Research Collection
Presentation
Tracing the pandemic in agent-based models Status and plans
Author(s):
Axhausen, Kay W.
Publication Date:
2021-03
Permanent Link:
https://doi.org/10.3929/ethz-b-000474171
Rights / License:
In Copyright - Non-Commercial Use Permitted
This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.
Preferred citation style
Axhausen, K.W. (2021) Tracing the pandemic in agent-based models:
Status and plans, presentation at the Hong Kong Polytechnic University
Smart Cities Research Institute, online, March 2021.
Tracing the pandemic in agent-based models: Status and plans
KW Axhausen IVT
ETH
Zürich
Collaborators and partners NFP78
ETH Zürich
• M Balac
• G Kagho
• S Pennazi
• A Sallard
ETH Zürich (Thomas van Boeckel) FHNW (Erath Alexander)
University Basel (Melissa Penny) University Geneve (Olivia Keiser) Senezon (Michael Balmer)
Kai Nagel (TU Berlin), Zbigniew Smoreda (ORANGE), Sebastian
Bonhoeffer (ETH Zürich)
Acknowledgements
Sebastian Müller,. TU Berlin
Steffano Penazzi, ETH Zürich
Joseph Molloy, ETH Zürich
COVID19 in Switzerland – Some numbers
MOBIS/COVID19 sample since March 2020
Share of mobile persons by day since September 2019
0 100 200 300 400
0%
25%
50%
75%
100%
0 28 56 84 112 140 168 196 224 252 280 308 336 364
N ew c o n fi rm ed h o sp ti al iz ed c as es
M o b ile p e rs o n s [% ]
Day of year
MOBIS control group COVID19 2020
COVID19 2021
Hospital cases 2020
Hospital cases 2021
Weekly average number of trips and size of activity space
0 1 2 3 4 5
0 100 200 300 400 500
02.03 02.05 02.07 02.09 02.11 02.01
N u m b er o f tr ip s/ d ay
A ct iv it y sp ac e [k m 2 ]
Week starting with Monday
Weekday - Activity space
Weekend&holidays
Weekday - Trips
Weekend&holidays
Δ% change in PKm by work location: WFH impact?
-75%
-50%
-25%
0%
25%
02.03 30.03 27.04 25.05 22.06 20.07 17.08 14.09 12.10 09.11 07.12 04.01 01.02
% c h an ge in P K m a ga in st b as el in e
Home office
Mixed
On-site
How to assess the measures against a pandemic ?
How to assess ..? Model types
• Aggregate SEIR (Susceptible-Exposed-Infectious-Removed)
• Agent-based with contact networks SEIR
• Aggregate networks of co-presence in vehicles
• Repeated networks of co-presence based on agent-based transport
models
How to assess ..? Model types
• Repeated networks of co-presence based on agent-based transport models
Smieszek, T., M. Balmer, J. Hattendorf, K.W. Axhausen, J. Zinsstag and R.W. Scholz (2011) Reconstructing the 2003/2004 H3N2 influenza
epidemic in Switzerland with a spatially explicit, individual-based
model, BMC infectious diseases, 11 (1) 1-18.
How to assess ..? Networks of co-presence
Co-presence is defined by
• A space
• A starting time
• A joint duration
• Other persons present for the joint duration
How to assess ..? Networks of co-presence
How to assess ..? MATSim
MATSim delivers
• An agent-population
• The network of co-presence for one day (at activities and “in vehicle”)
• The computing infrastructure to replan the days
How to assess ..? Simple measures
Prohibit or limit co-presence by
• Quarantine
• Home-office; home-schooling
• Limits on group size
• Range restrictions
How to assess ..? episim
episim delivers
• The repeated instances of the same day
• The tracing of the person and its health status
• The possible infection process during joint activities
• The simple measures by removed the affected agents from the
activities
How to assess ..? Episim’s infection model
episim delivers
• Repeated instances of the same day
• Tracing of the person and its health status over the days
• Possible infections during joint activities
• Simple measures by removing the affected agents from the
activities
How to assess ..? Episim’s infection model
episim assumes a random process among 4 randomly chosen agents:
𝑃
𝑖𝑛𝑓𝑒𝑐𝑡𝑖𝑜𝑛= 1 − 𝑒
− Θ・ContactIntensity ・Shedding・Intake・Duration・Susceptibility・ Infectivity・OutdoorFactor• ContactIntensity ∼ Flow per agents
• Shedding, Intake = 1, unless mask wearing
• Duration = Activity time from MATSim
• Susceptibility, Infectivity = 1, but reduced for children
• OutdoorFactor = 1, unless outdoor activity, e.g. dining
How to assess ..? Progression process after infection
Exposed Infectious
Symptomatic
Seriously sick Critical
Seriously sick
Recovered
Recovered
Recovered
Recovered
[3.5 (3.5)]
[2.0 (2.0)]
[4.0 (4.0)]
[1.0 (1.0)]
[21.0 (21.0)]
[4.0 (4.0)]
[8.0 (8.0)]
[14.0 (14.0)]
[3.5 (3.5)]
[7.0 (7.0)]
[Median (st. dev.)]
Switzerland implementation
Calibrated model, well used in previous projects
• OSM network
• Public transport timetables (SBB and all regional providers)
• STATPOP population (registry data)
• Travel diaries (Microzensus 2015)
• Validated against counts, modes*distance
Switzerland: Mask adoption factor
Switzerland: Imported cases (calibrated)
Infectiousness
• SARS_COV_2 = 1.0
• B.1.1.7 = 1.5
• B.1.351 = 1.5
• P.1 = 1.5
Switzerland: Calibrated model results
Switzerland: Seriously sick by example region (canton)
Switzerland: Average ICU cases by hospital
Switzerland: Vaccination strategy – ICU patients
3000 6000 9000 12000
Persons/day
Next steps: Hypertension + diabetes
Age