Early in the A(H1N1)pdm09 pandemic, it was clear from early data from the US, Mexico and the UK that transmission was heavily focused in children [26-29], which was later confirmed from epidemiologic data from other countries as they became affected. Initial media reports from Mexico in early April 2009 caused concerns about the severity of the emerging pandemic. While the severity of the pandemic was difficult to assess early on, a very severe scenario (e.g., mortality similar to what was reported during the 1918 pandemic) was quickly excluded [27,30]. More precise assessment of severity was harder to make and it was appreciated that severity was indeed a complicated concept.

Those epidemiological characteristics of relatively low (yet uncertain) severity and high transmission (especially in children) meant that the A(H1N1)pdm09 pandemic fell within the ambiguous or “grey” zone for scientific advisers and decision makers (Figure  1). While closing schools during the A(H1N1)pdm09 pandemic was expected to have an impact on transmission because of the age-specific immunity profile, it was unclear whether the potential benefits were worth the high economic and social costs. The absence of clear answers to this and other questions, the complexity of decision making and school systems left fertile ground for divergent views and interpretations on the relevance of school closures in the 2009 context. This was partly reflected by the heterogeneity in the policy options taken by the different countries.

Closing schools during seasonal influenza epidemics is a standard policy in two of the 12 countries and administrative region that participated in this review (Japan and Bulgaria). Japan implements a policy of closure of classes, grades and schools (C-CGS) during seasonal influenza epidemics. This is a gradualist policy. For example, Japan will close a class if a certain percentage (usually 10-20%) of students are absent; close a grade if ≥ 2 classes in the grade meet the above criterion; and close a school if ≥ 2 grades of the school meet the same criterion. The exact criteria for closures are usually defined by the local board of education; but a final decision is made by each school. In Japan, there is no nationwide recommendation apart from the notification to schools from the Ministry of Education (Item No. 1125) about the prevention of influenza-like illness (ILI) that dates back to 1982 and indicates that “class closure should be considered when the rate of student absenteeism due to infection reaches approximately 15-20%”. This notification from 1982 shows the long history of the C-CGS policy in the country and references to the policy can be found in reports dating back to the 1957 pandemic. Figure  2 shows the extent to which the C-CGS policy was implemented during seasonal influenza epidemics between 1997 and 2008. In that period, on average, 8,746 classes were closed each year. 3,166 and 594 classes affected by grade and school closures, respectively. The average proportion of classes affected by full school closures was low (5%).

Bulgaria also has a long history of closing schools during seasonal influenza epidemics. Such policies were first recommended in the 1970s. If more than 30% of schoolchildren are absent because of illness, a temporary school closure of individual schools or of all schools in the region may be considered by regional authorities. Although the recommended threshold of 30% absenteeism rate to trigger closure is high, in practice, many schools close each year during the annual seasonal influenza epidemics.

The other countries and administrative regions participating in this review do not routinely close schools during seasonal influenza epidemics. Some of these have relatively different uses of school closures to deal with public health crises. Notably HK has a considerable experience of closing schools during infectious disease outbreaks. For example, authorities in HK closed all schools during the SARS outbreak in 2003 [25] and all primary schools and kindergartens during a seasonal influenza outbreak in 2008 following two influenza-related deaths in children [20]. The NPI is usually accepted locally and indeed expected by the population including by politicians in HK. In other countries, it may be recognized that individual schools may choose to close temporarily for operational reasons (that is, if they have substantial student or staff absenteeism); but school closure seems to be used only rarely to manage infectious disease outbreaks [31]. In other countries, pre-scheduled school closures may coincide with the local seasonal influenza epidemics, which may have an unintentional impact on virus transmission. For example, South African schools were closed for the duration of the 2010 FIFA World Cup (a period of one month), which coincided with the seasonal influenza epidemic that year.

All twelve countries and administrative regions discuss school closure as a mitigation measure during influenza pandemics; however, only three (Japan, Bulgaria and Thailand) indicate that they would certainly close schools during an influenza pandemic (Table  1). The other nine countries and administrative regions left the option open, stating it would depend on circumstances.

During the A(H1N1)pdm09 pandemic, the decision to close schools was a process that involved both national and local policy makers and school administrators. All countries and administrative regions made recommendations on school closure at the national level. There was often a lead agency/Ministry, which consulted with other agencies/Ministries to prepare recommendations.

In addition to recommendations made at the national level, recommendations were sometimes also made at a more regional (sub-national) level. This was, for example, the case in the US, South Africa and Bulgaria, where health is designated responsibility of the states/provinces.

For reactive closures, although recommendations were essentially made at the national level, decision making on school closure was always undertaken at the local level. In the UK and Thailand, closure was decided at the school level (by school principals, headmasters or school boards). In China, France and South Africa, the decision was made by local governments or local representatives of the state.

In some regions, decisions were made at different levels depending on the type of school or the pandemic phase. For example, in the US, different States and cities had different school closure policies, and the decision as to whether to close schools or not were being made locally. In New York City, the NYC Department of Health and Mental Hygiene made the decision for public schools in consultation with the Department of Education; but private schools made their decisions independently. In Japan, New Zealand and Thailand, regional policy makers were in charge of making the decision during the initial phase; but later on, closure was the responsibility of school officials.

For countries and administrative regions that provided information, Figure  3 shows the total number of schools and number of schools per 1 million inhabitants, which were affected by school, grade or class closures. Relative to population size, Serbia and HK, which closed all the schools, were by far the areas which implemented a school closure policy most widely. 160,742 closures were reported in Japan (1,258 reported closures per million inhabitants). No school closure count was available for Bulgaria, the last country that closed schools proactively. However, multiple reporting from the same school was frequent so that the Japanese figure is an over estimate (there are about 60,000 schools in Japan). Even so, it indicates that closures were on a massive scale in Japan.

It is interesting to note that although France, Thailand, China and Italy made relatively similar recommendations of closing schools reactively, the extent to which the policy was implemented locally varied markedly. For example from 0.13 (Italy) to 16 (France) schools affected per 1 million inhabitants though for those two countries there was no strong evidence of differences in impact of the pandemic [34]. For countries and administrative regions that provided information, the effective duration of closure was often shorter than the duration of seven days (or sometimes 14 days) recommended in many locations (Figure  4).

First, there were important differences in the management of school closures across the participating countries. The processes (e.g., who was in charge of making recommendations, who was in charge of making the decision to close, etc) varied between countries. There were also marked differences in the school closure strategies with three countries and the European Union recommending to not use school closure as a mitigation strategy from relatively early on, while some countries recommended some sort of reactive closure, and still others implemented proactive closures at a large scale. Even among countries that made similar recommendations (e.g., allowing reactive closures) the actual extent of closure that took place varied substantially.

In the original plans of this research, we aimed to report on the reasoning behind each national policy decision to try to better understand why outcomes were so diverse. However, we quickly realized that this would be difficult to document in an objective way and that besides the outcome might actually provide limited insight. Indeed, the arguments in favor or against closure are already quite well known: it is a matter of finding the right balance between mitigating and delaying spread versus paying the potentially high cost associated with closure. Therefore, the question is not so much about the arguments used by countries to justify their decisions but more about why they made different appreciations of the health benefits and the economic and social costs of the intervention. In the end, we believe that the single most important factor to explain these differences was the very different public health practises and public expectations concerning school closures and influenza in the countries selected. For example, Japan and Bulgaria consider school closure as routine during seasonal influenza epidemics; and in HK, closures are expected from the population and politicians during large-scale infectious disease outbreaks. By contrast, it may require a severe pandemic for the intervention to be considered as a policy option, for example, in the UK. Obviously, these different interpretations were made possible because of the absence of a clear cut scientific/public health answer on the anticipated impact of school closures in a pandemic of moderate severity such as the A(H1N1)pdm09 pandemic, illustrated by the “grey zone” in Figure  1. Though closing schools was expected to have a significant impact on transmission dynamics in 2009 because of the age-specific transmission profile of the pandemic virus, one must ask if the benefits outweigh the high economic and social costs of closure, especially given that disease severity was relatively limited. This left considerable room for different views on the effectiveness to be transformed into policy. A last factor was the complexity of school systems and hence decision making with co-existing state and religious systems plus private schools in a number of countries.

Implementing reactive closures of schools on a relatively large scale requires that surveillance systems are in place in schools to monitor illness and absenteeism rates. In NYC, for example, the decision to close schools was based on trends in influenza like illness (ILI) visits to school nurses (sustained or sudden increase) and absenteeism. School health nurses who were in charge of gathering the data for their school were often overwhelmed so that the school data was generally not available for review until late afternoon and the decisions on closure could not take place until the evening - too late to inform parents.

Since school closures to mitigate influenza epidemics have been standard policy in Japan for many years, the country has developed an efficient system to monitor absenteeism in schools, and make decisions on closure on the basis of that system. The information about each school passed to relevant education boards (municipal board for most elementary and junior high schools and prefectural board for most high schools), and directly to the Ministry of Education for most private schools. The information is analyzed at the Ministry of Education and shared with the Ministry of Health. This process is performed on a daily basis. In contrast, pro-active closures as implemented in Osaka and Hyogo prefectures in Japan or in HK were simply triggered by the first local pandemic cases that were not linked to importation.

School closure was associated with a range of communication challenges.

One of the challenges faced when managing school closures was that different schools systems (e.g., public, private and parochial) often coexist. For example in NYC the public (state) system is centrally operated and closures in that system were decided by the Chancellor or the like; but that is not the case of the private and parochial systems for which closures were done at the discretion of the school.

In many countries, the national government could impose nationwide closures. But this is not necessarily straightforward. For example, in semi-autonomous or federal countries like South Africa or the USA, each state/province has the authority to make their own decision even if this means a seeming inconsistency across the country.

The countries and administrative regions participating in this review are not a representative sample of countries around the world. Indeed, our intention was not to make a systematic review but to describe the diversity of school closure experiences and provide examples from national and local perspectives. Therefore it is not possible to generalize on the extent of school closure around the world from this work.

In this review, we described national and local experiences on school closure. We presented the various policy processes leading to closure (from the elaboration of recommendations to the implementation of the policy) as well as the extent of closure (e.g., how many schools closed and for how long). We also discussed policy challenges associated with the intervention. However, important areas of research on school closures, such as the impact on spread and health care provision as well as the economical and social cost of closing schools, were left out of the review. A review of those aspects of school closure can be found in [12].