Dr. Aaron Scott - Advantages of Probability-Based Approaches to Animal Disease Surveillance
31
NIAA March 27, 2012 Aaron Scott, DVM PhD DACVPM (epi) Center Director, NSU Probability based approaches to animal health surveillance 1
description
Advantages of Probability-Based Approaches to Animal Disease Surveillance - Dr. Aaron Scott, Center Director, USDA/APHIS/VS/CEAH, from the 2012 Annual Conference of the National Institute for Animal Agriculture, March 26 - 29, Denver, CO, USA. More presentations at: http://www.trufflemedia.com/agmedia/conference/2012-decreasing-resources-increasing-regulation-advance-animal-agriculture
Transcript of Dr. Aaron Scott - Advantages of Probability-Based Approaches to Animal Disease Surveillance
1
NIAAMarch 27, 2012
Aaron Scott, DVM PhD DACVPM (epi) Center Director, NSU
Probability based approaches to animal health surveillance
2
National Surveillance Unit“Building Partnerships and Leading Change”
Safeguarding Animal Health
3
Using risk factors and probability to get the most
information at the lowest cost
4
Why?
5
Federal/State budgets
6
What is surveillance?
Find disease
if its not there…
Prove it!!!
7
Surveillance is evidence!
8
Reverend Thomas Bayes
9
Evidence may be combination of sample testing and
other knowledge about risk factors
10
“Risk” = likelihood * consequence
= animals more likely to get disease? = animals more likely to have disease? = animals more likely to be detected?
= more dangerous to people????
Risk-based versus probability
11
Inference …
…is drawing a conclusion about the population from one or
more pieces of evidence
12
Sampling and inference
Convenience
Random
Representative
13
Random Selection
•Inference is based on random sampling•Listing of all subjects (sample frame)•Each has equal chance of selection•Is always representative
14
Random Sample size
“n” = 3X when 1:X
3,000,000 samples to detect 1 in 1,000,000
15
Representative – non-biased
A representative sample could be just anybody!
17
Geographic representation
18
Representative
If not random, must be other supporting evidence for
representativeness…
19
Targeted (“risk based”) sampling
Concept: Every sample gives information
Corollary: Some are worth more than others
Targeted = intentional bias!
20
Must clearly define sub populationsMust have knowledge of risk factors
Must know magnitude of risk factorsTargeting represents sub-populationsMust know relationship
Targeting and inference
Population
Clear risk factor
Clear risk factor
3X
24
What is “n”?
40 H-risk * value 3 = 12010 L-risk * value 0.5 = 5
50 tests = 125 “virtual” samples
25
What are the risk factors?
Intentionally biased sample
+ other information
26
Sows and boars (value = 2.5x to 5x ) Exposure to feral pigs (10x ? 100x ?) Non-confinement (??)
Some samples from all populationsMarket swine (< 1x)
PRV Targeting
27
Slaughter sampling
From 3 herds: 5,000, 30, 10
Need “n” of 30/farm for inference30 + 30 + 10 =70 samples
Without additional info,n = 5,040!!!
28
Collect all with traceable IDSend to lab and sortOnly knowledge is State of originDiscard all but 5% from each State
Current PRV
29
PIN tag pilot
Collaboration:• NPB, AASV, industry reps• 6 State Veterinarians• Veterinary Services
30
High value (feral swine exposure)Limit to number neededCollect “pink tag”, scan ID, combine
zip with feral risk, target # needed Other tagsSpend the money on other diseases
PIN Pilot
31Safeguarding Animal Health
http://www.aphis.usda.gov/vs/ceah/ncahs/nsu/
Thanks!!
