modeling the ebola outbreak in west africa, december 22nd 2014 update

33
DRAFT – Not for a.ribu2on or distribu2on DRAFT – Not for a.ribu2on or distribu2on Modeling the Ebola Outbreak in West Africa, 2014 December 22 nd Update Bryan Lewis PhD, MPH ([email protected] ) presen2ng on behalf of the Ebola Response Team of Network Dynamics and Simula2on Science Lab from the Virginia Bioinforma2cs Ins2tute at Virginia Tech Technical Report #14131

Upload: ndsslvt

Post on 03-Aug-2015

50 views

Category:

Science


2 download

TRANSCRIPT

Page 1: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Modeling  the  Ebola    Outbreak  in  West  Africa,  2014  

December  22nd  Update    

Bryan  Lewis  PhD,  MPH  ([email protected])  presen2ng  on  behalf  of  the  Ebola  Response  Team  of    

Network  Dynamics  and  Simula2on  Science  Lab  from  the  Virginia  Bioinforma2cs  Ins2tute  at  Virginia  Tech  

Technical  Report  #14-­‐131  

Page 2: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

NDSSL  Ebola  Response  Team  Staff:  Abhijin  Adiga,  Kathy  Alexander,  Chris  Barre.,  Richard  Beckman,  Keith  Bisset,  Jiangzhuo  Chen,  Youngyoun  Chungbaek,  Stephen  Eubank,  Sandeep  Gupta,  Maleq  Khan,  Chris  Kuhlman,  Eric  Lofgren,  Bryan  Lewis,  Achla  Marathe,  Madhav  Marathe,  Henning  Mortveit,  Eric  Nordberg,  Paula  Stretz,  Samarth  Swarup,  Meredith  Wilson,Mandy  Wilson,  and  Dawen  Xie,  with  support  from  Ginger  Stewart,  Maureen  Lawrence-­‐Kuether,  Kayla  Tyler,  Kathy  Laskowski,  Bill  Marmagas    Students:  S.M.  Arifuzzaman,  Aditya  Agashe,  Vivek  Akupatni,  Caitlin  Rivers,  Pyrros  Telionis,  Jessie  Gunter,  Elisabeth  Musser,  James  Schli.,  Youssef  Jemia,  Margaret  Carolan,  Bryan  Kaperick,  Warner  Rose,  Kara  Harrison          

2

Page 3: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

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.  

3

       Cases  Deaths    Guinea      2,416  1,525    Liberia      7,830  3,290    Sierra  Leone    8,759  2085    Total      19,005  6,900  

Page 4: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Liberia  –  Case  Loca2ons  

4

Page 5: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Liberia  infec2on  rate  

5

Page 6: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Liberia  Forecast    

6

11/03  to  

11/09  

11/10  to  

11/16  

11/17  to  

11/23  

11/24  to  

11/30  

12/1  to  

12/7  

12/8  to  

12/14  

12/15  to  

12/21  

12/22  to  

12/28  

12/29  to  

1/04  

1/05  to  

1/11  

1/12  to  1/8  

Reported   362   185   187   156   369   172   -­‐-­‐  

Newer  model   457   444   431   419   407   270   254   240   226   214   201  

Reproduc2ve  Number  Community      0.23  Hospital        0.3  Funeral          0.2  Overall            0.8    

   

Page 7: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Liberia  long  term  forecasts  

7

Date   Weekly  forecast  

12/08   270  

12/15   255  

12/22   240  

12/29   227  

1/05   213  

1/12   202  

1/19   190  

1/26   179  

2/2   169  

2/9   160  

2/16   142  

2/23   127  

Page 8: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Sierra  Leone  –  County  Data  

8

Page 9: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Sierra  Leone  infec2on  rate  

9

Page 10: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Sierra  Leone  Forecast  

10

10/06  to  

10/12  

10/13  to  

10/19  

10/20  to  

10/26  

10/27    to  

11/02  

11/03    to  

11/09  

11/10  to  

11/16  

11/17  to  

11/23  

11/24  to  

11/30  

12/01  to  

12/07  

12/08  to  

12/14  

12/15  to  

12/21  

12/22  to  

12/28  

Reported   468   461   454   580   480   684   643   577   598   621   582  

Forecast  original   566   690   841   1025   1250   1523   1856  

Forecast  change  txm   430   524   513   543   566   588   612   636   660   713   740  

35%  of  cases  are  hospitalized  

ReproducMve  Number  Community  0.8  Hospital    0.3    Funeral    0.1    Overall    1.1    

Page 11: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

SL  longer  term  forecast  

11

Sierra  Leone  –  Newer  Model  fit  –  Weekly  Incidence   Date   Weekly  forecast  

12/08   660  

12/15   686  

12/22   713  

12/29   740  

1/05   769  

1/12   799  

1/19   830  

1/26   862  

2/2   895  

2/9   929  

2/16   965  

2/23   1002  

Page 12: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Sierra  Leone  -­‐  Prevalence  

12

Date   People  in  H  +  I  

12/08   898  

12/15   932  

12/22   968  

12/29   1006  

1/05   1045  

1/12   1085  

1/19   1128  

1/26   1171  

2/2   1216  

2/9   1263  

2/16   1312  

2/23   1363  

Page 13: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Guinea  Forecasts  

13

40%  of  cases  are  hospitalized  

ReproducMve  Number  Community  0.7    Hospital    0.1    Funeral    0.1    Overall    0.9    

10/09  to  

10/15  

10/16  to  

10/19  

10/23to  

10/29  

10/30to  

11/05  

11/06    to  

11/12  

11/13  to  

11/19  

11/20  to  

11/26  

11/27  to  

12/03  

12/04  to  

12/10  

12/11  to  

12/17  

12/18  to  

12/24  

12/25  to  

1/01    

Reported   175   129   143   12   136   121   142   69   86   55126   53  

Forecast   118   118   115   112   109   106   103   87   84   80   77   77  

Page 14: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Guinea  –  longer  term  forecast  

14

Date   Weekly  forecast  

12/08   81  

12/15   78  

12/22   75  

12/29   72  

1/05   69  

1/12   66  

1/19   63  

1/26   60  

2/2   58  

2/9   55  

2/16   53  

2/23   51  

Page 15: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Guinea  Prevalence  

15

Date   People  in  H+I  

12/08   98  

12/15   94  

12/22   90  

12/29   86  

1/05   82  

1/12   79  

1/19   76  

1/26   72  

2/2   69  

2/9   66  

2/16   64  

2/23   61  

Page 16: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Agent-­‐based  Model  Sierra  Leone  Calibra2on  

16

Incorporates  behavioral  change  around  Oct  15th  

Page 17: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Imported  into  SIBEL  

17

Page 18: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Sierra  Leone  Base  Case  

•  Bias  towards  household  members  – 70%  less  likely  to  transmit  outside  the  household  

•  Hospital  Isola2on  – 57%  isolated  with  82%  efficacy  

•  Proper  Burial  – 58%  buried  with  77%  reduc2on  

•  Behavioral  Change  –  mid  Oct  – 16%  reduc2on  in  effec2ve  contacts  

18

Page 19: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Agent-­‐Based  Model  Next  Steps  

•  Targeted  Vaccines    – Debug  

•  Sierra  Leone  upda2ng  – Na2onal  spread  calibrated  – Work  on  Geographic  spread  – Run  similar  prelim  study  

•  Genera2on  of  Synthe2c  disease  spread  as  public  resource  for  designing  vax  trials  and  campaigns  

19

Page 20: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

FORECAST  COMPARISON  TOOL  EpiViewer  

20

Page 21: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

EpiViewer  

•  Portal  for  exploring,  sharing,  comparing,  and  dissemina2ng  forecasts  of  infec2ous  diseases  

•  Currently  set  up  for  Ebola  Forecasts  •  Intended  as  a  community  tool  – Hosted  on  public  facing  URL  (once  finalized)  – Offers  simple  controls  for  exploring  the  data  – Allows  independent  uploading  by  3rd  par2es  to  facilitate  sharing,  as  well  as  consolidated  downloading  of  data  

21

Page 22: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

EpiViewer  –  Explore  Data  

22

Page 23: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

EpiViewer  –  Compare  Forecasts  

23

Page 24: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

EpiViewer  –  Upload  Data  

24

Page 25: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

EpiViewer  –  Download  All  Data  

25

Columbia_Liberia_MW_11-­‐09.csv  Columbia_Liberia_MW_11-­‐16.csv  Columbia_Liberia_MW_11-­‐23.csv  Columbia_Liberia_MW_11-­‐30.csv  Columbia_SL_MW_10-­‐05.csv  Columbia_SL_MW_10-­‐12.csv  Columbia_SL_MW_10-­‐19.txt  Columbia_SL_MW_10-­‐26.csv  Columbia_SL_MW_11-­‐02.csv  HealthMap_Liberia_Aug.31_MW  HealthMap_Liberia_Sept.07_MW  HealthMap_Liberia_Sept.14_MW  HealthMap_Liberia_Sept.21_MW  HealthMap_SL_Aug.17_MW  HealthMap_SL_Aug.31_MW  Mobs_Guinea_10-­‐01_MW.txt  MoBS_Liberia_Oct.1_MW.csv  MoBS_SL_10-­‐1_MW.csv  MoBS_SL_11-­‐1_MW.csv  MoBS_SL_11-­‐10_MW.csv  NDDSL_SL_Dec16.csv  NDSSL_SL_06-­‐07_MW.csv  NDSSL_SL_08-­‐09_MW.csv  NDSSL_SL_10-­‐11_MW.csv  Sur_SL_06-­‐07.csv  Sur_SL_08-­‐09.csv  Sur_SL_10-­‐11.csv  

Downloads  a  zip  file  Containing  all  data  

Page 26: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

APPENDIX  Suppor2ng  material  describing  model  structure,  and  addi2onal  results  

26

Page 27: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

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.  

27

Page 28: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

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.  

28

Page 29: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

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  

29

Page 30: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

DRAFT  –  Not  for  a.ribu2on  or  distribu2on    

Parameters  of  two  historical  outbreaks  

30

Page 31: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

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  

31

Page 32: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

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  

32

Page 33: Modeling the Ebola Outbreak in West Africa, December 22nd 2014 update

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  

33