Outcome	Number of studies	Overview of effect by study	Comparison used in each study	Effect direction per study (positive ▲; negative ▼; no change/mixed effects/conflicting findings ◀▶)
Outcome category: transmission‐related outcomes
Number or proportion of cases	7 modelling studies (Cohen 2020; Di Domenico 2020a; Head 2020; Lyng 2020; Panovska‐Griffiths 2020a; Tupper 2020; Williams 2020)	Cohen 2020: in the absence of any countermeasures in schools, 6% to 25% of teaching and non‐teaching staff and 4% to 20% of students would be infected with COVID in the first three months of school, depending upon the case detection rate. Implementing countermeasures that limit transmission and detect, trace, and quarantine cases within schools would lead to reductions in the cumulative SARS‐CoV‐2 infection rate among students, teachers, and staff over 14‐fold. Surveillance measures were implemented alongside classroom cohorting, face masks, physical distancing, and handwashing protocols in schools, so it is not possible to comment on the impact of these measures alone.	Full opening of schools with no measures in place	Positive ▲
		Di Domenico 2020a: the authors provide no effect estimates for testing, tracing and isolation, but provide results in a graphical way for both assumptions (25% and 50% case isolation through a 90% reduction of their contacts, simulating the result of rapid and efficient tracing and testing of cases). A sensitivity analysis performed indicates that a 25% case isolation compared to a 50% case isolation leads to a reduction in the daily number of new clinical cases under moderate social‐distancing interventions in all scenarios, except for the 100% reopening with no measures in place. Additionally, the effect is moderated by the relative transmissibility of pre‐school and primary school children: in scenarios with lower transmissibility in these younger children, fewer cases occur with each measure in place.	Full opening of schools with no measures in place	Positive ▲
		Head 2020: when faculty and/or students are tested (85% sensitivity, 100% specificity) on a weekly or monthly basis, with positive cases isolated, and their class quarantined for 14 days, the study predicted that in the absence of other interventions, testing and isolation/quarantine strategies have low effectiveness. When combined with strict social‐distancing measures, a modest reduction in community cases is possible as infectious individuals and their contacts identified in the school environment are quarantined (i.e. have their community contacts reduced by 75% for 14 days). The excess proportion of infections in teachers when only testing is employed is 37.77 (95% CI 10.64 to 53.31) compared to a scenario in which testing is combined with cohorting and mask wearing 1.45 (95% CI ‐2.36 to 5.69), compared to students where it is 52.07 (16.82 to 69.12) and 3.18 (95% CI 0.2 to 7.16), respectively. The excess proportion of infections in the community is 1.01 (95% CI ‐0.78 to 2.97). The effect sizes are moderated by the model parameters such as relative susceptibility and infectiousness of children, and extent of community transmission amid reopening. For weekly versus monthly testing, results are presented only in graphical form, but indicate that there would be a higher proportion of students with symptomatic infection with a monthly testing strategy than with a weekly testing strategy.	Single intervention component	Positive ▲
		Lyng 2020: at sensitivities of 98%, the models predict that a 2‐day delay in results will result in a just a 59% reduction in infections experienced at a 14‐day testing frequency. As the testing frequency is increased, the number of missed infections reduces rapidly by > 99% from no testing at all to a daily testing frequency, even with the 2‐day delay. Increasing testing frequency was associated with a non‐linear positive effect on cases averted over 100 days. While precise reductions in cumulative number of infections depended on community disease prevalence, testing every 3 days versus every 14 days (even with a lower sensitivity test) reduces the disease burden substantially.	Least intense measure	Positive ▲
		Panovska‐Griffiths 2020a: the study suggests that it might be possible to avoid a second pandemic wave if enough people with symptomatic infection can be tested, and contacts of those diagnosed can be traced and effectively isolated. Assuming 68% of contacts could be traced, the study estimates that 75% of those with symptomatic infection would need to be tested and isolated if schools return full time in September, or 65% if a part‐time rota system were used. If only 40% of contacts could be traced, these figures would increase to 87% and 75%, respectively.	Least intense measure	Positive ▲
		Tupper 2020: in all scenarios, if individuals have not already been identified through the relevant protocol, transmission stops when symptoms begin, as symptomatic individuals do not attend (or they leave when symptoms arise). In scenario 4, the mean cluster size was reduced from 11.9 to 6.5 in the asymptomatic case, whereas the group and two group protocols reduce it to 8.3 and 7.5 students, respectively. Over all the scenarios, the whole class protocol reduced cluster sizes roughly in half, with the contact and two‐group protocols performing slightly worse.	Least intense measure	Positive ▲
		Williams 2020: isolating household members of individuals who experience symptoms is estimating to avert 2.22 times more symptomatic cases than not isolating them. The multiplicative effect is slightly higher for surveillance/test/quarantine scenarios and highest for cluster sampling on schools, where 3.37 times more symptomatic cases are averted by isolating household members.	Least intense measure	Positive ▲
Reproduction number	1 modelling study (Panovska‐Griffiths 2020a)	Panovska‐Griffiths 2020a: across two scenarios of school reopening and different tracing levels, the test–trace–isolate strategy would need to test a sufficiently large proportion of the population with COVID‐19 symptomatic infection and trace their contacts with sufficiently large coverage, for R to diminish below 1.	Least intense measure	Positive ▲
Number or proportion of deaths	2 modelling studies (Head 2020; Panovska‐Griffiths 2020a)	Head 2020: the excess proportion of deaths in teachers when only testing is employed is 8.12 (95% CI 0.00 to 47.85), compared to 0 for students and 0.5 (95% CI ‐2.72 to 3.68) in the community. The effect sizes are moderated by the model parameters, such as relative susceptibility and infectiousness of children, and extent of community transmission amid reopening.	Full opening of schools with no measures in place	Positive ▲
		Panovska‐Griffiths 2020a: while results are only presented in a graphical way, they imply that more intense test, trace, and isolate strategies would lead to lower death rates than less intense strategies.	Least intense measure	Positive ▲
Shift in pandemic development	4 modelling studies (Landeros 2020; Panovska‐Griffiths 2020a; Panovska‐Griffiths 2020a (Preprint); Williams 2020)	Landeros 2020: the study found that reopening with a surveillance programme in place may provide 10 to 12 weeks of continuous instruction with low‐infection risk. Infections after closing are driven by a lack of interventions outside of school; testing and isolation in this context can curtail this growth. In general, the results support the importance of testing and complete school closure in preventing a major disease outbreak after reopening. Overall, this model also shows that reduction of class density and the implementation of rapid viral testing, even with imperfect detection, have greater impact than moderate measures for transmission mitigation.	Full opening of schools with no measures in place	Positive ▲
		Panovska‐Griffiths 2020a: the time point at which R diminishes depends on the degree to which the test–trace–isolate strategy had been implemented and the combination of testing and tracing.	Least intense measure	Positive ▲
		Panovska‐Griffiths 2020a: test–trace–isolate models, combined with mask wearing in the community and secondary schools were modelled. Results suggest that there is a greater benefit of mandatory masks in secondary schools if the effective coverage of masks is high (30%). Under current testing and tracing levels (24% testing, 47% tracing) and masks’ effective coverage of 30%, the predicted second COVID‐19 wave would be less than half of the original wave if masks were mandatory in secondary schools, as well as used in community settings. The minimum testing levels necessary to avoid a second wave, under scaled up test–trace–isolate strategies is 8% to 11% less when masks are mandatory in schools than if they are not, depending on the effective coverage of masks. The simulations suggest that the time point at which R diminishes depends on the degree to which the test–trace–isolate strategy had been implemented and the combination of testing and tracing.	Least intense measure	Positive ▲
		Williams 2020: isolation of symptomatic, asymptomatic individuals, and their household members can delay the peak prevalence. As with numbers of cases, the largest delays in peak prevalence occur when household members are isolated along with symptomatic and known asymptomatic cases. When all known infected cases and their household members are isolated, this delays the peak prevalence by 74 days.	Least intense measure	Positive ▲
Outcome category: healthcare utilisation
Number or proportion of hospitalizations	1 modelling study (Head 2020)	Head 2020: reopening schools with a weekly or monthly testing strategy for teachers and students would lead to a higher number of hospitalisations than reopening under strategies to reduce contacts, such as stable cohorts or alternating attendance. The excess proportion of hospitalisations in teachers when only testing is employed is 162.47 (95% CI 0.00 to 588.24), compared to students 0.58 (95% CI 0.00 to 15.27), and the community 3.68 (95% CI ‐7.27 to 15.54). The effect sizes are moderated by the model parameters, such as relative susceptibility and infectiousness of children, and extent of community transmission amid reopening.	Full opening of schools with no measures in place	Positive ▲
Outcome category: societal, economic and ecological outcomes
Numbers of days spent in school	1 modelling study (Gill 2020)	Gill 2020: in the absence of a school closure policy, quarantine of classmates and bus mates of infected students are likely to reduce in‐person attendance for the typical student by about 10% in a school open full‐time with precautions. High‐community infection rates were predicted to be more disruptive to schools operating full‐time in person than to schools using hybrid approaches. Even at 100 reported community infections per 100,000 per week, the typical student in a hybrid secondary school can expect to miss only a very few days due to quarantine, while the typical student in a secondary school open full‐time with precautions might be sent home for about 15% of possible days due to quarantine. Delays in testing would have large effects in schools implementing no precautions: as testing turnaround time increases from zero to 10 days. Policies that close the school (for 3 days or 14 days) when infections are detected substantially reduce the total number of days that students can attend in person. These effects are larger in schools operating full‐time than in schools using hybrid approaches because schools using hybrid approaches experience fewer infections that lead to quarantines or closures. In secondary schools where students attend daily, and the community infection rate is at a moderate level (50 per 100,000 per week), closing the school for 14 days for each detected infection would be highly disruptive, such that the typical student would be able to attend only about half of all school days.	Least intense measure	Negative ▼
Resources	3 modelling studies (Campbell 2020b; Lyng 2020; Williams 2020)	Campbell 2020b: the study predicts that universal testing for at‐risk populations would cost CAD 1.3 billion for each round of testing. The status quo testing approach from 8 to 17 July 2020 was predicted to require 41,751 tests per day and required 755 nurses, 213 nursing assistants, 172 other healthcare professionals, 3261 clerical and nonclinical staff, and 721 laboratory staff (5122 personnel total). Testing of at‐risk groups, in particular testing all 6,012,144 students and employees in primary and secondary schools over 1.5 months would require an added 20,956 healthcare professionals, 22,950 clerical staff and 2462 laboratory staff, costing CAD 816.0 million. A strategy of actively testing large population groups who are at increased risk of acquiring SARS‐CoV‐2 is feasible and affordable in Canada.	Least intense measure	Negative ▼
		Lyng 2020: frequent testing strategies can reduce the rate of new infections compared to scenarios where there is no testing at all. A 98% sensitive test with no delay in results administered every 3 days with pooling, and no confirmatory test offered by the institution costs less than USD 1.50 per person per day, with high performance. The model demonstrates that frequency of testing, test sensitivity, turn‐around time, and the external community prevalence are all important factors to consider, and there is often more than one testing strategy to achieve the desired level of performance.	Least intense measure	Positive ▲
		Williams 2020: if household members of symptomatic cases are also isolated (without testing them), a much lower eight tests are required to avert each one symptomatic case. Moving to the symptomatic testing and quarantine (STQ) scenarios, using simple random sampling and isolating only the cases that test positive with STQ, results in 145 tests required to avert one symptomatic case. This decreases to a low of 16 tests to avert one symptomatic case for pooled sampling of 5‐person pools and seven tests for pooled sampling of 5‐person pools if household members are also isolated. Notably, the STQ scenario of pooled sampling of 5‐person pools is slightly more efficient than the current status quo of testing and isolating symptomatic cases. All other STQ scenarios are less efficient than the status quo. However, instituting even these less efficient STQ scenarios is likely to avert a substantial number of cases (as described above) and could be more cost‐effective than the emergency room visits, long‐term care, lost labour, and other economic costs of symptomatic cases and deaths.	Least intense measure	No change/mixed effects/conflicting findings ◀▶