Model-based projections for COVID-19 outbreak size and student-days lost to closure in Ontario childcare centres and primary schools

There is a pressing need for evidence-based scrutiny of plans to re-open childcare centres during the COVID-19 pandemic. Here we developed an agent-based model of SARS-CoV-2 transmission within a childcare centre and households. Scenarios varied the student-to-educator ratio (15:2, 8:2, 7:3), family clustering (siblings together versus random assignment) and time spent in class. We also evaluated a primary school setting (with student-educator ratios 30:1, 15:1 and 8:1), including cohorts that alternate weekly. In the childcare centre setting, grouping siblings significantly reduced outbreak size and student-days lost. We identify an intensification cascade specific to classroom outbreaks of respiratory viruses with presymptomatic infection. In both childcare and primary school settings, each doubling of class size from 8 to 15 to 30 more than doubled the outbreak size and student-days lost (increases by factors of 2–5, depending on the scenario. Proposals for childcare and primary school reopening could be enhanced for safety by switching to smaller class sizes and grouping siblings.


Tables of the various scenarios considered
The scenarios considered in this study are combinations of factors potentially affecting the speed and efficiency of transition within both childcare centres and primary schools. In Column 1 of Tabs. S1 and S2, the case of 'high' transmission rate encapsulates behaviours facilitating disease spread (such as close contact, insufficient disinfection and ventilation of student spaces, etc), while the 'low' transmission case representing obedience to safety guidelines (such as physical distancing. mask usage, hand disinfection and other such measures). For considering different class sizes and composition, we change the numbers and ratios of students and teachers in each classroom of the centre or school (Column 2 of both Tabs. S1 and S2). Also common to both childcare centre and schools is the variation of the duration of the school day, as shown in Column 4 of both Tabs. S1 and S2. Students can either attend class for some typical duration (full), or can instead spend less time in class to lessen the number of contacts in the institution (reduced, B).
Childcare Centre (no student cohorts)  Table S1. Scenarios evaluated based on different assumptions about transmission probabilities, educator-student ratios, student assignment to classrooms and the duration of the school day in childcare centres.
Uniquely in childcare centres (in this study), students can be assigned to classrooms either without pattern (random assignment, RA) or by placing cohabiting students together where possible (siblings together, ST); this choice is shown in Column 3 of Tab. S1. Here, we assume that all students and teachers attend the childcare centre whenever possible. However, in the case of primary schools, we treat all classroom assignment of students as random, and instead model scenarios where students are placed in alternating cohorts (A) or allowed to attend class on all school days (no cohorting). This choice appears of Column 3 of Tab. S2.
Sensitivity Analysis: varying α 0 and B H The parameter β H represents the rate of interaction in the household, and thereby regulates the spread of the disease. For each value of α 0 , increasing the rate of interaction in the home β H increases the number of infections produces for both RA ( Supplementary Fig. S7) and ST ( Supplementary Fig. S8) assignment. In most scenarios (7:3 RA being one of the exceptions), varying α 0 (for constant β H ) produces a small increase in the number of infections produced throughout the simulation. The rate of increase also depends on the number of children in the classroom; for the scenario 31:1 RA, increasing β H from 0.0545 to its baseline value 0.109 almost triples the number of total infections.

Sensitivity Analysis -Varying α 0 and R init
The parameter R init refers to the proportion of individuals we presume are recovered from some previous period of infection spread, while α 0 is responsible for the rate of infection in common areas relative to the infection rate in the classroom. All other parameters are set to the baseline values given in Supplementary Tab. S4. These parameters were varied together by 50% in

2/15
Primary School (all students allocated randomly)   either direction. In Supplementary Figs. S9 and S10, increasing values of R init lower both the means and standard deviations of the total number of infections for each value of α 0 . Also, for each value of R init , the total number of infections produced increases with α 0 . This shows opposing interaction between increasing common area infection and increasing initial recovery rate; one increases infection and the other lowers it (respectively).
Sensitivity Analysis -Varying α 0 and λ i From Tab. S4, parameter λ i varies the amount of community infection in the model (infection due to other sources not modelled, such as public transport); be reminded that we assumed that the rate of community infection is effectively twice the baseline value for those individuals in the model not attending the school. For each value of α 0 in Supplementary Fig. S12, the total number of infections produced in the simulation increases with λ in each scenario with random assignment (RA), and also with grouping by household (ST, Supplementary Fig. S11). For each  λ , there is no consistent relationship between the numbers of infections and the value of α 0 . This result is intuitive; though the effect is not pronounced, increasing the rate of community infection increases the total number of infections in each tested scenario.  Table S3. Times at which the mean proportions of presymptomatic (P), exposed (E), symptomatically infected (I) and asymptomatically infected (A) school attendees peak during the first 30 days of simulation with secondary spread with respect to each of the scenarios tested, and the corresponding peak number of cases. β C transmission probability in classrooms β C = α C β H , 4 , assumption 3,4 , assumption

Supplementary Figures and Tables
infection rate due to other sources 1.16 × 10 −4 /day 5 , estimated R init initial proportion with immunity 0.1 assumption ξ probability of siblings attending same centre 0.8 assumption o proportion of childless educators 0.36 6 , assumption household size distributions 6   Table S4. Parameter definitions, baseline values and literature sources.    Text in boxes denotes the mean and standard deviation of the data corresponding to the parameters and error bars denote a single standard deviation of the data used.

12/15
Figure S10. Results of varying the parameters R init and α 0 by (50% each) on the total number of infections for RA assignment. Text in boxes denotes the mean and standard deviation of the data corresponding to the parameters and error bars denote a single standard deviation of the data used.

13/15
Figure S11. Results of varying the parameters λ i and α 0 by (50% each) on the total number of infections for ST assignment. Text in boxes denotes the mean and standard deviation of the data corresponding to the parameters and error bars denote a single standard deviation of the data used. 14/15 Figure S12. Results of varying the parameters λ i and α 0 by (50% each) on the total number of infections for ST assignment. Text in boxes denotes the mean and standard deviation of the data corresponding to the parameters and error bars denote a single standard deviation of the data used.