Upload
ethen-bullington
View
216
Download
0
Tags:
Embed Size (px)
Citation preview
Risk based optimization of the Danish pork Salmonella control program
H. Scott Hurd[1], Claes Enøe[2], Lene Sørensen[2], Henrik Wachmann[2], Steven M Corns[3], Kenneth M. Bryden[3], Helen H. Jensen[4]
[1] College of Veterinary Medicine, Iowa State University[2] Danish Meat Association, Copenhagen, Denmark
[3] Mechanical Engineering Department, Iowa State University[4] College of Agriculture, Dept of Economics, Iowa State University
AcknowledgmentsTine Hald, Matthias Greiner, Rene Bøker, Anne Bisp Lind
– Danish Institute for Food and Veterinary Research (DFVF)
Funded as a project of the International EpiLab – Guest Scientist at the EpiLab in Denmark, – Employed at the National Animal Disease Center:
USDA:ARS, Ames, Iowa, USA
Overview
• Evaluation of the on-farm pork Salmonella control program
• Has the last 10 years of investment significantly reduced human cases?
• Simulation/systems model with best data says……
U.S.
• 100.3 M pigs/year
• 71,000 producers
Iowa
• 144,700 km2
• 28.8 M pigs/year
Denmark
• 43,000 km2
• 23 M pigs/year
• 13,200 producers
Denmark is a small state with a UNIQUE form of pork production and marketing (Coop)
Context for this project
• “We are spending too much for on-farm surveillance”
• “Carcass decontamination is the cheapest way to control salmonella”
• “Washing all carcasses, not acceptable. Don’t want to eat feces”
• “The Level 3 herds are not important”• “Pressure on the L3 herds has improved all herds”• “Yeah, but has it improved public health?”• “The industry just cares about “cheaper”• “We are at a standstill in reducing carcass swab
prevalence”
BACKGROUNDThe current Salmonella herd
classification scheme
• Serological monitoring of all herds in the “coop” system – Form of a blood test (serum exudate, meat
juice)– (up to 60 pigs/herd/year)
• Assigned to 3 prevalence levels based on last 3 months of serology
The current Salmonella herd classification scheme
• High seroprevalence (~1.5%) assigned – to special hygienic slaughter conditions, – financial penalties, – and hot water decontamination
• Carcass swab (400cm2)– Pool of 5 samples– cultured for Salmonella– Similar to USDA:FSIS
Objectives
• Estimate the effect of various control interventions– On-farm (pre-harvest) vs.– Abattoir (post-harvest)
• Overtime – Previous 10 years (retrospective)– Next 10 years (prospective)
• Final outcome = human cases (unique).
Materials and Methods
• Stochastic systems model– Multiple Excel® spreadsheets
• Simulate the number of human Salmonella cases as a function of– Herd seroprevalence – Undefined abattoir interventions
• carcass swab culture status
– Human attribution model.
• Using best AVAILABLE data from Danish surveillance system
Systems model of pork attributable human cases (PAHC) of salmonellosis
Distribution of herds by
seroprevalence category and size
ProductionModule
Probability of carcass swab positive given
herd seroprevalence
category
Slaughter module
Attributes human cases as a function of
positive carcasses
AttributionModule
Pigs from each seroprevalence category
Human cases per year
Carcass swab positive pigs
0 5 15 2535 45 55 65 75 85 95
1995
1997
1999
2001
2003
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
Percentage
Seroprevalance Group
Year
Distribution of Herds by Average Annual Seroprevalence
1995
1996
1997
1998
1999
2000
2001
2002
2003
Changes in herd seroprevalence due to on-farm control program (actual data)
PRODUCTION module converts size distribution to number of pigs, using
stochastic version of average herd sizeOUTPUT --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------Number of PIGS from each ave. annual seroprevalence category by herd size
Size category 0 5 15 25 35 45 55 65 75 85 95 Chk total1-100 4818 147 196 98 0 0 147 0 0 0 49 5457101-200 67147 6919 6564 3105 1508 710 1774 177 266 177 266 88613201-500 248058 206715 45695 19040 7616 7072 4896 3264 1088 0 544 543988501-1000 371424 600728 106292 51354 22692 19109 11943 3583 4777 1194 0 11930961001-2000 900033 1783336 344622 160601 76954 46842 23421 10038 0 0 0 33458462001-3000 718598 1959196 352520 193208 98299 37286 13558 13558 3390 0 0 33896133001-5000 1188339 3253784 611146 305573 198056 73564 28294 5659 0 0 0 56644145000-10000 1111645 3911092 1091310 494817 155901 6778 0 0 0 0 0 6771544>10000 136736 768552 170920 84871 16503 0 0 0 0 0 0 1177581Tot pigs by seroprev 4.6E+06 1.2E+07 2.6E+06 1.2E+06 5.6E+05 1.9E+05 8.4E+04 3.6E+04 9.5E+03 1.4E+03 8.6E+02 2.10E+07
Systems model of pork attributable human cases (PAHC) of salmonellosis
Distribution of herds by
seroprevalence category and size
ProductionModule
Probability of carcass swab positive given
herd seroprevalence
category
Slaughter module
Attributes human cases as a function of
positive carcasses
AttributionModule
Pigs from each seroprevalence category
Human cases per year
Carcass swab positive pigs
Slaughter module
• “Creates” carcass swab positive pigs• As a function of average annual
seroprevalence• Combines data from all Danish abattoirs in
the program• Regression parameter
– adjusted to match historical carcass swab results with the distribution of average annual seroprevalence for that year
Note how coefficient improves over time due to changes in abattoir processes
Table II: Historical (1995-2003) carcass Salmonella swab data and resulting coefficient used in the model for determining positive carcasses.
Year Prevalence in Cuts1
Observed Carcass Swab Prevalence2
Resulting Coefficient3
1995 1.30% 2.47% 0.1207
1996 1.30% 2.47% 0.1100
1997 1.10% 2.09% 0.0971
1998 1.10% 2.09% 0.1467
1999 n/a 1.92% 0.1275
2000 n/a 1.43% 0.0570
2001 n/a 1.37% 0.0490
2002 n/a 1.38% 0.0530
2003 n/a 1.43% 0.0550
1Inflation factor for cut to carcass swap prevalence = 1.927 2From National Data of Danish Meat Association 3Estimated from the observed distribution of herds within average seroprevalence (Table 1) and the observed carcass swab prevalence
Systems model of pork attributable human cases (PAHC) of salmonellosis
Distribution of herds by
seroprevalence category and size
ProductionModule
Probability of carcass swab positive given
herd seroprevalence
category
Slaughter module
Attributes human cases as a function of
positive carcasses
AttributionModule
Pigs from each seroprevalence category
Human cases per year
Carcass swab positive pigs
Attribution module
• Uses attribution data 1999-2003– Methods shown in Hald et al. 2004. Risk Analysis
• Connects positive carcasses – (servings) produced that year with number of pork
attributed human Salmonella cases• Attribution factor
– pert distribution, • Same factor is used in retrospective (historical) and
prospective (future) simulations
Simulation • 5,000 iterations in @Risk• Outputs of interests
– pork attributable salmonellosis cases
• Retrospective (historical) simulations– (1995-2003)– Uses available historical trend data on
• On-farm seroprevalence or• Abattoir regression parameter
• Prospective (future) simulations– (2004-2013)
Systems model of pork attributable human cases (PAHC) of salmonellosis
Distribution of herds by
seroprevalence category and size
ProductionModule
Probability of carcass swab positive given
herd seroprevalence
category
Slaughter module
Attributes human cases as a function of
positive carcasses
AttributionModule
Pigs from each seroprevalence category
Human cases per year
Carcass swab positive pigs
Retrospective simulations(3 scenarios)
• Both combined (Historical)• Abattoir parameter improvements only
• Post-harvest• Farm at 1995 seroprevalence distribution• Farm at 2003 seroprevalence distribution
• Changes in herd seroprevalence over time• Pre-harvest• Abattoir 1995 regression parameter• Abattoir 2003 regression parameter
Historical: Combined effect of on-farm (pre-harvest) changes and abattoir (post-harvest)
Simulated pork attributable human cases
050
100150200250300350400450
Tota
l cas
es
Historical - Farm &Abattoir
Lower Historical
Upper Historical
Significant (P <0.05) changes in risk
occurred 1998-2000
Changes in herd seroprevalence over time: Abattoir parameter at 1995 level
= Effect of on-farm (pre-harvest) program only
Simulated pork attributable human cases
0
100
200
300
400
500
1995
1996
1997
1998
1999
2000
2001
2002
2003
To
tal
case
s
Historical - Farm &Abattoir
Lower Historical
Upper Historical
Abattoir95
19% decrease, 69 cases (P>0.05)
Abattoir (post-harvest) improvements only:
If no on-farm control program? (i.e. seroprevalence stayed at 1995 level)Simulated pork attributable human cases
0
100
200
300
400
500
1995
1996
1997
1998
1999
2000
2001
2002
2003
To
tal
case
s
Historical - Farm &Abattoir
Lower Historical
Upper Historical
Farm 1995
A difference of 81 cases (P>0.05). Due to abattoir
improvements only
Where do we go from here?Prospective simulations (2004-2013)
• Compare on-farm to abattoir– Improvements on-farm at same rate as the
past– Improvements in abattoir at same rate as
the past
• Carcass decontamination in slaughter– Reduced Pr (CS+) by 90%
Improve abattoir only,
40 cases
Improve farm only = no change
Human cases based on holding on-farm seroprevalence levels at 2003 levels and improving abattoir methods (Farm2003ImpAb); reverting on-farm to 1995 values while keeping abattoir values constant (RevF); and continuing to improve on-farm while keeping abattoir values constant (ImpF).
Reduce all herds to <25% ave. annual seroprevalence
0 5 15 25 35 45 55 65 75 85 95
2004
2006
2008
2010
2012
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
45.0%
50.0%
Percentage
Seroprevalance Group
Year
Distribution of herds by average annual seroprevalence
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
Decrease cases by ~10 to 190/yr
Cases while steadily improving abattoir methods for all on-farm values set at 25% or less (AllFarm25); reverting on-farm to 1995 values while keeping abattoir values constant (RevF); and continuing to improve on-farm while keeping abattoir values constant (ImpF).
Still end up with 110 cases/yr
What else can be done?Carcass decontamination
Number of pork attributable human cases
Percentage of the national herd decontaminated
Equivalent annual seroprevalence
cuttoff Lower Mean Upper
0% NA 136 152 168
1.50% Level 3, >45% 131 147 163
10.00% > 25% 110 126 142
100.00% all 14 15 16
Summary of results - Retrospective
• On-farm program reduced risk in early years (‘95-’98)
• Improvements since ‘98 due to unspecified abattoir improvements
• On-farm program ALONE would have only reduced average annual cases by ~69
Summary of results - Prospective
• Massive on-farm improvements needed to further reduce risk (new technology?)
• Don’t go back (on-farm seroprevalence)• Steady improvement in general slaughter
hygiene will make minimal reduction in risk
• Some form of carcass decontamination has the most significant impact on risk
Interesting points to ponder
• Denmark is a health policy “lab”• Lessons on the economics of food
safety• Use of “attribution” model• Shows what happens when “connect all
the dots”• Systems models provide a tool for
policy and research decision making
Future work: Add economics to a U.S model
Contaminated servings or potential infectious servings