A total of 2,519 students participated in the study at one elementary school (grades K-6), a combined school (grades K-8), and a high school (grades 9-12) from the charter district–District A, and from five elementary schools (grades K-4) and an intermediate school (grades 5-6) in the public school district–District B (Table 1 and S0). As there were only 12 students in the 12th grade, we combined grades 11 and 12 for all analyses (termed ‘11+’). We achieved good participation rates across the schools: 95.2% in school A1, 99.6% in school A2; 96.9% in school A3; 76.5% in school B1; 89.6% in school B2; 85.0% in school B3; 90.0% in school B4; 84.5% in school B5; 86.8% in school B6. An additional three students dropped out during the study period. After accounting for children for whom we could not achieve follow-up following an absence (see Absenteeism section below), we were able to successfully surveil 2,077 children during the study period.
In total, 1,772 children generated 4,720 absence events during the study period. We were unable to successfully call the home and determine the health status of 442 (24.9%) absent children who did not present to the school nurse with illness; collectively, they were responsible for 970 (20.6%) absence events. Our ability to successfully characterize illness in children following absence was greatest for schools in district A and among younger children (Table S2), We were able to characterize fully 3,750 absence events absences among 1,330 children. Older children (grades 7-12) had fewer absences with ascertained cause (i.e. observable for ILI outcome) (see Table 1).
Time course of ILI and confirmed infections
Of the children who were absent from school or reported to the school nurses, 408 were diagnosed with ILI and swabbed, of which 271 children were PCR positive for at least one virus. Some children had multiple ILI episodes and were swabbed more than once: 39 were swabbed twice, and six swabbed three times. Of those 271 children testing positive at least once with confirmed virus, 180 (66.4%) were positive for influenza (either subtype A or B), while 56 (20.7%) and 132 (48.7%) were positive for influenza A and B, respectively, during the surveillance period. A small wave of influenza A preceded a larger wave of influenza B cases, corresponding to distinct epidemics of A and B occurring in the wider community (Figure 1). A substantial proportion of children with were positive for respiratory pathogens other than influenza. Among the children testing positive for any virus, 64 (23.6%) were positive for coronavirus, 54 (19.9%) rhinovirus, 28 (10.3%) respiratory syncytial virus (RSV), 6 (2.2%) adenovirus, 3 (1.1%) picornavirus, and 3 (1.1%) metapneumovirus.
Cumulative attack rates (CARs)
In our study, we found an overall cumulative attack rate of 19.7% (95% CI, 18.0-21.5%) for ILI, 13.1% (11.7-14.6%) for infection with a respiratory virus, and 8.7% (7.5-10.0%) for influenza infection. We found significant variation in CAR between grades and schools for children identified with ILI, children testing positive for any of the viruses tested, and children testing positive for influenza (Figure 2; Figures S1, S2, S3). The highest rate of ILI was in grades 1 (26.6%, 95% CI 21.4%-32.3%) and 2 (25.3%, 20.5%-30.7%), while the rate ranged between 10.3% (95% CI, 7.0%-14.4%) and 30.7% (24.2%-37.8%) between schools. Children in grade 1 had the highest CAR for all viruses (20.6%, 95% CI 15.9%-26.0%), and the rate decreased with increasing grade, although kindergarten children had a lower CAR than children in grade 1 (8.2, 95% CI, 5.3%-12.1%). We also found significant differences in CAR for infection by any respiratory virus between the schools (Figure 2B). For influenza, similar patterns were observed: there were significant differences in CAR between grades and between schools, with the highest CAR being in younger aged children in grades 1 to 4, and CAR ranging in individual schools from 3.2% (95% CI, 1.5%-6.0%) to 17.4% (95% CI, 12.4%-23.4%). There were different patterns of CAR for influenza A and B (Figure S1) and other respiratory viruses across grades and schools (Figure S2), and across grade within the same school (Figure S3).
Long delays between symptom onset and sampling may introduce bias in the surveillance of acute respiratory viruses, as viral shedding rates diminish as individuals recover. While the delay between symptom onset and swabbing in our study ranged between 0 and 22 days, the median delay was 4 days and 83% of samples were taken within 6 days of symptom onset. Overall, we found no strong indication that positivity rates declined with increased delay in sampling (Figure S4, Table S8). We also found no secular trends in sampling delays (Figure S5).
CARs are a crude measure of true infection rates, due to the potential for confounders and the pooling of data across all participating schools. To explore the impact of multiple factors in influencing the infection rates of children, we fitted a series of mixed-effect regression models that explicitly incorporated the hierarchical nature of the observations within schools and districts. We tested whether case status (either having ILI, any virus, influenza or specific influenza subtype) was associated with the sex, school grade, self-reported vaccination status of students, or the duration of their instruction at school (full day vs half day). Half day duration only occurred in one of the school districts studies (district B).
We found broad agreement in the effect sizes of covariates and random variables across the five modelled outcomes (Table 2). Across all the different models, there was greater variation between schools than between districts. The proportion of variation associated with schools ranged from 11.2% (respiratory virus infection) to 26.5% (influenza B infection). Increasing school grade, a proxy for age, was associated with a reduction in risk in all outcomes. We did not find a significant effect of sex in any of the models. Vaccination status was not associated with viral infection risk but was associated with a reduction in the risk of ILI. Kindergarten children attending school for a half day were at significantly reduced risk of all infection outcomes except for influenza A infection. Half day attendance was most strongly associated with the risk of influenzas B infection, representing a reduction in risk of between 55% and 92%. These associations were supported further when we used a model selection process to consider alternative model formulations (Tables S3 and S4; supplementary information). An additional variable, percentage of students eligible for free or reduced price lunch, was selected in final models for all outcomes except influenza A. The association in all these models suggests students in schools with higher percentages of students eligible for free or reduced price lunch tend to be at reduced risk of ILI, respiratory virus and influenza B infections.
Additional analyses of half-day kindergarten attendance
To further explore the impact of half-day school attendance, we fitted models which explore the association between grade (Table S5) and duration of attendance (Table S6) and infection outcome in schools with half day kindergarten attendance. We also fitted models of infection outcomes with variables for attendance duration, grade, sex and vaccination status, in schools which have both full day and half day kindergarten students (Table S7). We found Kindergarten students at significantly reduced risk of infection compared to higher (older) grade students in schools with half day attendance (Table S5). However, in the same schools, an analysis that did not include grade and looked only at half day attendance, we did not find a statistically significant reduction in risk associated with half day attendance (Table S6). Thus, our results indicate that accounting for decreasing risk with increasing grade was important to detecting a statistically significant effect associated with half day status. When we fitted the full model (including terms for attendance, grade, sex and vaccination status) to the four schools in our study with both half and full day kindergarten students, we found half day kindergarten students were at significantly reduced risk of all infection outcomes (Table S7), though we did not find a significant association between infection risk and grade in these models.