Open or closed?
Agent-based Simulations of SARS-CoV-2 Prevention Measures in Austrian Schools

Graz University of Technology     |     Complexity Science Hub Vienna

Jana Lasser     |     jana.lasser@tugraz.at     |     @janalasser

slides: https://janalasser.at/talks/school_covid_talk_TU/


Motivation

Im autum 2020, keeping schools open or closing them has become a very ideological discussion in Austria:

... opening schools is no problem since children are less susceptible to the virus ...

... children get infected in schools and spread the virus to their families ...

... wearing masks and frequently airing the rooms is harmful for children ...

... keeping schools closed is harmful for children ...

Scientific questions

We aimed to provide scientific evidence to enable informed decisions:

Can outbreaks in schools be controlled with non-pharmaceutical intervention measures at all?

What measures work best?

How many measures are necessary?

Are children less infectious?

Modelling SARS-CoV-2 in schools

Modelling SARS-CoV-2 in schools

Modelling SARS-CoV-2 in schools

Modelling SARS-CoV-2 in schools

Model of the infection

Model of the infection

Model of the infection

Model of the infection

Model of the infection

Model of the infection

Model of the infection

Model of the school

Model of the school

Model of the school

Model of the school

Model of the school

Model of the school

School types

Type # classes # students # teachers
primary 8 19 12
primary with daycare 8 19 16
lower secondary 8 18 20
lower secondary with daycare 8 18 24
upper secondary 10 23 29
secondary 28 24 70

Contact types

Interaction Duration Proximity Type
household very long very close household
table neighbours long close K1
teachers, long meeting long close K1
teaching, supervision long close K1
classmates, daycare mates long far K2
teachers, short meeting short close K2

simulation of infection dynamics

















Calibration

Ratio of asymptomatic cases in schools.
Contact tracing data from AGES.

Remaining free parameters:

Transmission risk for K1 contacts

Transmission risk for K2 contacts

Age dependence of transmission risk


Calibration using observational data.

Remaining free parameters:

Transmission risk for K1 contacts

Transmission risk for K2 contacts

Age dependence of transmission risk


Calibration using observational data.

Cluster tracing by AGES

A cluster traced by AGES contains (i) the setting (ii) the index case (iii) the infection chain and (iv) demographic characteristics of all involved people.

Cluster tracing by AGES

A cluster traced by AGES contains (i) the setting (ii) the index case (iii) the infection chain and (iv) demographic characteristics of all involved people.

Cluster tracing by AGES

A cluster traced by AGES contains (i) the setting (ii) the index case (iii) the infection chain and (iv) demographic characteristics of all involved people.

Cluster tracing by AGES

A cluster traced by AGES contains (i) the setting (ii) the index case (iii) the infection chain and (iv) demographic characteristics of all involved people.

Observational data


Data from AGES cluster tracing.

Observational data


Data from AGES cluster tracing.

Why do we trust this data?

The data we use was collected in September and Oktober 2020.

At the time, Austria performed rigorous testing of all K2 contacts (until the system was overhwelmed at the beginning of November).

Children seem to be well represented in this sample, even though they tend to be asymptomatic.

AGES follows up on infections to determine whether people stay asymptomatic.

Calibration – Results

A K1 contact has a 6.3% chance to transmit an infection.

A K2 contact has a 5.5% chance to transmit an infection.

Children are 2% less likely per year younger than 18 to transmit an infection.

Recap

We have a calibrated model

of different school types

that we can use to test interventions.

Interventions

Interventions

Interventions

Interventions

Interventions

Results

Single measures

Single measures

Single measures

Single measures

Single measures

Single measures

Single measures

Single measures

Measure combinations

Measure combinations

Measure combinations

sensitivity of results to parameter choices

Efficiency of individual measures is very uncertain.

Class size reduction: How well does it work?

Masks: Are they worn correctly?

Ventilation: How efficient is it really?

Preventive testing: How many participate voluntarily?

Test technology: How sensitive are the tests?

Virus: Mutants with higher infectivity?

Linear decrease in sensitivity leads to exponential increase in cases.

Scenario 1: Conservative assumptions about measure implementation

  • AG test sensitivity: 40% (instead of 100%)
  • 50% participation in voluntary tests (instead of 100%)
  • Only 30% of students stay at home (instead of 50%)
  • Room ventilation reduces transmission risk by 20% (instead of 64%)
  • Masks reduce transmission risk by 40% [20%] for exhaling [inhaling] (instead of 50% [30%]).

    Scenario 2: Mutant with increased transmissibility

  • Keep optimistic assumptions about measure implementation.
  • Increase the base transmission risk β by 50%.

    • Scenario 1: Conservative assumptions about measure implementation

  • AG test sensitivity: 40% (instead of 100%)
  • 50% participation in voluntary tests (instead of 100%)
  • Only 30% of students stay at home (instead of 50%)
  • Room ventilation reduces transmission risk by 20% (instead of 64%)
  • Masks reduce transmission risk by 40% [20%] for exhaling [inhaling] (instead of 50% [30%]).

    Scenario 2: Mutant with increased transmissibility

  • Keep optimistic assumptions about measure implementation.
  • Increase the base transmission risk β by 50%.

    • Single measures

    Baseline: scenario with literature values

    X: X-fold increase of mean outbreak size over baseline

    R: Number of transmissions from the index case

    Single measures

    Baseline: scenario with literature values

    X: X-fold increase of mean outbreak size over baseline

    R: Number of transmissions from the index case

    Single measures

    Primary schools seem to be safe, other school types depend on the scenario.

    Summary of results

  • Children are slightly less infectious the younger they are.
  • Measure combinations can prevent frequent and large outbreaks.
  • Risk in primary schools is lower, mainly due to org. aspects.
  • Testing and ventilation are sufficient to contain (most) outbreaks.
  • Results are highly sensitive to bad measure implementation and more infectious variants.

    • Publication preprint:
    https://doi.org/10.1101/2021.04.13.21255320

    Simulation package "small comunity SEIRX" (Python):
    https://pypi.org/project/scseirx

    Application to schools: https://github.com/JanaLasser/school_SEIRX

    Application to nursing homes: https://github.com/JanaLasser/nursing_home_SEIRX

    slides:
    https://janalasser.at/talks/school_covid_talk_TU/