Quantifying threat from COVID-19 infection hazard in Primary Schools in England

Abstract

SummaryWe have constructed a COVID-19 infection hazard model for the return of pupils to the 16,769 state Primary Schools in England that takes into account uncertainties in model input parameters. The basic probabilistic model estimates likely number of primary schools with one or more infected persons under three different return-to-school circumstances. Inputs to the infection hazard model are: the inventory of children, teachers and support staff; the prevalence of COVID-19 in the general community including its spatial variation, and the ratio of adult susceptibility to that of children. Three scenarios of inventory are: the counts on 1st June when schools re-opened to Nursery, Reception, Year 1 and Year 6 children, when approximately one-third of eligible children attended; a scenario assuming a full return of eligible children in those cohorts; and a return of all primary age children, scheduled for September. With a national average prevalence, we find that for the first scenario between 178 and 924 schools out of 16,769 in total (i.e. about 1% and 5.5% respectively) may have infected individuals present, expressed as a 90% credible interval. For the second scenario, the range is between 336 (2%) and 1873 (11%) schools with one (or more) infected persons, while for the third scenario the range is 661 (4%) to 3310 (20%) schools, assuming that the prevalence is the same as it was on 5th June. The range decreases to between 381 (2%) and 900 (5%) schools with an infected person if prevalence is one-quarter that of 5th June, and increases to between 2131 (13%) and 9743 (58%) schools for the situation where prevalence increases to 4 times the 5th June level. Net prevalence of COVID-19 in schools is reduced relative to the general community because of the lower susceptibility of primary age children to infection. When regional variations in prevalence and school size distribution are taken into account there is a slight decrease in number of infected schools, but the uncertainty on these projected numbers increases markedly. The probability of having an infected school in a community is proportional to the local prevalence and school size. Analysis of a scenario equivalent to a full return to school with an average national prevalence of 1 in 1700 and spatial prevalence variations, estimated from data for late June, indicates 82% of infected schools would be located in areas where prevalence exceeds the national average. The probability of having multiple infected persons in a school increases markedly in high prevalence areas. Assuming national prevalence characteristic of early June, individual, operational and societal risk will increase if schools reopen fully in September due to both increases in numbers of children and the increased challenges of sustaining mitigation measures. Comparison between incidents in primary schools with positive tests in June and July and our estimates of number of infected schools indicates at least an order of magnitude difference. The much lower number of incidents reflects several factors, including effective reduction in transmission resulting from risk mitigation measure instigated by schools.