Our data demonstrate that ILI-SCs occur annually in the US with temporality that reflects the general patterns of influenza activity on both national and regional levels as observed through routine surveillance of medically attended ILI. In most seasons prior to the COVID-19 pandemic, school winter breaks temporarily reduced ILI activity with seasonal peaks usually following within a few weeks of the start of second semester. ILI-SC annual peaks occurred in the same timeframe, presumably once the number of ill students reached a critical mass a few weeks after school re-opening for spring semester. A similar pattern was exhibited around the winter break of the 2019/20 school year; however, the advent of the COVID-19 pandemic and subsequent preemptive closure of nearly all K-12 schools in the US likely impacted the remaining weeks of both the season’s ILI activity and related SCs. During the subsequent two COVID-affected years, 2020/21 and 2021/22, ILI-SCs reflected the temporality and low levels of ILI activity.

Prior to the COVID-19 pandemic, the highest numbers of ILI-SCs occurred in seasons when influenza A (H3N2) was predominant either nationally or regionally. Among these, the 2017/18 and 2018/19 seasons each reached a level of closures previously only seen in the 2009 pandemic where a reported 1,947 schools closed [11]. Meanwhile, exceptionally mild seasons, such as 2011/12 or those dominated by influenza A (H1N1) generally had far fewer closures.

However, the largest annual number of ILI-SCs during the study period occurred during the 2019/20 school year, which was affected by an intense seasonal outbreak predominated first by influenza B and subsequently by influenza A (H1N1). This record was set despite the school year being abruptly interrupted, by the implementation of measures used to mitigate the spread of SARS-CoV-2 as early as February 2020 and ultimately by the near simultaneous preemptive closure of all K-12 schools beginning in the week ending March 14, 2020 in order to slow transmission of SARS-CoV-2 [10,12].

In addition to variations in seasonal influenza epidemiology, the number of ILI-SCs per year may have been influenced by differences in seasonal vaccine effectiveness. The higher counts of ILI-SCs in H3N2-dominated years may, at least in part, be related to a relatively higher effectiveness of the influenza A (H1N1) vaccine component compared with the A (H3N2) component [13]. Furthermore, because the predominant circulating virus types and subtypes can vary geographically [6], vaccine effectiveness may also differ by locale.

Though vaccine effectiveness may be a contributing factor to seasonal distribution of ILI-SC occurrence, uneven geographic distribution could also have been related to variations in vaccine uptake or certain programmatic factors that prompt school authorities to close schools, which may be common to a state or locale.

There were observable regional differences in vaccine coverage across the study period. Among HHS Regions with relatively few ILI-SCs across the study period, particularly regions 1, 2, 3, 7, and 8, estimated vaccine coverage across the study period tended to be higher than national estimates. Meanwhile HHS Region 4, the region with the most ILI-SCs, reported vaccine coverage estimates consistently below the national-level [14]. This may suggest that seasonal vaccine uptake is a contributing factor in ILI-SC occurrence, and, while not a key driver, potentially has a bigger impact in some years than others.

There are likely also programmatic factors that impact school closure decisions. For example, regional variability could be affected by differences in state policies related to school funding or the varying costs to run schools. Student enrollment is typically a primary driver of funds provided to schools by the state, and, depending on how the number of enrolled students is calculated in each state, school closure decisions may be swayed by illness-related absenteeism given the connection between attendance and funding [15]. Additionally, factors such as the higher costs of food service and transportation in rural schools [16] may impact the cost-effectiveness of running differently located schools with high levels of illness-related absenteeism. More research is needed to analyze the effect of policies and programmatic factors on the reasons for and barriers to closure. Disproportionate burden of ILI-SCs appears to be concentrated in socio-economically disadvantaged communities of two HHS regions, which adds urgency to further evaluate and, where needed, address programmatic factors that may be influencing these trends.

During the COVID-dominated period (school years 2020/21 and 2021/22), ILI-SC data collection continued but remarkably few ILI-SCs were documented. The 2020/21 school year had among the lowest number of ILI-SCs, comparable only to the 2013/14 school year, a moderate influenza A (H1N1)-dominated season. Moreover, all documented ILI-SCs in the 2020/21 school year were simultaneously attributed to COVID-19, and the weight of each disease’s contribution to the closure decisions is unknown. The low number of observed ILI-SCs in the 2020/21 school year is credible, because the ILI activity that year did not breach the national baseline, a first during the study period.

While the 2021/22 school year saw an increase in ILI-SCs from the previous year, the numbers were only a fraction of the number of ILI-SCs in the three school years prior to the COVID-19 pandemic, all dominated by influenza A (H3N2). This pattern is consistent with the generally low ILI activity observed during the COVID-dominated years [12].

In response to the growing levels of community SARS-CoV-2 transmission in the summer and fall of 2020, many schools and districts across the country offered a variety of reduced-contact learning options in the 2020/21 school year including hybrid (partly in-person / partly distance learning) and fully virtual (online) education models, thereby limiting the density, frequency, and duration of in-person student congregation in schools [17]. In schools where in-person learning did take place, various measures such as social distancing, facemasks, and increased cleaning were recommended to reduce transmission of SARS-CoV-2 [18]. The nearly total return to in-person learning in the 2021/22 school year [19] coincided with an uptick in both ILI-activity and ILI-SCs.

ILI-SCs described in this multi-year study are reactive closures associated with increased ILI activity, i.e. they occurred as a consequence of increased disease transmission as reported in most school closure announcements. In contrast, during the 2009 Influenza A (H1N1) pandemic, both preemptive and reactive closures were implemented in the US and other countries [20]. Subsequently performed systematic literature reviews support pre-emptive school closures as an NPI to reduce virus spread in schools and surrounding communities [21, 22], but the effectiveness of closures implemented reactively remains unclear. Two US-based studies reported that reactive closures have little effect during seasonal [23] or pandemic [24] influenza outbreaks; conversely, studies implemented elsewhere indicate that transmission would be even greater if they are not implemented [25, 26], though some with only minor effect [26]. Reactive closures have for decades been routinely used in Japan and Russia to control outbreaks of seasonal and pandemic influenza where they are triggered locally by defined levels of heightened ILI-related student absenteeism [25, 27]. However, it has yet to be shown under what conditions reactive closures might help control local outbreaks in the US. Meanwhile, recent US-based research has shown that routine monitoring of cause-specific student absenteeism can act as an early warning signal of increased ILI activity in schools and surrounding community [28]. For communities which are disproportionally burdened year-after-year with ILI-SCs, it may be particularly worthwhile to explore more effective alternatives to late reactive ILI-SCs, even if only via better (earlier) timing of the interruptions of in-person instruction to obtain greater reduction of within-school disease transmission.

This evaluation is subject to at least four limitations. First, ILI-SC data only captured closures reported through publicly available online media sources, potentially missing announcements made exclusively through other communication means (e.g. text messages, listservs, etc.). Second, because data points for ILI-SCs were abstracted exclusively from publicly available announcements, some information may not have been complete or entirely accurate. However, the method of data collection imposed no reporting burden on schools and was relatively inexpensive. Third, data on outpatient healthcare provider visits for ILI and laboratory-confirmed influenza hospitalizations were subject to limitations that are characteristic of surveillance systems. Additionally, not all states were included in hospitalization data and therefore results are not generalizable. However, this was mitigated by the fact that Tennessee, the state most greatly impacted by ILI-SCS, was accounted for in the hospitalization data. Lastly, we compared the relationships of ILI-SCs and ILI surveillance data on national and sub-national levels. Such high level of comparisons may not have fully reflected relationships on the local level (i.e., communities around closed schools), and further studies exploring these relationships on the local level are warranted.

To our knowledge, this study is the first of its kind in the US, describing the frequency and characteristics of school closures attributed to seasonal influenza/ILI over eleven consecutive influenza seasons, and thereby establishing an inter-influenza-pandemic baseline of such closures. This data collection also provided real-time situational awareness information during severe seasonal influenza outbreaks and, after suitable modification, was instrumental in documenting COVID-19-associated preemptive and reactive K-12 school closures in the United States from February 2020 through June 2022 [10]. The ILI-SC data, systematically collected from public sources, have been congruent with trends observed via national disease surveillance systems both during high ILI activity seasons as well as during the unusually low ILI incidence in COVID-affected years. Of note, during the study implementation ILI-SC data were available about a week sooner than ILINet data, i.e. ILI-SC data were being compiled in near-real time.