Challenges in Modeling SARS-CoV-2: Bridging the Best of Both Worlds - - PowerPoint PPT Presentation

challenges in modeling sars cov 2 bridging the best of
SMART_READER_LITE
LIVE PREVIEW

Challenges in Modeling SARS-CoV-2: Bridging the Best of Both Worlds - - PowerPoint PPT Presentation

Challenges in Modeling SARS-CoV-2: Bridging the Best of Both Worlds Between Models and Reality Michael Li - McMaster University mcmaster.ca | Background Math and Statistics MacTheoBio Theoretical, statistical and computational approaches to


slide-1
SLIDE 1

mcmaster.ca

|

Challenges in Modeling SARS-CoV-2: Bridging the Best of Both Worlds Between Models and Reality

Michael Li - McMaster University

slide-2
SLIDE 2

mcmaster.ca

|

Background

2 June 26, 2020

Theoretical, statistical and computational approaches to study biology MacTheoBio Infectious diseases Spatial ecology Evolution Math and Statistics

slide-3
SLIDE 3

mcmaster.ca

|

Background

3 June 26, 2020

Theoretical, statistical and computational approaches to study biology MacTheoBio Infectious diseases Math and Statistics

Malaria Ebola Influenza Canine Rabies HIV etc..

slide-4
SLIDE 4

mcmaster.ca

|

Background

4 June 26, 2020

Theoretical, statistical and computational approaches to study biology MacTheoBio Infectious diseases Math and Statistics

Malaria Ebola Influenza Canine Rabies HIV etc..

I can read Chinese!

slide-5
SLIDE 5

mcmaster.ca

| 5

2019–2020 WuHan Novel Corona Virus Outbreak Part 1

June 26, 2020

slide-6
SLIDE 6

mcmaster.ca

|

Background

6

2019–2020 WuHan Novel Corona Virus Outbreak

  • Late December 2019, several

suspicious cases of pneumonia in Wuhan City, Hubei Province of China.

  • Unknown virus
  • Early January 2020, Chinese

authorities confirmed that they had identified a new virus.

June 26, 2020

slide-7
SLIDE 7

mcmaster.ca

|

Data collection

7 June 26, 2020

Jan 21st 24:00 105 new cases, 3 discharge and 3 death 375 Cumulative cases 28 Discharged 9 death

slide-8
SLIDE 8

mcmaster.ca

|

The Basic Reproductive Number/Ratio (R0)

8 June 26, 2020

  • Expected number of new cases per case
  • Good index of risk at the population level
slide-9
SLIDE 9

mcmaster.ca

|

The Basic Reproductive Number/Ratio (R0)

9

Most Valuable Piece

  • f information in disease modelling

June 26, 2020

  • Expected number of new cases per case
  • Good index of risk at the population level
slide-10
SLIDE 10

mcmaster.ca

|

WH outbreak R0 Estimates

10 June 26, 2020

Majumder, M.S. and Mandl, K.D., 2020. Early in the epidemic: impact of preprints on global discourse about COVID-19 transmissibility. The Lancet Global Health, 8(5), pp.e627-e630.

slide-11
SLIDE 11

mcmaster.ca

|

  • Exponential growth rate (r)
  • Generation Interval

Exponential Fitting Framework

11 June 26, 2020

slide-12
SLIDE 12

mcmaster.ca

|

  • Exponential growth rate (r)
  • Generation Interval

Exponential Fitting Framework

12 June 26, 2020

slide-13
SLIDE 13

mcmaster.ca

|

Exponential growth rate (r)

13 June 26, 2020

  • Estimate from time series

data

slide-14
SLIDE 14

mcmaster.ca

|

Exponential growth rate (r)

14 June 26, 2020

  • Estimate from time series

data

slide-15
SLIDE 15

mcmaster.ca

|

Exponential growth rate (r)

15 June 26, 2020

  • Estimate from time series

data

X

slide-16
SLIDE 16

mcmaster.ca

|

Exponential growth rate (r)

16 June 26, 2020

  • Estimate from time series

data

  • Fitting to incidence data
  • Logistic model with

negative binomial noise

  • epigrowthfit package in R

Ma, J., Dushoff, J., Bolker, B.M. and Earn, D.J., 2014. Estimating initial epidemic growth rates. Bulletin of mathematical biology, 76(1), pp.245-260.

slide-17
SLIDE 17

mcmaster.ca

|

  • Exponential growth rate (r)
  • Generation Interval

Exponential Fitting Framework

17 June 26, 2020

slide-18
SLIDE 18

mcmaster.ca

|

Generation interval

18 June 26, 2020

  • Time between

infections

  • Focal individual
slide-19
SLIDE 19

mcmaster.ca

|

Generation interval

19 June 26, 2020

  • Time between

infections

  • Focal individual
  • Infection are hard to
  • bserve
  • Serial intervals
  • Time between

symptom onsets

slide-20
SLIDE 20

mcmaster.ca

|

  • Exponential growth rate (r)
  • Generation Interval

Exponential Fitting Framework

20 June 26, 2020

slide-21
SLIDE 21

mcmaster.ca

|

  • Exponential growth rate (r)
  • Generation Interval

Exponential Fitting Framework

21 June 26, 2020

Euler-Lotka equation

Wallinga, J. and Lipsitch, M., 2007. How generation intervals shape the relationship between growth rates and reproductive numbers. Proceedings of the Royal Society B: Biological Sciences, 274(1609), pp.599-604.

slide-22
SLIDE 22

mcmaster.ca

|

  • Exponential growth rate (r)
  • Generation Interval (GI)
  • Mean GI ( )
  • Dispersion ( )

Gamma approximation framework

22 June 26, 2020

Park, S.W., Champredon, D., Weitz, J.S. and Dushoff, J., 2019. A practical generation-interval-based approach to inferring the strength of epidemics from their speed. Epidemics, 27, pp.12-18.

slide-23
SLIDE 23

mcmaster.ca

|

R0 estimates

23 June 26, 2020

Park, S.W., Bolker, B.M., Champredon, D., Earn, D.J., Li, M., Weitz, J.S., Grenfell, B.T. and Dushoff, J., 2020. Reconciling early-outbreak estimates of the basic reproductive number and its uncertainty: framework and applications to the novel coronavirus (SARS-CoV-2) outbreak. (In press)

slide-24
SLIDE 24

mcmaster.ca

|

Parameter Uncertainties

24 June 26, 2020

slide-25
SLIDE 25

mcmaster.ca

|

Parameter Uncertainties

25 June 26, 2020

  • Uncertain of R0 comes from these

parameters

  • Important to propagate uncertainties

from all the parameters

slide-26
SLIDE 26

mcmaster.ca

|

Forecasting

26 June 26, 2020

Li, M., Dushoff, J. and Bolker, B.M., 2018. Fitting mechanistic epidemic models to data: A comparison of simple Markov chain Monte Carlo approaches. Statistical methods in medical research, 27(7), pp. 1956-1967.

Reports

  • Bayesian Discrete-time SIR
  • Dual beta-binomial process
  • Transmission and
  • bservation
  • 2 week projection
slide-27
SLIDE 27

mcmaster.ca

|

Summary

27

  • Easy to make a model
  • Extremely hard to make models that:
  • are informative
  • predict well
  • work in real time
  • Discrepancies between R0 estimates is

hard to understand

  • Method
  • Data
  • Hard to communicate clearly.
  • E.g. R0 ~ 3 (2,4)

June 26, 2020

slide-28
SLIDE 28

mcmaster.ca

|

Summary

28

  • Easy to make a model
  • Extremely hard to make models that:
  • are informative
  • predict well
  • work in real time
  • Discrepancies between R0 estimates is

hard to understand

  • Method
  • Data
  • Hard to communicate clearly.
  • E.g. R0 ~ 3 (2,4)

June 26, 2020

  • Predictions of the short-term

trajectories are useful

slide-29
SLIDE 29

mcmaster.ca

|

Moving forward and preparing for the global pandemic

29

  • Interesting and important to figure out the discrepancies
  • What else? What do people want to know?

June 26, 2020

slide-30
SLIDE 30

mcmaster.ca

|

Preparing for the global pandemic

30 June 26, 2020

slide-31
SLIDE 31

mcmaster.ca

|

Preparing for the global pandemic

31 June 26, 2020

slide-32
SLIDE 32

mcmaster.ca

| 32

Distributions of delays associated with COVID-19 healthcare in Ontario, Canada Part 2

June 26, 2020

slide-33
SLIDE 33

mcmaster.ca

| 33

Preparing for the global pandemic

June 26, 2020

slide-34
SLIDE 34

mcmaster.ca

|

Distributions of delays associated with COVID-19 healthcare in Ontario, Canada

34 June 26, 2020

slide-35
SLIDE 35

mcmaster.ca

|

Time intervals

35 June 26, 2020

slide-36
SLIDE 36

mcmaster.ca

|

Time intervals

36

Isolation time Specimen collection time Reporting time ER time Hospital admission time ICU time

June 26, 2020

slide-37
SLIDE 37

mcmaster.ca

|

Chronological Episodes (Symptom onset)

37

Days

Delay

June 26, 2020

slide-38
SLIDE 38

mcmaster.ca

| 38

Chronological Episodes (Isolation)

Days Days

Delay

Duration June 26, 2020

slide-39
SLIDE 39

mcmaster.ca

| 39

Chronological Episodes (Specimen collection)

Days

Delay

June 26, 2020

slide-40
SLIDE 40

mcmaster.ca

| 40

Chronological Episodes (ER)

Days Days

Delay

Duration June 26, 2020

slide-41
SLIDE 41

mcmaster.ca

| 41

Chronological Episodes (Hospitalization)

Days Days

Delay

Duration June 26, 2020

slide-42
SLIDE 42

mcmaster.ca

| 42

Chronological Episodes (ICU)

Days Days

Delay

Duration June 26, 2020

slide-43
SLIDE 43

mcmaster.ca

| 43

Chronological Episodes (Death)

Days

Delay

June 26, 2020

slide-44
SLIDE 44

mcmaster.ca

| 44

Chronological Episodes (Intubation)

Days Days

Delay

Duration June 26, 2020

slide-45
SLIDE 45

mcmaster.ca

| 45

Chronological Episodes (Ventilator)

Days Days

Duration

Delay

June 26, 2020

slide-46
SLIDE 46

mcmaster.ca

| 46 June 26, 2020

Admissions, Durations and Occupancies

Daily Admissions

slide-47
SLIDE 47

mcmaster.ca

| 47 June 26, 2020

Admissions, Durations and Occupancies

Daily Admissions Durations

slide-48
SLIDE 48

mcmaster.ca

| 48 June 26, 2020

Admissions, Durations and Occupancies

Daily Admissions Durations Occupancy

slide-49
SLIDE 49

mcmaster.ca

|

Distributions of delays associated with COVID-19 healthcare in Ontario, Canada

49

Changes over time

June 26, 2020

slide-50
SLIDE 50

mcmaster.ca

|

New Confirmations

50 June 26, 2020

Ontario, Canada

slide-51
SLIDE 51

mcmaster.ca

|

Testing capacity

51 June 26, 2020

slide-52
SLIDE 52

mcmaster.ca

|

Testing capacity

52 June 26, 2020

slide-53
SLIDE 53

mcmaster.ca

|

Backlog ratio

53 June 26, 2020

(test/day) (test) Backlog newTests ~ X days to clear the backlogs

Backlog ratio

slide-54
SLIDE 54

mcmaster.ca

|

Changes in Delays over time

54 June 26, 2020

Backlog ratio

slide-55
SLIDE 55

mcmaster.ca

|

Summary

55

  • Early detection is key in controlling the spread of covid
  • Reduce delay -> earlier isolation/treatment
  • What components we can take back to the modelling world to improve

models

  • e.g. Hospitalization data
  • Duration spent in treatment
  • Combining different data streams to model

June 26, 2020

slide-56
SLIDE 56

mcmaster.ca

|

McMaster Pandemic

56

  • SEIR model
  • Including Hospitalization

time series

  • Using empirical delay/

duration distributions to parameterize the flow of the compartments

June 26, 2020

Additional features

  • mobility from google and apple

https://github.com/bbolker/McMasterPandemic

slide-57
SLIDE 57

mcmaster.ca

|

Distributions of delays associated with COVID-19 healthcare in Ontario, Canada

57 June 26, 2020

slide-58
SLIDE 58

mcmaster.ca

|

Final remarks

58

  • Easy to be tunnel visioned in the model world
  • Important to get a sense what is happening in reality
  • Bridging the knowledge gap in both directions (Public Health and Modelling)

June 26, 2020

slide-59
SLIDE 59

mcmaster.ca

|

Final remarks

59

  • Easy to be tunnel visioned in the model world
  • Important to get a sense what is happening in reality
  • Bridging the knowledge gap in both directions (Public Health and Modelling)

June 26, 2020

  • Small piece of the pie
slide-60
SLIDE 60

mcmaster.ca

|

Acknowledgments

60

  • Dr. Jonathan Dushoff
  • Dr. Ben Bolker
  • Dr. David Earn

Sang Woo Park (Princeton University) David Champerdon (University of Western Ontario) Morgan Kain (Stanford University) Irena Papst (Cornell University) MacTheoBio Steve Cygu Chyun Shi Martin Stelmach McMaster University CIHR IIDR @ McMaster U PHAC PHO

June 26, 2020