Hubble Asteroid Hunter III. Physical properties of newly found asteroids
Ebodea usda-slides-2015-01-20
1. Investigating the transmission pathways of
porcine epidemic diarrhea virus (PEDV) using
outbreak incidence and virus sequence data
Eamon O’Dea
Department of Biology
Georgetown University
2. A large foot and mouth disease virus (FMDV) outbreak in
the U.S. could cause a large economic shock
data from Paarlberg et al., 2005
4. Key parameters for models are often uncertain
APHIS Overview of Modeling and Assessment Tools:
The data and information needed to properly estimate
parameters are often sparse, dated, and not readily
available. Researchers typically address these shortcomings
with expert opinion and informed assumptions.
5. Key parameters for models are often uncertain
McReynolds et al., 2014:
The estimates of the probability of indirect transmission
and achievable movement controls are uncertain
parameters, based solely on USDA subject matter expert
opinion. Model outputs are quite sensitive to these
parameters and an improved knowledge of the efficacy of
biosecurity practices and the ability to achieve movement
controls to limit direct and indirect transmission are
necessary for more focused planning of optimal control
efforts.
6. PEDV provides an example of a rapidly spreading pathogen
Red text gives positive accessions as of Jan. 2014.
AL:0AZ:0 AR:0
CA:1
CO:35
CT:0
DE:0
FL:0
GA:0
ID:0
IL:71
IN:67
IA:770
KS:143
KY:4
LA:0
ME:0
MD:1
MA:0
MI:12
MN:217
MS:0
MO:18
MT:0
NE:5
NV:0
NH:0
NJ :0
NM:0
NY:2
NC:301
ND:0
OH:60
OK:272
OR:0
PA:28 RI:0
SC:0
SD:5
TN:6
TX:26
UT:0
VT:0
VA:0
WA:0
WV:0
WI:4
WY:1
[10 to 74) [74 to 166) [166 to 728) [728 to 7,550]
Farm count
data from USDA APHIS VS NVSL National Animal Health Laboratory Network
12. Outline
Do pairs of states with large flows have similar case dynamics?
What variables seem relevant for predicting PEDV burdens?
Do flows improve the fit of an epidemiological model?
What do sequence data tell us about transmission routes?
13. Do pairs of states with large flows have similar
case dynamics?
14. AASV has been publishing weekly counts of positive test
results
MN KS
IL OK
IA NC
0
10
20
0
5
10
0
10
0
10
20
0
50
0
10
20
Jul Oct Jan Jul Oct Jan
Date
Cases
31. We used total cases and litter size changes as burdens
32. It is not easy to identify the best predictors
33. We used regularized regression and stability selection to
see which variables were relevant
Identifies variables with the most robust predictive power
Balances goal of finding small sets of variables while letting
correlated variables enter into model together
42. Modeling assumptions
Infected farms are infectious only the first week they are
infected
Consistent with other PEDV model (ANSES, 2014)
Best fit to the data
After being infective, farms are no longer susceptible
Reasonable for the time window we consider (38 weeks)
43. Our time-series susceptible-infected-recovered model
E(infectivesi,t+1) = (transmission rate)i,t
× [ jweighti,j (infectives)j,t + (other risks)]b0
× (susceptibles)i,t
E(Ii,t+1) = βi,t( jwi,jIj,t + η)b0 Si,t
(transmission rate)i,t = exp(b1 + Zi + b2t)
× (N2
i farm densityi )b3
× flowb4
i N−2
i
with Si,t = Ni − t−1
n=0 Ii,n and η, b, Z unknown.
47. Conclusions
Including estimates of flows significantly improves the fit of a
model of PEDV spread among farms.
Undirected flows fit better than directed flows, which suggests
we are not seeing the effects of the movement of live animals.
48. What do the sequence data tell us about
transmission routes?
50. In a preliminary analysis, we found that some pairs of
states have significantly higher transition rates
51. We are developing methods to efficiently estimate the
effects of candidate predictors on these transition rates
52. Overall conclusions
The incidence data support a model in which flows of animals
are correlated with transmission routes.
Time- and location-tagged sequence data contains additional
information about transmission routes, which we are developing
methods to extract more easily.
53. Acknowledgments
My supervisor Shweta Bansal has played a large part in the
development of this work.
We thank John Korslund and Harry Snelson for useful feedback on
veterinary and swine industry subject matter.
This work was supported by DHS Contract # HSHQDC-12-C-0014
and the RAPIDD Program of the Science & Technology Directorate,
Department of Homeland Security and the Fogarty International
Center, National Institutes of Health.
The views and conclusions contained in this document are those of
the author and should not be interpreted as necessarily representing
the official policies, expressed or implied, of the US Department of
Homeland Security.