calculation of R0
S I
dS/dt = -βSI
R βSI αI
dI/dt = βSI - αI
dR/dt = αI
Modelling deaths
S I
dS/dt = -βSI + μN - μS
R βSI αI
dI/dt = βSI – αI - μI
dR/dt = αI - μR
μS μI μR
μN
SIR model and calculation of R0
S I
βSI = (α + μ)I
R βSI αI
R (Effective reproductive ratio) = {β/(α + μ)}*S
μS μI μR
μN
β/(α + μ)*S = 1
SIR model and calculation of R0
S I
*In case all individuals are susceptible(S0 = N)
R βSI αI
R0 = {β/(α + μ)}*N
μS μI μR
μN
Force of infection
Infection dynamics in SIR model
Herd immunity
S I R
Vaccination
Herd immunity threshold =
0
11
R
*Epidemic does not occur above this level of immunity
Acknowledgements
• CSRS – Professor Bassirou Bonfoh
– Sylvain Traore
– Sylvain Koffi
– All other staff
• Swiss Tropical Public Health Institute
• And the participants!!
48