swann gt nc guinea worm july 2019 short like carter · cost-benefit analysis of interventions 22 by...
TRANSCRIPT
Guinea Worm Ecology Model & Simulation
Pinar Keskinocak, Zihao Li, Julie Swann,Tyler Perini, Natashia BolandJuly 2019
GoalsGoals
Assist in understanding Guinea Worm (GW) disease spread in Chad and evaluate the effectiveness of potential interventions
ApproachAgent-based simulation model that tracks the disease spread of GW in dogsModel complex interactions (e.g., humans, dogs, hosts, and water) over multiple yearsFlexible model that can be fine tuned as more data become available
2
Example Life Cycle of GW with Paratenic Host
Key AssumptionsInfections occur through water or paratenic host (e.g., fish, tadpole, frog, lizard)Lifetime of host and L3 larvae in hostTiming of rainy season, link with consumption patterns of water or food, and corresponding infection rate
Eberhard et al, “The Peculiar Epidemiology of Dracunculiasis in Chad”, Am J Trop Med (2014)
3
Disease Incubation (e.g., 12 months)
Worms exuded monthly
Environment (e.g., Rainfall, Temp)
Dog population and behavior
Existing or Potential paratenic hosts
Interventions
Disease Modeling
4
Worms exuded monthly the following year
Today’s Focus: Timing,
Interventions
ExperimentsModel “tracks” lifecycle of larvae and worms and infections of dogs in systemExperiments explore different parameters and assumptions
Timing of high infectivity (e.g., June to Oct) Length of L3 availability (e.g., through paratenic host)Interventions (ABATE, tethering, other)Hundreds to thousands of computational experiments (so far)
Results are compared to Chad data (dogs) from 2014 to 2017
Weighted Mean Squared Error (WMSE) is an important quantification
L3 & host consumed
for infection
(I)
L3 penetrate, reproduce
Female worm
migrates and
exudes (E)
L1 larvae released
into water
L1 larvae consumed
by copepod
L1 larvae become L3 larvae I
E
I
E
050
100150200250300350400450
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58
Rain
fall
Months (Years 2014 to 2017)
Dogs Exuding Worms (visual)HEIGHT
INTERVALWIDTH
VALLEY > 0
Seasonality & Environmental Factors
6
Hypothetical Environmental Factors affecting High Infectivity Periods
7
0
0.2
0.4
0.6
0.8
1
J F M A M J J A S O N D
Weather Factors
No Effect
Temperature
Rainfall
Rainfall (acc.)
Combined
Dry SeasonFishing
Some Fishing
Rainy SeasonFarming
Model Fit (Examples)
8
No Seasonality (WMSE = 1069.88)
Combined (WMSE = 444.68)
Conclusion: Seasonality of infections is driven by more than just the life cycle. Infectivity is high (or low) in particular time periods.
Hypothetical Environmental Factors affecting Seasonality of Infections
9
0
0.2
0.4
0.6
0.8
1
J F M A M J J A S O N D
Weather Factors
No Effect
Temperature
Rainfall
Rainfall (acc.)
Combined
Weighted MSE
1069.88
620.35
1259.65
538.52
444.68
Dry SeasonFishing
Some Fishing
Rainy SeasonFarming
1) Decreases with more rain
2) Combined includes hot temperature
Calibrated Select Environmental Factors
10
0
0.2
0.4
0.6
0.8
1
J F M A M J J A S O N D
Weather Factors
Temperature
Combined
Rainfall (acc.)
Calibrated
Calibrated Environmental Factor
0
0.2
0.4
0.6
0.8
1
J F M A M J J A S O N D
Weather Factors
Temperature
Combined
Rainfall (acc.)
Calibrated
2.6% 3.1% 5.5% 10% 13% 16% 15% 12% 8.3% 5.9% 4.0% 2.6% Worm burden in water
Calibrated Select Environmental Factors
0
0.2
0.4
0.6
0.8
1
J F M A M J J A S O N D
Weather Factors
Temperature
Combined
Rainfall (acc.)
Calibrated
2.6% 3.1% 5.5% 10% 13% 16% 15% 12% 8.3% 5.9% 4.0% 2.6% Worm burden in water
68% of the worm burden between April-August. Also when the “good” environmental factors are most similar.
Transmission rate is decreasing (even with worms being exuded), e.g., due to changes in behavior or water or hosts
Estimates of reproductive rate vary with time
13
1 infected dog mayresult in 0 new infections or 4 to 10 new infections.
In the peak infectivity time periods or large villages, 1 infected dog is likely to infect more dogs.
Infections in 2018 (Increase)
14
Greater surveillance in-countryGap between reported intervention and “effective” intervention
E.g., 40% less tethering or 30% less effective tethering and ABATE
Impact from earlier peak in rainfall in 2017?
Predicting Future Years“It’s tough to make predictions, especially about the future”. (Yogi Berra)
15
16
ABATE Tether Other
40% 76% 17%
Initialization Period2014-2017
BASELINEInfections continue
Intervention What-If Analysis
17
ABATE Tether Other
40% 76% 17%
70% 95% 17%
Initialization Period2014-2017
BASELINEInfections continue
INCREASE interventions significantly
Infections decrease
Intervention What-If Analysis
18
ABATE Tether Other
40% 76% 17%
70% 95% 17%
20% 50% 17%
Initialization Period2014-2017
BASELINEInfections continue
INCREASE interventions significantly
Infections decrease
DECREASE interventions significantly
Infections explode
Intervention What-If Analysis
What might be needed for eradication within 10 years?
19
When comparing various levels of ABATE & Tethering coverage, we can identify the few combinations that are likely to eradicate within 10 years. 99% coverage is understandably impractical.
Likelihood of eradication within 10 years
20
When comparing various levels of ABATE & Tethering coverage, we can identify the few combinations that are likely to eradicate within 10 years. 99% coverage is understandably impractical.
When ABATE is at its highest level, all 6 scenarios reach eradication.
When Tethering is at its highest level, only 4 out of 6 scenarios reach eradication.
Likelihood of eradication within 10 years
21
When comparing various levels of ABATE & Tethering coverage, we can identify the few combinations that are likely to eradicate within 10 years. 99% coverage is understandably impractical.
Few practical combinations where Tethering 95% And ABATE 85%
Cost-benefit analysis of interventions
22
By incorporating the costs of each intervention, we can do cost-benefit analysis.We have assumed tethering costs $100 per dog, and the cost of ABATE is piecewise linear along the following graph:
$50,000 for 50% coverage
$150,000 for 70% coverage
Cost-benefit analysis of interventions
23
When comparing the practical combinations, we can observe big picture patterns about the strengths of increasing ABATE vs. tethering coverage.
Number of remainingdog infections when not eradicated:Maximum = 526Median = 162
Cost-benefit analysis of interventions
24
When comparing the practical combinations, we can observe big picture patterns about the strengths of increasing ABATE vs. tethering coverage.
Number of remainingdog infections when not eradicated:Maximum = 526Median = 162
Eradication will take time
Least-costly solutions in the long-term are those that invest at highest levels immediately
Current ConclusionsTiming of infectivity (and worm burden) is important to understanding causes and effective interventions
E.g., Shallow pools or tadpoles or fish entrailsE.g., Providing dogs water to drink, burying, tethering (possibly proactively)
Eradication may take yearsEarly, full-level interventions are ultimately cheaper
Continue current interventions while trying others“Contain cases” and “Clean water”Proactive tethering and providing dogs water to drink could also keep them away from shallow pools, especially during high infectivity periods
25
Contact InformationNatasha Boland ([email protected]) Pinar Keskinocak ([email protected])Zihao Li ([email protected])Tyler Perini ([email protected])Julie Swann ([email protected])
26