DRAFT – Not for a.ribu2on or distribu2on Modeling the Ebola Outbreak in West Africa, 2014 November 7 th Update Bryan Lewis PhD, MPH ([email protected]) Caitlin Rivers MPH, Eric Lofgren PhD, James Schli., Alex Telionis MPH, Meredith Wilson MPH Henning Mortveit PhD, Dawen Xie MS, Samarth Swarup PhD, Hannah Chungbaek, Keith Bisset PhD, Maleq Khan PhD, Chris Kuhlman PhD, Stephen Eubank PhD, Madhav Marathe PhD, and Chris Barre. PhD Technical Report #14114
43
Embed
Modeling the Ebola Outbreak in West Africa, November 7th 2014 update
Researchers at the Network Dynamics and Simulation Science Laboratory have been using a combination of modeling techniques to predict the spread of the Ebola outbreak.
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
DRAFT – Not for a.ribu2on or distribu2on
Modeling the Ebola Outbreak in West Africa, 2014
November 7th Update
Bryan Lewis PhD, MPH ([email protected]) Caitlin Rivers MPH, Eric Lofgren PhD, James Schli., Alex Telionis MPH, Meredith Wilson MPH
Henning Mortveit PhD, Dawen Xie MS, Samarth Swarup PhD, Hannah Chungbaek, Keith Bisset PhD, Maleq Khan PhD, Chris Kuhlman PhD,
Stephen Eubank PhD, Madhav Marathe PhD, and Chris Barre. PhD
Technical Report #14-‐114
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Currently Used Data
● Data from WHO, MoH Liberia, and MoH Sierra Leone, available at h.ps://github.com/cmrivers/ebola
● MoH and WHO have reasonable agreement ● Sierra Leone case counts censored up
to 4/30/14. ● Time series was filled in with missing
dates, and case counts were interpolated.
2
Cases Deaths Guinea 1906 997 Liberia 6454 2705 Sierra Leone 5235 1500 Total 13,617 5210
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Liberia – Case Loca2ons
3
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Liberia – County Case Incidence
4
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Learning from Lofa
15
Model fit to Lofa case series up Aug 18th (green) then from Aug 19 – Oct 21 (blue), compared with real data (red)
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Learning from Lofa
16
Model fit to Lofa case with a change in behaviors resul2ng in reduced transmission sta2ng mid-‐Aug (blue), compared with observed data (green)
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Learning from Lofa
17
Model fit to Liberian case data up to Sept 20th (current model in blue), reduc2on in transmissions observed in Lofa applied from Sept 21st on (green), and observed cases (red)
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Learning from Lofa
18
Model fit to Liberia case with a change in behaviors resul2ng in reduced transmission sta2ng Sept 21st (green), compared with observed data (blue)
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
• New GUI deployed for running ABM expts • Ini2al calibra2on with travel for all Liberia – Plausible base case determined – Ini2al a.empts with spa2al spread
19
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Regional Travel -‐ Liberia • Mobility data comes from flowminder.org
– Probability Matrix of county to county trips by week (15x15) – Number of trips probably high, ra2os be.er – Es2mates available for several model fits – Data converted to daily probabili2es
• Method: Make dynamic schedules for EpiSimdemics – Each person has a home county based on home loca2on – Each person is matched with a person in each non-‐home county, based on gender and age bin
– For each person and non-‐home county, a new schedule is created that shadows the schedule of the matched person
– A scenario file is created that contains rules for each source/des2na2on pair (15 x 14 = 210 for Liberia)
20
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Regional Travel -‐ Example
21
# Travel from Grand_Kru (2042) to Maryland (2082) with prob 0.008036427trigger repeatable person.County = 2042 and person.isTraveling = -1 apply travel_to_2082 with prob=0.008036427
intervention travel_to_2008 set person.isTraveling = 2008 set person.daysLeft = 3 set tripsTo2008++ set traveling++ set trips++ schedule county2008 1
# return from travelintervention return unschedule 1 set person.isTraveling = -1 set person.daysLeft = -1 set traveling--
trigger repeatable person.daysLeft > 0 set person.daysLeft—
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Auto-‐Calibra2on of ABM
26
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
SIBEL – New version
27
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
SIBEL – New features
• Generic interven2on supports more possible interven2ons
• Dura2on and logis2cal rates of interven2on added
• Many more…
28
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Plausible Base Case
29
• Hospital isola2on for 50% -‐ reduces txm by 80% • Proper burial for 50% -‐ reduces txm by 80% • Ebola Mode: Transmission in household 3x more likely than
outside the household
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Transmission calibra2on
30
4665 cases
Day 158 Day 27 22 cases
131 days Burn in period
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Calibra2on – Spa2al Spread
31
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Simula2on Comparison
32
Cases per 100k popula2on
Mean simula2on results Ministry of Health Data
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Simula2on Comparison
33
Total Cases
Single Simula2on result Ministry of Health Data
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Agent-‐based Next Steps • Spa2al spread calibra2on – Incorporate degraded road network to help guide finng to current data
– Guide with more spa2ally explicit ini2al infected seeds and intervenBons
– Incorporate road degrada2on for rainy season • Planned Experiments: – Impact of hospitals with geo-‐spa2al disease
• Study design / implementa2on under construc2on – Vaccina2on campaign effec2veness
• Framework under development
34
DRAFT – Not for a.ribu2on or distribu2on
APPENDIX Suppor2ng material describing model structure, and addi2onal results
35
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Legrand et al. Model Descrip2on
Exposednot infectious
InfectiousSymptomatic
RemovedRecovered and immune
or dead and buried
Susceptible
HospitalizedInfectious
FuneralInfectious
Legrand, J, R F Grais, P Y Boelle, A J Valleron, and A Flahault. “Understanding the Dynamics of Ebola Epidemics” Epidemiology and Infec1on 135 (4). 2007. Cambridge University Press: 610–21. doi:10.1017/S0950268806007217.
36
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Compartmental Model
• Extension of model proposed by Legrand et al. Legrand, J, R F Grais, P Y Boelle, A J Valleron, and A Flahault. “Understanding the Dynamics of Ebola Epidemics” Epidemiology and Infec1on 135 (4). 2007. Cambridge University Press: 610–21. doi:10.1017/S0950268806007217.
37
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Legrand et al. Approach
• Behavioral changes to reduce transmissibili2es at specified days
• Stochas2c implementa2on fit to two historical outbreaks – Kikwit, DRC, 1995 – Gulu, Uganda, 2000
• Finds two different “types” of outbreaks – Community vs. Funeral driven outbreaks
38
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Parameters of two historical outbreaks
39
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
NDSSL Extensions to Legrand Model
• Mul2ple stages of behavioral change possible during this prolonged outbreak
• Op2miza2on of fit through automated method
• Experiment: – Explore “degree” of fit using the two different outbreak types for each country in current outbreak
40
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Op2mized Fit Process • Parameters to explored selected – Diag_rate, beta_I, beta_H, beta_F, gamma_I, gamma_D, gamma_F, gamma_H
– Ini2al values based on two historical outbreak • Op2miza2on rou2ne
– Runs model with various permuta2ons of parameters
– Output compared to observed case count
– Algorithm chooses combina2ons that minimize the difference between observed case counts and model outputs, selects “best” one
41
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on
Fi.ed Model Caveats
• Assump2ons: – Behavioral changes effect each transmission route similarly
– Mixing occurs differently for each of the three compartments but uniformly within
• These models are likely “overfi.ed” – Many combos of parameters will fit the same curve – Guided by knowledge of the outbreak and addi2onal data sources to keep parameters plausible
– Structure of the model is supported
42
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on