The safe return of students and teachers to schools during the transition of SARS-CoV-2 to endemicity necessitates an evidence-based public health approach. While previous studies indicated relatively low SARS-CoV-2 transmission in schools, particularly among younger students, outbreaks have been reported, especially in secondary schools and regions with high community transmission. The emergence of the highly transmissible delta variant further complicates the situation, particularly given the varying vaccine eligibility among different age groups. Countries employ various NPIs to mitigate in-school transmission, including mask-wearing, class size reduction, room ventilation, and school entry testing using rapid antigen (AG) tests. While modeling studies suggest the potential of frequent AG testing to offset lower sensitivity compared to PCR tests, there's limited evidence on the combined effectiveness of these measures, especially with the dominance of the delta variant and varying vaccination rates. Existing school cluster descriptions are often limited in scope or lack the detailed information needed for proper model calibration and validation. This study addresses these gaps by developing and calibrating an agent-based model to Austrian school cluster data to evaluate the effectiveness of NPI combinations under the conditions of the delta variant dominance and partial vaccination coverage.

The literature review section of the paper summarizes existing research on SARS-CoV-2 transmission in schools, highlighting the relatively low transmission rates observed in some studies, particularly among younger children. However, the review also acknowledges the documented occurrences of outbreaks, especially in secondary schools and areas with high community transmission rates. The increased transmissibility of the delta variant is emphasized as a significant factor influencing the effectiveness of preventative measures. The review covers the existing literature on different NPIs, including mask-wearing, class size reduction, improved ventilation and regular testing, but highlights the lack of comprehensive data on the combined effectiveness of these measures under various conditions (i.e., delta variant, partial vaccination). The study acknowledges the limitations of existing models and data sets, which often lack the detail and scale necessary for accurate calibration and validation.

The researchers developed an agent-based model calibrated to Austrian data on school-related SARS-CoV-2 clusters. The model incorporated six school types: primary, lower secondary, upper secondary, and secondary schools, with and without daycare. A cluster was defined as a group of at least two epidemiologically linked cases with at least one in-school transmission. The model coupled in-host viral dynamics with population dynamics on contact networks determined by school type, class size, and timetable. The contact networks were time-dependent and multi-relational, reflecting the varying intensity and type of contacts in different settings (classes, common areas, households). Viral dynamics allowed for a more accurate representation of testing strategies. The model was calibrated to reproduce realistic cluster sizes and the ratio of infected teachers to students. The transmission risk for household and school contacts, and its age dependence, was calibrated using data from autumn 2020. To adapt the model to the delta variant, the household transmission risk was multiplied by 2.25, reflecting the increased infectivity of the delta variant, and the incubation and latent periods were adjusted. The model evaluated the effectiveness of four NPIs (room ventilation, mask usage, student cohorting, and entry AG testing) and their combinations, along with different vaccination rates. The effectiveness of individual NPIs was estimated from existing literature. A reference scenario with only test-trace-isolate strategy was used as a benchmark. The impact of each NPI was assessed in terms of cluster size reduction and the reproduction number (R). The sensitivity of results to varying NPI stringency and vaccination rates was also analyzed. An online simulation viewer was created to communicate results effectively to various stakeholders.

The analysis of 616 clusters revealed that 40% involved only two cases, and 3% had more than 20 cases. The model demonstrated that combinations of NPIs alongside vaccination are essential to control SARS-CoV-2 spread in schools under sustained community transmission of the delta variant. Primary schools needed at least two NPIs, while secondary schools required at least three, even with vaccination of 80% of teachers and 60% of family members. Among individual NPIs, frequent AG testing among students yielded the most significant cluster size reduction. Room ventilation, class size reduction, and mask usage also contributed substantially. The study further explored various combinations of NPIs, showing that in scenarios where all NPIs (excluding testing) were implemented, only primary schools achieved R<1 (R being the reproduction number). The most effective strategies for controlling transmission included the combinations of room ventilation and frequent AG testing. The sensitivity analysis revealed an exponential increase in cluster size with reduced NPI effectiveness. In the “worst-case scenario” with low NPI effectiveness, even with vaccination, no NPI combination alone achieved R<1 in all school types. Vaccination significantly reduced transmission risk, with vaccination scenario II (80% of teachers, 60% of family members, and 50% of students vaccinated) achieving R<1 for all school types and source cases except for teacher source cases in secondary schools. The ratio of infections in schools relative to households decreased significantly with vaccination.

The findings highlight the necessity of combining multiple NPIs for effective SARS-CoV-2 control in schools, particularly secondary schools, where the more complex contact networks and older age of students contribute to higher transmission rates. The study’s results align with existing evidence suggesting that schools reflect community transmission dynamics, but with younger children contributing less to virus spread. The increased R value for teacher-source cases underlines the importance of targeted mitigation measures for teachers, especially due to their role as connectors between classes and extended speaking time. The sensitivity analysis underscores the critical role of consistent and stringent NPI implementation. While the model provides valuable insights, the reliance on preliminary literature for NPI effectiveness necessitates cautious interpretation of the results. Future research could focus on more accurate estimations of NPI effectiveness through rigorous field studies and refining the model to include additional factors like ventilation systems and social dynamics outside school settings.

This study demonstrates the importance of a multi-faceted approach to managing SARS-CoV-2 transmission in schools. The findings show that combinations of NPIs are needed, with the specific combination dependent on school type and vaccination rates. The effectiveness of these NPIs heavily relies on consistent and stringent implementation. Future work could refine the model to incorporate additional factors and validate the findings through further empirical research. The online visualization tool is a significant contribution, allowing for practical application and easier interpretation of the findings for stakeholders.

The study acknowledges several limitations. The age-dependent transmission risk remains a topic of debate, and potential biases in the data used for calibration due to variations in testing strategies are acknowledged. The model simplifies several aspects of school life, such as the time resolution of one day and the representation of various interactions. The generalizability of the findings to other countries is also limited by the focus on the Austrian context. Parameter choices for NPI effectiveness relied on existing literature, which constitutes a source of uncertainty. Finally, constant vaccine effectiveness and static vaccination rates were assumed in the model, which is a simplification compared to real world situations.