Calibrating school models to cluster tracing data

Graz University of Technology     |     Complexity Science Hub Vienna

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

Slides available at https://www.janalasser.at/talks/calibrating_school_models/

Agent based modelling of SARS-CoV-2 in schools

Agent based modelling of SARS-CoV-2 in schools

Agent based modelling of SARS-CoV-2 in schools

Agent based modelling of SARS-CoV-2 in schools

Agent based modelling of 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

School co-location network

School co-location network

School co-location network

School co-location network

School co-location network

Different school types


school 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

Source: Austrian school statistics.

simulation of infection dynamics















The data

536 clusters* with 3342 cases recorded in Austrian schools between
2020-08-31 and 2020-11-02 collected by AGES.

Age School type Clusters Cases
< 10 years primary 67 286
10-15 years lower secondary 180 762
> 15 years upper secondary 116 388
> 10 years secondary 70 810
otherwise inconclusive 103 1097

*"school cluster": at least two cases of which at least one transmission ocurred in a school context.

Data available at https://doi.org/10.5281/zenodo.4706876

Cluster size distributions

Student case distributions

Ratio of asymptomatic cases*

*Follow-up to exclude initially pre-symptomatic cases.

Conditions in autumn 2020

  • Infection detected: isolation and quarantine of all K1 contacts.
  • Rigorous testing of K2 contacts: all students in the same class.
  • No mask mandate in schools yet.
  • No air ventilation, preventive testing or class cohorting.
  • Strict measures to prevent mixing between classes.
  • Original SARS-CoV-2 strain was circulating.
  • Aim: build a model of infection transmission in Austrian schools that reflects this data a closely as possible.

    Use the calibrated model to test the effectiveness of additional intervention measures.

    Calibration of household SAR

    [1] Madwell et al. 2020 Household Transmission of SARS-CoV-2 A Systematic Review and Meta-analysis.

    Calibration of household SAR

    [1] Madwell et al. 2020 Household Transmission of SARS-CoV-2 A Systematic Review and Meta-analysis.

    Remaining free parameters

    Assumption: q2 = q3 = qage.

    Model qage as linear decrease in infection risk for every year younger than 18.

    Error term to optimize


    Simulations for all school types:
    (-) Draw source cases from known distribution of teachers and students.
    (-) Use known age-dependence of asymptomatic courses.
    (-) Simulate with known conditions (only TTI) at data collection time.

    Calibration of school contact & age dependence

    Calibration of school contact & age dependence

    Contact weight: 0.30 [0.26; 0.34].

    Age dependence: -0.005 [-0.0225; 0.0] per year younger than 18.

    Interventions

    Results (delta)

    Results (delta)

    Results (delta)

    Results (delta)

    Results (delta)

    Publication 1: Lasser et al. 2022. Assessing the impact of SARS-CoV-2 prevention measures in Austrian schools using agent-based simulations and cluster tracing data. Nature Communications 13:554

    Publication 2: Lasser, Hell, Garcia 2022. Assessment of the effectiveness of Omicron transmission mitigation strategies for European universities using an agent-based network model. Journal of Clinical Infectious Diseases, accepted

    Team: Peter Klimek, Johannes Sorger, Lukas Richter, Stefan Thurner, Daniela Schmid

    Simulation package: https://pypi.org/project/scseirx

    Data: https://doi.org/10.5281/zenodo.4706876

    Slides: https://www.janalasser.at/talks/calibrating_school_models/