Background
The 2009 influenza pandemic caused disruption to schools, businesses, and governmental entities. Numerous reports from US CDC and European countries document the central role of school-aged children in community-wide transmission of pandemic virus [
1‐
10]. In the United States and Australia, the onset of pandemic influenza incidence was linked to school opening dates [
11]. Children experience higher rates of infection, shed influenza virus longer than adults, and have social mixing patterns conducive to the propagation of respiratory viruses [
5,
12]. However, the risk distribution of influenza and other respiratory infections throughout childhood is poorly understood. Current influenza vaccination programs in the United States target individuals 6 months of age and older. If vaccine supply is limited, children 6–59 months of age are an age group of focus for vaccination [
13]. Programs in the United Kingdom target children 17 years of age and younger, with a program that phases universal coverage in beginning with younger children [
14].
A number of seasonal pathogens cause respiratory symptoms consistent with influenza-like illness (ILI) [
15]. Given the importance of school-aged children for influenza transmission, a better understanding of the etiology of respiratory infections among children, their associated burden, and the interaction of these infections with influenza is required to improve current public health strategies to reduce transmission of infections in schools and communities.
Despite the role of school-aged children in driving influenza epidemics, relatively little fine-grained analysis has been conducted to assess how different ages within the school demographic are affected by seasonal influenza [
16]. Factors other than age may also be important in determining risk of infection. Females have been shown to be at greater risk of influenza infection in adults [
17], but it is unknown if school-aged females are also at elevated risk compared to male pupils in the same schools. Small-scale spatial variation in influenza attack rate has been observed [
18], though it is unclear whether nearby schools (within the same district) experience similar epidemic dynamics. The acquisition of immunity during childhood is thought to drive patterns of age specific incidence of influenza [
19]; few detailed studies have examined how infection risk changes with age during childhood.
We conducted the SMART study (Social Mixing And Respiratory Transmission study) in two school districts in the Pittsburgh metropolitan area. This study included testing school-aged children for respiratory viruses following an ILI-related absence during the winter influenza season of 2012–2013 in a cohort of schools. We relate infection outcomes to attributes of individual children and grade and classroom properties, as well as school districts and the individual schools within them.
Methods
Study population
We conducted our study in a charter school system in an urban area (district A: eight schools and 2000 kindergarten to 12th grade [K-12] students) and a public school district in a suburban area surrounding an urban core (district B: nine schools and approximately 4700 K-12 students). We conducted a purposive sample of schools to study a selection that broadly represented the diversity of socioeconomic and racial compositions of public schools across the Pittsburgh metropolitan area. School districts and schools were approached and those willing to participate were selected for inclusion. We approached only two school districts, and both agreed to participate. We worked with nine schools selected from these two districts: three from district A and six from district B.
Recruitment and study procedures
Project staff met with schoolboards, parent-teacher organizations, school staff, and school nurses to provide information about the project. Investigators provided students and parents a summary of the study, including a disclosure form and a signature section for opting out of participation. All parents received information sheets ahead of the study period, offering them the option to withdraw their child from the study. All participating children gave verbal assent prior to swabbing and interviewing.
Surveillance and virologic sampling
The 2012–13 influenza transmission season started unusually early nationwide [
20]. In schools participating in our study, influenza cases began to rise in mid-December. The SMART study team initiated surveillance in three schools during the week of December 17–21, 2012, prior to the scheduled winter break, and in the first week of January 2013. Surveillance activities were implemented in all nine participating schools from the start of the spring semester on January 7, 2013, concluding on March 27, 2013. Schools closed for winter break December 24, 2012 to January 2, 2013; spring break closure was March 28 to April 1, 2013.
Trained project staff were deployed in participating schools to monitor student absentee data and receive daily attendance reports from schools. We defined an absence event as absence from school by a student for either an entire day or missing part of the day due to leaving school early due to illness. This excluded individuals who were tardy but present for part of the day. Students with influenza-like illness at school either self-identified or were identified as ill by their teachers and sent to a school nurse. School nurses were present in all schools. School nurses contacted the parents of ill children to coordinate treatment as well as to confirm parental assent for swabbing of their children. Children of assenting parents were swabbed for virological testing either on presentation to the school nurse when ill or on return to school when ILI was identified through contact of parents. All contiguous days for which a student was absent were classified as a single absence event. Study staff telephoned parents of absent children to determine the reason for student absences. If an absent student was reported ill, staff then inquired about the symptoms the student had (categories recorded were: fever, sore throat, cough, runny nose/congestion, headache, muscle or joint pain, nausea/diarrhea/vomiting). Students with ILI using the CDC Case Definition - fever of at least 37.8 °C, and either cough or sore throat - were eligible for nasal swabbing by project or school staff upon return to school. Temperature was requested and recorded for those individuals who reported a recorded temperature, however, to increase sensitivity of case finding, children that a parent or guardian reported as having a fever, even in the absence of a measurement were included as meeting our case definition. Students presenting to the school nurse with ILI were swabbed immediately, otherwise students were swabbed on their return to school following illness absence.
Sampling and laboratory analysis
Staff used sterile wound polyester-tip swabs to collect specimens from the anterior nares of participants; the swab was placed in sterile transport media, and transported to University of Pittsburgh Medical Center Clinical Virology Laboratory, where PCR tests were performed for; influenza A/H1 and influenza A/H3 (Influenza A), influenza B (Influenza B), RSV A and B (RSV), rhinovirus (HRV), coronavirus 229E, NL63, HKU1 and OC43 (CV), and adenovirus B, C and E (ADNO) (see supplement for further details).
Questionnaire
Participating students were surveyed on a repeated basis to determine influenza vaccine status.
Statistical analysis
Cumulative attack rates (CARs) were calculated from pooled data across all schools and the entirety of the surveillance period. CARs were calculated for participants who were identified as: 1) having ILI, 2) testing positive for any of the respiratory viruses in the panel, or 3) testing positive for influenza virus, regardless of subtype and testing positive for each of subtypes of influenza separately. Denominator values included the number of participating students, students with absence periods who reported to the school nurse, and students with absence periods who did not report to the school nurse but whose home we were able to contact. We excluded students whose homes we were not able to contact to determine their health status even after multiple attempts. Binomial confidence intervals were calculated for all CARs.
We performed a variety of generalized linear mixed-effect regressions, in which outcome variables were the binary infection status of a participant over the entire surveillance period given a particular definition of outcome: whether they were identified as having ILI; testing positive for any respiratory virus; testing positive for influenza; or testing positive for a specific subtype of influenza (A and B). We included hierarchical random intercept terms for school and district in all models unless specified otherwise, where schools were nested within one of the two districts. We performed hypothesis-driven regression models, in which the variables included in the model were decided a priori. These included: a linear term for a child’s school grade (to explore the effect of increasing age, where Kindergarten grade is replaced with a zero and grade treated as a numeric term), attendance duration (full-day or half-day, to explore the effect of reduced time spent at school), sex (to identify any sex difference in risk within this age group), and vaccination status (to estimate vaccine efficacy). Random effect terms for school and district were included to identify variation in risk between schools and districts. To assess the robustness of our hypothesis-driven modelling, we also performed an exploration of alternative models using a forward selection process in which the final model’s variables were selecting from a series of candidate variables including those listed above as well as a the percentage of students eligible for free or reduced price lunch (school level), social mixing rate and class size (student level) (described in supplement; Tables S
1, S
2, S
3 and S
4).
Discussion
We conducted surveillance for influenza-like illness within nine schools during a single influenza season (2012–2013). Nearly 20% of surveilled children experienced at least one episode of ILI during the study period; several experienced multiple episodes. Influenza was the most common viral pathogen identified among these children, although we found the presence of other respiratory viruses in a substantial proportion of detected ILI episodes, suggesting that the burden of ILI in school age children during the normal influenza season may not be entirely due to infection with influenza. ILI and infections were concentrated in younger children, and CARs varied between schools. Modelling of the risk of infection with respiratory viruses and, specifically, influenza identified associations with the grade and the attendance duration of children, where younger children are at higher risk of infection than older children, and kindergarten children attending school for half days were at lower risk than kindergarten children attending for full day instruction.
Contacting parents of absent students to document reasons for absence was a major challenge during the study. However, our rate of follow-up for an absence episode (47.2% of cases) compared favorably to an earlier study in which only 28.6% of absences were verified [
21], though we did observe variation in documentation rates by school and grade. The primary problem in contacting parents was inaccurate phone numbers. This may be associated with the socio-economic status of the parents (and likely the school catchment). Nonetheless, the pattern of influenza-associated absences confirmed during the study period mirror emergency department and outpatient provider-reported, virologically confirmed influenza from the Allegheny County Department of Health, suggesting that absentee-triggered surveillance was an effective methodology (Fig.
1).
Our study found overall CARs for respiratory virus infection of 13.1% (95% CI, 11.7–14.6%) and for influenza infection of 8.7% (95% CI, 7.5–10.0%); these rates are consistent with those found in other studies of children in non-clinical settings [
16]. Our study also shows that the cumulative attack rate of respiratory viruses decreases with increasing grade of school-aged children. This age pattern of infection suggests that respiratory infectious disease prevention measures, including efforts to further increase seasonal influenza vaccine uptake, should be especially emphasized for the younger age groups: kindergarten through grade 6. We found the strongest association of grade and time spent in school on the risk of infection with influenza B; this was the outcome with the largest number of cases and, thus, the greatest power to detect significant associations. We found infection outcomes were clustered within schools but varied little between districts, after adjusting for other factors. This observation is consistent with a greater risk of influenza due to school-based rather than community-based transmission in our study population, though the clustering of other risk factors in schools could also explain our results. In the model selection process, an additional variable (percentage of students within a school eligible for free or reduced price lunch) was selected for several of the infection outcomes: lower percentages were associated with increased risk of ILI, respiratory virus and influenza B. This may reflect higher levels of immunity generated by greater exposure to viral pathogens earlier in life, or reduced detection of viral infections in our study through reduced access of study staff to parents to obtain reasons for absence or consent to participate. Our hypothesis driven models accounted for the potential for differences in infection rates between schools through random effects. The consistency of our estimated coefficients in hypothesis driven models and ‘agnostic’ selected models increases our confidence that school and district level random effects accounted for systematic differences between schools that may be driven by socio-economic factors or other unmeasured demographic variables.
Kindergarten students who attended school for half-day duration were at significantly lower risk of respiratory infection suggests the reduced risk of infection in kindergarten students may be due to reduced time spent within school. Further study is required to determine if these findings can be replicated in other settings, and if half-day school attendance would similarly affect respiratory viral infection in other grades. It is possible that our results reflect the experiences and exposures of kindergarten students within a single district and may not be generalizable to other kindergarten students or other age groups and school grades. Nonetheless, the prospect of a truncated school day as a potential alternative to full school closure, as a control option to reduce the transmission of seasonal and pandemic influenza [
4,
22], is intriguing. Pre-emptive, school closures are recommended as a countermeasure during severe influenza pandemics [
23], but are associated with significant educational, social and economic impacts [
24]. Prior studies of the impact of school closures on respiratory viral transmission have focused on observations around planned and unplanned closures [
4,
6,
25], but none have characterized grade specific differences or half-day attendance. Further work is required to assess the efficacy of a half-day attendance regime in non-kindergarten grades and in other school populations on influenza infection and transmission.
There are several important limitations to this study. Our detection of virus relied on identification of an ILI episode within a child; we may, therefore, have underestimated the of incidence of viral infections, particularly if infections can be asymptomatic. While we found little evidence of an effect of vaccination status on infection status, we relied on the children to self-report their vaccination status. This may have introduced bias through misclassification error and masked any true effect vaccines may have had in altering individual infection risk. Our findings may also be biased due to differences in our ability to characterize illness in absent children from different ages and schools. Follow-up for absence was harder to achieve in schools with higher percentage students eligible for free or reduced-price lunches, indicating a possible socio-economic reason. ILI and infection in older children and in schools with students from poorer backgrounds may, therefore, be underestimated in our study. A further limitation is that we have not measured other variables that confound or better explain the observation that half-day kindergarten children are at reduced risk of infection. Half-day kindergarten enrollment may be associated with socioeconomic status of families that may in turn be associated with the risk of acquisition and/or detection of respiratory viral illness in our study. The availability of half-day kindergarten programs by school may also be associated with socioeconomic characteristics of our study locales that could be associated with our outcomes. Kindergarten children in particular schools were reported by the school as attending for either a half or full school day; we had no further information on the time actually spent in school or daycare by these individuals.
Acknowledgements
We would like to thank the children, parents, and staff of our study schools for generously supporting and participating in this study. We are grateful to the virology laboratory of Dr. Charles Rinaldo, Jr. at the University of Pittsburgh and to Arlene Bullotta, UPMC Clinical Virology Lab for processing viral specimens, and to Peter Diggle and Chris Jewell for statistical advice.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.