Developing guidelines for school closure interventions to be used during a future influenza pandemic
© Halder et al; licensee BioMed Central Ltd. 2010
Received: 24 March 2010
Accepted: 27 July 2010
Published: 27 July 2010
The A/H1N1 2009 influenza pandemic revealed that operational issues of school closure interventions, such as when school closure should be initiated (activation trigger), how long schools should be closed (duration) and what type of school closure should be adopted, varied greatly between and within countries. Computer simulation can be used to examine school closure intervention strategies in order to inform public health authorities as they refine school closure guidelines in light of experience with the A/H1N1 2009 pandemic.
An individual-based simulation model was used to investigate the effectiveness of school closure interventions for influenza pandemics with R0 of 1.5, 2.0 and 2.5. The effectiveness of individual school closure and simultaneous school closure were analyzed for 2, 4 and 8 weeks closure duration, with a daily diagnosed case based intervention activation trigger scheme. The effectiveness of combining antiviral drug treatment and household prophyaxis with school closure was also investigated.
Illness attack rate was reduced from 33% to 19% (14% reduction in overall attack rate) by 8 weeks school closure activating at 30 daily diagnosed cases in the community for an influenza pandemic with R0 = 1.5; when combined with antivirals a 19% (from 33% to 14%) reduction in attack rate was obtained. For R0 >= 2.0, school closure would be less effective. An 8 weeks school closure strategy gives 9% (from 50% to 41%) and 4% (from 59% to 55%) reduction in attack rate for R0 = 2.0 and 2.5 respectively; however, school closure plus antivirals would give a significant reduction (~15%) in over all attack rate. The results also suggest that an individual school closure strategy would be more effective than simultaneous school closure.
Our results indicate that the particular school closure strategy to be adopted depends both on the disease severity, which will determine the duration of school closure deemed acceptable, and its transmissibility. For epidemics with a low transmissibility (R0 < 2.0) and/or mild severity, individual school closures should begin once a daily community case count is exceeded. For a severe, highly transmissible epidemic (R0 >= 2.0), long duration school closure should begin as soon as possible and be combined with other interventions.
There is a continuing threat of a future novel influenza pandemic having high morbidity (in terms of hospitalization) and mortality (in terms of case fatality) rates. Major pandemics occurred in the past century due to antigenic shift and reassortment of influenza viruses causing millions of deaths; among them 1918's caused the most death . There were also influenza pandemics that were moderate in terms of mortality in 1957 and 1968 . In 2009, influenza A/H1N1 virus, first identified in Mexico, rapidly circulated around the world causing an influenza pandemic . The A/H1N1 2009 influenza pandemic has caused at least 16,455 deaths in 213 countries as of 28th February, 2010 . The 2009 influenza pandemic may cause public health authorities to review their pandemic mitigation guidelines in the light of the limited success in containing and controlling the pandemic. Therefore, improved pandemic guidelines are especially required for future highly pathogenic pandemics, such as may occur if a human transmissible H5N1 virus emerges.
The influenza pandemic guidelines of many countries [5–7] and the World Health Organization (WHO)  suggest a series of non-pharmaceutical and pharmaceutical interventions. Among those intervention strategies, school closure is commonly suggested as a key intervention strategy to slow down the spread of a pandemic within a community, particularly at the early stages of its advancement. The rationale for considering school closures as a frontline intervention is that children and young adults are thought to be the most susceptible to any influenza virus due to their high contact rates within school clusters and limited (or no) immunity to a circulating virus strain when compared to adults. The significance of school closure is also reflected in the A/H1N1 2009 influenza pandemic where about a 60% of cases infected with influenza A/H1N1 virus are 18 years old or younger , and many of the disease transmissions took place in school clusters [9–12]. In addition, school closure can be quickly adopted with a high degree of compliance. School closure strategies are therefore significant for controlling the spread of a pandemic within a community, by breaking the chain of disease transmission among school children and young adults. This frontline intervention is also intended to allow sufficient time for the distribution of antiviral drugs and development of new vaccines.
During the early progression phase of the A/H1N1 2009 influenza pandemic, school closure interventions and application of antiviral drugs for treatment and prophylaxis were implemented in Australia and other parts of the world [13–15]; however, in certain countries the activation of these measures appears to have been somewhat haphazard. Modelling studies [16–22, 10, 23] have shown that influenza virus strains which are more transmissible than A/H1N1 2009 influenza pandemic, would be susceptible to attack rate reduction through school closure strategies, and might be contained if combined with other social distancing and pharmaceutical interventions. Benefits, in terms of reduced attack rates (cumulative illness attack rate and daily incidence rate) and avoided hospitalizations and deaths, would be expected from closing schools during an influenza pandemic; however, school closure must be weighed against the potential high economic and social costs. Studies [9, 24] suggested that a 12 weeks school closure might cost in range of 0.2% - 1.0% of GDP in the UK. Strong hesitance may be experienced in the implementing school closure unless the virus strain is severe, i.e. has a high case fatality ratio. In addition there is no clear agreement on operational issues such as, when school closure should be initiated (its activation trigger), how long schools should be closed (length of closure or duration) and what types of school closures (either individual school closure or simultaneous school closure) should be adopted. In this context, Cauchemez et al. reviewed historical approaches of school closure strategies as a public health policy in . No modelling study has yet extensively and systematically evaluated the operational issues of school closure interventions in terms of how they impact on the effectiveness of alternative school closure strategies; the objective of this study is to do just this. In light of the varied school closure responses [11, 14, 15] to the A/H1N1 2009 influenza pandemic taken by many countries we believe that the outcomes from this detailed evaluation of school closure strategies will help better inform public health policy makers as to the optimal use of school closure measures during a future influenza pandemic.
A detailed simulation model of a real community of Albany (a small city in Western Australia) with ~ 30,000 population was used to simulate local epidemics for R0 values of 1.5, 2.0 and 2.5 with the dynamics of influenza A/H1N1 2009 swine flu virus. Details and an extensive methodological description of the population model used here was described in [21, 23, 25]. Using this model, we conducted stochastic, individual-based simulations of local epidemics, assuming that an average of one new infection per day was randomly introduced into the population for the whole epidemic. The simulation period was divided into 12 hour day/night cycles and during each simulation cycle a nominal location of each individual was calculated, taking into consideration the cycle type (day/night, weekday/weekend), the infection state of each individual and whether child supervision was needed to look after a child at home. Individuals residing in the same location during the same period of time (cycle) were assumed to come into potential infective contact.
School Closure and Antiviral interventions
The school closure interventions modelled were individual school closure and simultaneous school closure strategies. In the individual school closure strategy, we assumed that upon a single diagnosed symptomatic case within a primary school, the whole school was closed; if there was one or two cases diagnosed in a high school only the class members of the affected class were isolated; and finally if there were more than two cases diagnosed in a high school the entire school was closed. However, this school closure policy was only activated when the daily number of diagnosed cases in the community reached an activation trigger. Cases occurring in schools before this time did not result in school closure. We also evaluated a simultaneous school closure strategy where all schools in the community were closed simultaneously, using a similar community activation trigger. School closure interventions were modelled for fixed durations from 2, 4 or 8 weeks. The intervention activation trigger was modelled in a way such that the intervention would come into effect when there was a certain reported number of diagnosed cases per day in the community (for example, after reported cases reached 20 per day the public health authority may announce 2 weeks of school closure).
We also evaluated the application of antiviral drugs to allow us to examine intervention strategies which combine school closure with antiviral treatment and prophylaxis, as occurred during the A/H1N1 2009 influenza pandemic . We assume that 50% of symptomatic individuals were diagnosed and treated with antiviral drugs and their close contact household members were given antivirals for prophylaxis. We also assumed that the antivirals were distributed continually throughout the pandemic period, restricted to 1 course for treatment and maximum 2 courses for prophylaxis per individual. The antiviral administration strategy began once the reported cases in the community reached 10 per day. This would provide sufficient time to the public health authorities to become aware of the arrival of the epidemic in the community and allow time for antiviral distribution and so forth. Detailed modelling explanation and parameters for antiviral drug interventions are given in  with key details being repeated below.
The baseline transmission coefficient βwas initially chosen to give an epidemic with an attack rate of 17.4% which is consistent with seasonal influenza. To achieve simulations under a range of reproductive numbers, βwas increase from this baseline value to achieve epidemics of various R0 magnitudes. Details of the procedure for estimating R0 calibrating βare given in .
The disease infectivity parameter Inf( I i ) was set to 1 for symptomatic individuals at the peak period of infection and then to 0.5 for the rest of the infectivity period. The infectiousness of asymptomatic individuals was also assumed to be 0.5 and this applied to all infected individuals after the latent period but before onset of symptoms. The infection profile of a symptomatic individual was assumed to last for 6 days as follows: a 0.5 day latent period (with Inf( I i ) set to 0) is followed by 1 day asymptomatic and infectious, where Inf( I i ) is set to 0.5; then 2 days at peak infectiousness (with Inf( I i ) set to 1.0); followed by 2.5 days reduced infectiousness (with Inf( I i ) set to 0.5). For an infected but asymptomatic individual the whole infectious period (of 5.5 days) is at the reduced level of infectiousness with Inf( I i ) set to 0.5. This infectivity profile is a simplification of the infectivity distribution found in a study of viral shedding . Following infection an individual is assumed to be immune to re-infection for the duration of the simulation. We further assume that influenza symptoms develop one day into the infectious period , with 20% of infections being asymptomatic among children and 32% being asymptomatic among adults. These percentages were derived by summing the age-specific antibody titres determined in table five of . Symptomatic individuals withdrew into the home with the following probabilities; adults 50% and children 90%, which is in keeping with the work of [16, 17].
The susceptibility parameter Susc( I s ) is a function directly dependent on the age of the susceptible individual. It captures age-varying susceptibility to transmission due to either partial prior immunity or age-related differences in contact behavior. To achieve a realistic age specific infection rate, the age-specific susceptibility parameters were calibrated against the serologic infection rates for seasonal H3N2 in 1977-1978 in Tecumseh, Michigan .
The antiviral efficacy factor AVF( I i , I s ) = (1 - AVE i )*(1 - AVE s ) represents the potential reduction in infectiousness of an infected individual (denoted by AVE i ) induced by antiviral treatment, and the reduction in susceptibility of a susceptible individual (denoted by AVE s ) induced by antiviral prophylaxis. When no antiviral intervention was administrated the values of both AVE i and AVE s were assumed to be 0, indicating no reduction in infectiousness or susceptibility. However, when antiviral treatment was being applied to the infectious individual the value of AVE i was set at 0.66, capturing a reduction in infectiousness by factor of 66% . Similarly, when the susceptible individual was undergoing antiviral prophylaxis the value of AVE s was set to 0.85 indicating a reduction in susceptibility by a factor of 85% .
Simulated characteristics of the epidemics under the un-mitigated scenarios
Three separate influenza epidemics with the basic reproductive number, R0 of 1.5, 2.0 and 2.5 were simulated using the Albany population model. The outcomes of the simulated epidemics varied stochastically due to the random location of infectious individuals seeded into the model as index cases and the probabilistic model of influenza transmission. We further assumed that a continuous influx of infectious individuals was introduced from outside of the simulation boundary at a rate of one infectious case per day to achieve a sustained epidemic for each simulation. Therefore we determined the results of all simulated epidemics from the average of 40 separate simulation runs, each with stochastic choices made using a different random number sequence. The mean cumulative illness attack rates (or attack rate) of 40 simulation runs were 33% (standard deviation 0.57%), 50% and 59% of the total simulated population corresponding to R0 values of 1.5, 2.0 and 2.5 respectively; while peak daily incidences were 120, 333 and 564 cases per 10,000 population. The mean serial intervals or generation times are 2.49 days, 2.36 days and 2.21 days corresponding to R0 values of 1.5, 2.0 and 2.5 respectively.
Impact of duration on school closure interventions
For R0 = 1.5 a maximum of 8% (from 33% to 25%), 10% (from 33% to 23%) and 14% (from 33% to 19%) reduction in attack rate can be achieved at 2, 4 and 8 weeks of school closure respectively (see Figure 1 for R0 = 1.5; blue, orange and dark red lines with circle and square marker).
For R0 = 2.0, a maximum of a 5% (from 50% to 45%), a 7% (from 50% to 43%) and a 9% (from 50% to 41%) reduction in attack rate achieved for 2 weeks, 4 weeks and 8 weeks of school closure respectively (see Figure 1 for R0 = 2.0; blue, orange and dark red lines with circle and square marker).
Similarly for R0 = 2.5, a maximum of a 3% (from 59% to 56%), a 4% (from 59% to 55%) and 5% (from 59% to 54%) reduction in attack rate obtained for 2 weeks, 4 weeks and 8 weeks of school closure respectively (see Figure 1 for R0 = 2.5; blue, orange and dark red lines with circle and square marker).
Combining antiviral treatment and household prophylaxis (T+H) with school closure strategies decreases the attack rates for all duration scenarios. For example, for R0 = 2.0 and 2.5, ~10% reduction in attack rate would be achieved by coupling T+H with the school closure strategies compared to the purely school closure strategies (see Figure 1 for R0 = 2.0 and 2.5; comparing the results with T+H (diamond and triangle marker) and without T+H (circle and square marker) for all blue, orange and dark red lines); whereas for R0 = 1.5, ~8%, ~7% and ~5% reductions in attack rate would be achieved by 2, 4 and 8 weeks school closure strategies respectively (see Figure 1 for R0 = 1.5).
Our simulation results show that for R0 = 1.5, a maximum reduction of 64 (from 120 to 56), 80 (from 120 to 40), 87 (from 120 to 33) cases per 10,000 population would be achieved by 2, 4 and 8 weeks school closure respectively. For R0 = 2.5, a maximum reduction of 184 (from 564 to 380), 194 (from 564 to 370), 214 (from 564 to 350) cases would be achieved by 2, 4 and 8 weeks school closure respectively (see Figure 2; dark blue, orange and dark red lines with circle and square marker).
Coupling antiviral treatment and household prophylaxis (T+H) with school closure strategies would further reduce the peak daily incidence. For example, at R0 = 2.5, a maximum reduction of 304 (from 564 to 260) cases per 10,000 would be achieved by 2 weeks school closure strategies with an activation trigger of 30 daily diagnosed cases (see Figure 2; dark blue lines with diamond and triangle marker).
Impact of activation trigger on school closure interventions
The reductions in attack rate achievable by school closure reported above depend on schools closing at the right time. One of the observable measures of the advance of an influenza epidemic used in public heath forecasting is the number newly infected cases per day, as estimated by the reported number of daily diagnosed cases. In our simulation results, the reported number of daily diagnosed cases has been used as an activation trigger used to initiate the school closure interventions.
Our study indicates that the activation trigger which gives the maximum reduction in attack rate depends on R0, the closure duration, and the type of school closure intervention in a non-trivial way. The impact of the activation trigger on the reduction of attack rate is shown in Figure 1.
If schools close for 2 weeks an activation trigger of at least 50 cases per day gives an 8% (from 33% to 25%) reduction in attack for R0 = 1.5 (see Figure 1 for R0 = 1.5; blue lines with circle and square marker); whereas the optimal trigger is ~80 cases per day for the epidemics with R0 >= 2.0, (see Figure 1 for R0 >= 2.0; blue lines with circle and square marker).
For 4 weeks of school closure, activation triggers of 40, 50 and 10 cases per day should be used to obtain optimal attack rates (see Figure 1; orange lines with circle and square marker) for R0 of 1.5, 2.0 and 2.5 respectively. With 8 weeks of school closure, activation triggers of 30, 5 and 1 case(s) per day give the maximum reduction in attack rate for R0's of 1.5, 2.0 and 2.5 respectively (see Figure 1; dark red lines with circle and square marker).
When adding antiviral treatment and household prophylaxis (T+H) to school closure strategies, simulation suggests that school closure activation triggers of 20, 40 and 50 cases per day are within 1% of optimal for R0 of 1.5, 2.0 and 2.5 respectively (see Figure 1; lines with diamond and triangle marker) and these activation triggers apply for all durations from 2 to 8 weeks. When antivirals are added to school closure, there is less variation in attack rate due to variation in the school closure activation trigger. This is because when antivirals are used, most (at least half, for the scenarios simulated) of the reduction in attack rate is due to antivirals, which have the same effect regardless of the school closure activation trigger. As a result, achieving optimal reduction in attack rates for the combined strategies is less dependant on precise choice of the activation trigger. It turns out that this is true to such a degree that for each R0 a single activation trigger can be chosen for all school closure durations that gives a final attack rate that is within 1% of the best possible outcome. The concurrent use of this antiviral strategy thus effectively eliminates the dependency of optimal activation trigger on school closure duration.
Relationship between activation trigger, cumulative diagnosed cases and intervention activation delay.
Cases diagnosed per day
% of population diagnosed per day
Cumulative diagnosed cases (%)
Activation delay in days
Cumulative diagnosed cases (%)
Activation delay in days
Cumulative diagnosed cases (%)
Activation delay in days
Impact of individual school vs. simultaneous school closure interventions
Optimal attack rate reductions and sensitivity to activation trigger for school closure strategies.
School closure duration
Activation trigger range
Activation trigger range
Activation trigger range
30 - 70
20 - 50
10 - 30
60 - 80
40 - 60
20 - 30
ISC + AV
1 - 30
SSC + AV
30 - 40
40 - 90
10 - 70
1 - 10
70 - 90
30 - 90
10 - 30
ISC + AV
20 - 90
1 - 10
SSC + AV
70 - 90
50 - 90
10 - 50
30 - 90
1 - 20
1 - 50
80 - 90
10 - 90
1 - 50
ISC + AV
30 - 90
1 - 90
1 - 10
SSC + AV
60 - 90
20 - 90
1 - 30
School closure and age-specific attack rates
Epidemiological data for pandemic influenza are often uncertain, sparse, limited and widely distributed. There are uncertainties in characteristics such as mortality, morbidity, disease transmissibility, contact behavior and behavioral changes among people during a pandemic and also in antiviral and vaccine efficacies, pre-existing immunities and well as other uncertainties. The simulation model upon which the current study is based has been subject to an analysis determining its sensitivity to alternative estimates for parameters relating to serial interval, age-specific attack rates and other assumptions (see , electronic supplementary material, Text S2). This analysis shows sensitivity of the un-mitigated attack rate and of the attack rate when mitigated by school closure. The current simulation model differs from the earlier model only in that it has a more refined model of viral shedding (see Methods, Transmission model).
The most notable outcome is that at least a 40% diagnosis ratio is required for the maximum reduction in attack rate (see Figure 4; green and purple line). A lower diagnosis ratio would result in an excessive delay of school closure activation. In contrast, underestimating the diagnosis ratio (and consequently introducing school closure too soon) has less serious consequences. The relationship between activation trigger and activation delay, assuming a diagnosis ratio of 50%, is given in Table 1.
Optimal school closure activation triggers
Optimal triggers in the number of daily diagnosed cases per 30,000 population (% of population newly infected per day)
School closure strategy
For example, for an epidemic with an R0 of 1.5, with a 4 weeks school closure strategy (either individual school closure or simultaneous school closure) triggered by a single diagnosed case, the simulation results indicate that school closure slows the initial epidemic growth, but when school closure interventions are relaxed after 4 weeks the epidemic restarts (see Figure 5 for R0 = 1.5; blue and light green lines). Therefore, determining the optimal school closure trigger is crucial when the number of times schools close and their duration is limited. The optimal school closure activation triggers for a range of reproduction numbers, school closure types and duration scenarios are given in Table 3.
A greater number of activation triggers (for the reproduction numbers and durations considered) have the individual school closure strategy performing better (in terms of reduction in attack rate) than that involving simultaneous school closure. For the trigger scenarios where simultaneous closure is the better strategy, difficulty arises if the optimal case load trigger is missed, possibly due to under- or mis-diagnosis of cases. If this is the situation then individual school closure performs better, and hence is a safer strategy to adopt.
Given that at the beginning of an influenza pandemic we are unlikely to know how many asymptomatic or un-diagnosed cases exist, it may be difficult to determine when the optimal activation trigger has been reached. Since individual school closure does almost as well over a greater number of activation triggers, this suggests that it is a more reliable strategy to adopt in the face of uncertainty about the true degree of epidemic spread. Furthermore, individual school closure is more likely to be socially acceptable for mild pandemic strains, as parents will be aware of the local cases triggering the school closures that affect them.
The marginal increase in effectiveness of additional weeks of school closure decreases as the duration increases. For example, at R0 = 1.5, 2 weeks school closure gives an 8% reduction (4% per week) in attack rate. An additional 2 weeks school closure gives an additional 2% (1% per week) reduction. A further 4 weeks school closure (8 weeks total) gives an additional 4% (1% per week) reduction (see Figure 1 for R0 = 1.5; blue, orange and dark red lines with circle and square marker).
School closure and antiviral drug use for treatment and prophylaxis are complementary. The combination of antiviral strategies together with school closure always gives an increased reduction in attack rate compared to either in isolation, across different transmissibility characteristics, school closure activation triggers (except for significantly delayed school closure at R0 = 1.5), school closure types and school closure durations (see Figure 1 for R0 = 1.5).
Modelling features and relationship to other studies
We have developed an individual-based simulation model of Albany (a small city of ~30,000 populations in Western Australia) utilizing Australian Bureau of Statistics Census data  to simulate virtual epidemics with a range of characteristics and to systematically evaluate school closure strategies. Related models ranged in scale from the whole world , through large [17, 18] and smaller [16, 33] countries, to actual  and synthetic [19, 10] small communities. Our model differs from the differential equations-based deterministic and compartmental models [34–37] which consider homogeneous mixing patterns within and between subpopulations; for a survey on mathematical modelling of influenza pandemics see . Our model encompasses significant complexity including spatial contact structure; age specific susceptibility; mixing groups and community wide random contacts [21, 23]. For the purposes of this study the model has been augmented with sophisticated models of school closure and antiviral prophylaxis and treatment.
Empirical evidence and previous modelling studies indicate that school closure can reduce the magnitude of an influenza epidemic; for a review see . The effectiveness of school closure clearly depends on operational issues such as what type of school closure is adopted (simultaneous school closure or individual school closure), when school closure should be triggered, and how long schools should be closed.
Previous modelling studies have made particular assumptions, plausible or idealized, about these issues, and found school closure to be effective to a lesser or greater extent. In Ferguson et al.  3 weeks of individual school closure strategy was chosen for multiple times and the strategy was activated after a single community case was detected. In Germann et al. , Glass et al.  and Davey et al. [20, 39], simultaneous school closure strategy was assumed for whole pandemic period with an activation trigger following single or several community case(s) detection (in Germann et al. triggered at 1 community case, in Glass et al. at 10 community case and in Davey et al. a sensitivity for 10, 30 and 100 community cases was shown). In Milne et al.  a comparison of school closure interventions evaluated using different individual-based models was studied. In that study the authors emphasized the benefit of long duration school closure in containing an influenza pandemic. However long term school closure has an adverse economic impact; Sadique et al.  estimated a cost in the range 0.2% - 1.0% of GDP in the UK following a 12 weeks of school closure. The effectiveness of school closure evaluated in these modelling studies depended largely on the underlying assumptions made about their models.
In this study we have explicitly considered three key school closure operational issues which were identified in  as being significant. These are when school closure should be initiated (activation trigger), how long schools should be closed (duration) and what type of school closure should be adopted and determined their impact on influenza pandemics: no previous modelling study has systematically evaluated such school closure operational issues. We believe that the detailed evaluation of such strategies will contribute to further development of school closure policy guidelines for future influenza pandemics.
Antiviral drug strategies also play an important role in controlling disease spread at the early stages of an outbreak. In addition to school closure (in isolation), we simulated the application of antiviral drug treatment and prophylaxis to household members of an infected case layered with school closure. A detailed evaluation of the effectiveness of antiviral drug strategies applied to a pandemic similar to the A/H1N1 2009 influenza pandemic has been investigated in .
There are some limitations of the model used in this study. As the model is based on a population in a developed country the outcomes may not be applicable to populations in a developing country, where populations may be less mobile and have higher population densities. We have focused on the reduction in the number of daily symptomatic cases and the cumulative illness attack rate as they are used for determining intervention effectiveness rather than focusing on influenza-related adverse events such as hospitalizations and deaths. We also do not take account of possible antiviral drug resistance [40, 41] that may arise due to the implementation of antiviral drug strategies, as our main goal is to suggest refinements to policy guidelines for school closure.
Public health policy implications
School closure policy recommendations
School closure duration
low (R0 = 1.5)
medium (R0 = 2.0)
high (R0 = 2.5)
Schools should be closed individually when cases are identified in each school; this policy should delayed until the first day on which 13 new cases per 10,000 population are diagnosed (assuming 50% diagnosis of symptomatic individuals).
Schools should be closed individually when cases are identified in each school; this policy should be instituted as soon as possible once the pandemic has reached the community. Antivirals should be dispensed to slow the spread.
Schools should be closed individually when cases are identified in each school; this policy should delayed until the first day on which 6 new cases per 10,000 populationare diagnosed (assuming 50% diagnosis rate). However, if the activation trigger is missed all schools should close as soon as possible. Antivirals should be dispensed together with school closure and home isolation.
All schools should close simultaneously as soon as possible once the pandemic reaches the community. Antivirals and other non-pharmaceutical interventions should also be applied.
All schools should close simultaneously as soon as possible once the pandemic reaches the community Antivirals should be dispensed in larger extent to slow down disease spread. Other social distancing based interventions should be rigorously applied.
We assume that a key factor which will determine the duration of school closure will be the perceived severity of the pandemic (i.e. morbidity and mortality given as case hospitalization and case fatality ratios). For a 'mild' pandemic similar to the A/H1N1 2009 pandemic and similar in severity to seasonal influenza , it is unlikely that school closure would be tolerated for periods longer than two weeks. In contrast, during a pandemic with severity similar to the 1918 pandemic (case fatality ratio estimated as > 2.5%) , the public might accept (or even demand) school closure for several months.
Recommendations for school closure interventions should reflect that in the case of a severe pandemic with high transmissibility (R0 >= 2.0), where schools would be closed for at least 8 weeks, all schools should be closed at the same time, as early as possible once the epidemic reaches the local community. In the case of a moderately severe pandemic (case fatality ratio < 1.0%) where schools would be closed for at least 4 weeks, schools should close individually once cases are detected in each school and this policy should be instituted as soon as possible. In the case of a mild pandemic similar to the A/H1N1 2009, where at most 2 weeks of school closure would be adopted, schools should close individually, but the policy should be delayed until the first day on which 0.13% of the population becomes newly infected (on that day), assuming 50% of symptomatic cases would be diagnosed (see Table 4 for a set of possible school closure policy recommendations).
The additional use of antiviral treatment and prophylaxis will give a greater reduction in attack rate than school closure in isolation for influenza epidemics with R0 >= 2.0, (see Figure 1 for R0 >= 2.0).
For epidemics with R0 <= 1.5 where antivirals are used, there is no advantage having periods longer than 2 weeks (eg. 4 or 8 weeks) of school closure if school closure is delayed for 40 days or more (see Figure 1 for R0 = 1.5). In all other scenarios, the combination of antivirals and school closure is substantially more effective than either strategy alone. For highly transmissible epidemics (R0 = 2.5), fixed period school closure alone will be ineffective (i.e. have a limited ~4% reduction in the attack rate) (see Figure 1 for R0 = 2.5); therefore, additional rigorous social distancing based interventions would need to be applied [21, 23]. If the virus strain is matched to stockpiled vaccines then vaccination strategies suggested in  should be applied at an early stage.
Selecting a school closure activation trigger for short closure periods using a reported number of diagnosed cases per day (rather than a fixed period of time, say 2 weeks) means that the timing adapts to faster developing epidemics (those with higher reproduction numbers). For example, the 40 daily diagnosed cases per 30,000 populations occur at 33, 20 and 15 days for R0 = 1.5, 2.0 and 2.5 respectively (see Table 1). These results show that the optimal use of school closure depends both on pandemic severity (which will determine the feasible duration of school closure) and its transmissibility characteristics.
Our approach to modelling a generic influenza pandemic was to base epidemic characteristic on those of seasonal influenza strains (for which data is available), but to increase the overall transmissibility (by increasing the basic transmission probability) in order to represent higher infectivity of immunologically novel strains. We thus assumed that the age specificity of infection followed a pattern typical of seasonal influenza. With the advent of the 2009/2010 pandemic, data from an actual pandemic influenza strain has become available, albeit a strain that has turned out to be less pathogenic and immunologically novel than first thought. Our R0 = 1.5 simulations align with estimates for the 2009 pandemic [44, 42, 45–47], as does our derived serial interval of 2.49 days. The A/H1N1 2009 pandemic exhibited a somewhat different pattern of age-specific infectivity than our seasonal influenza baseline, with higher attack rates in children and young adults but lower attack rates in older adults. We conducted an alternate set of simulations based on the A/H1N1 2009 age-specific attack rate pattern and R0 (results not reported here). We found that the alternative assumptions lead to a lower overall attack rate (compared to a "seasonal" epidemic with the same R0), but that the proportional effectiveness of school closure and optimal activation triggers were essentially identical.
Our simulation results give guidance as to public health policy decisions in the refinement of school closure strategies to be used in a future influenza pandemic. We have systematically evaluated school closure operational issues to determine when schools should be closed and re-opened to achieve the maximum reduction in influenza spread. We found that the optimal timing of school closure depends both on the duration of school closure (which we assume will depend on the severity of the influenza strain, with strains that are more severe in terms of serious infection outcomes making longer periods of school closure acceptable) and on the transmissibility of the influenza strain (which influences the rate of growth and spread of the epidemic). Accurate early estimates of epidemic characteristics such as the basic reproduction number and disease severity are thus necessary to achieve the maximum case reduction from school closure. We found that a policy of allowing schools to close individually was much less sensitive to the precise timing of the intervention than a policy of simultaneous community-wide school closure, a valuable observation given the difficulty in determining the true degree of epidemic spread in the early stages of an outbreak.
We would like to thank the reviewers for their helpful comments. No external funding supported this study.
- Taubenberger JK, Morens DM: 1918 Influenza: the Mother of All Pandemics. Emerg Infect Dis. 2006, 12: 15-22.PubMed CentralView ArticlePubMed
- Glezen WP: Emerging infections: pandemic influenza. Epidemiol Rev. 1996, 18: 64-76.View ArticlePubMed
- WHO: Influenza-like illness in the United States and Mexico. [http://www.who.int/csr/don/2009_04_24/en/index.html]
- WHO: Pandemic (H1N1) 2009 - update 87. [http://www.who.int/csr/don/2010_02_12/en/index.html]
- Pandemic Influenza Plan. [http://www.hhs.gov/pandemicflu/plan/]
- Pandemic flu: A national framework for responding to an influenza pandemic. [http://www.dh.gov.uk/en/Publicationsandstatistics/Publications/PublicationsPolicyAndGuidance/DH_080734]
- Australian Health Management Plan for Pandemic Influenza. [http://www.flupandemic.gov.au/internet/panflu/publishing.nsf/Content/ahmppi-1]
- World Health Organization: Pandemic influenza preparedness and response. [http://www.who.int/csr/disease/influenza/pipguidance2009/en/index.html]
- Cauchemez S, Ferguson NM, Wachtel C, Tegnell A, Saour G, Duncan B, Nicoll A: Closure of schools during an influenza pandemic. The Lancet Infectious Diseases. 2009, 9: 473-481. 10.1016/S1473-3099(09)70176-8.View ArticlePubMed
- Sypsa V, Hatzakis A: School closure is currently the main strategy to mitigate influenza A(H1N1)v: a modelling study. Euro Surveill. 2009, 14: ppi = 19240
- Nishiura H, Castillo-Chavez C, Safan M, Chowell G: Transmission potential of the new influenza A (H1N1) virus and its age-specificity in Japan. Euro Surveill. 2009, 14: pii = 19227
- Effler PV, Carcione D, Giele C, Dowse GK, Goggin L, Mak DB: Household responses to pandemic (H1N1) 2009 - related school closures, Perth, Western Australia. Emerg Infect Dis. 2010, 16: 205-PubMed CentralView ArticlePubMed
- Nicoll A, Coulombier D: Europe's initial experience with pandemic (H1N1) 2009 - mitigation and delaying policies and practices. Euro Surveill. 2009, 14: pii = 19279
- BBC News: UK Five new UK flu cases confirmed. [http://news.bbc.co.uk/2/hi/uk_news/8032697.stm]
- Australian Commonwealth Department of Health and Aging: Pandemic (H1N1) 2009 update bulletins for 7 to 13 September 2009. [http://www.healthemergency.gov.au/internet/healthemergency/publishing.nsf/Content/bulletins-7-13Sep]
- Longini IM, Nizam A, Xu S, Ungchusak K, Hanshaoworakul W, Cummings DAT, Halloran ME: Containing pandemic influenza at the source. Science. 2005, 309: 1083-1088. 10.1126/science.1115717.View ArticlePubMed
- Ferguson NM, Cummings DAT, Fraser C, Cajka JC, Cooley PC, Burke DS: Strategies for mitigating an influenza pandemic. Nature. 2006, 442: 448-452. 10.1038/nature04795.View ArticlePubMed
- Germann TC, Kadau K, Longini IM, Macken CA: Mitigation strategies for pandemic influenza in the United States. PNAS. 2006, 103: 5935-5941. 10.1073/pnas.0601266103.PubMed CentralView ArticlePubMed
- Glass RJ, Glass LM, Beyeler WE, Min HJ: Targeted social distancing design for pandemic influenza. Emerg Infect Dis. 2006, 12: 1671-1681.PubMed CentralView ArticlePubMed
- Davey VJ, Glass RJ, Min HJ, Beyeler WE, Glass LM: Effective, robust design of community mitigation for pandemic influenza: a systematic examination of proposed US guidance. PLoS ONE. 2008, 3: e2606-10.1371/journal.pone.0002606.PubMed CentralView ArticlePubMed
- Milne GJ, Kelso JK, Kelly HA, Huband ST, McVernon J: A small community model for the transmission of infectious diseases: comparison of school closure as an intervention in individual-based models of an influenza pandemic. PLoS ONE. 2008, 3: e4005-10.1371/journal.pone.0004005.PubMed CentralView ArticlePubMed
- Halloran ME, Ferguson NM, Eubank S, Longini IM, Cummings DAT, Lewis B, Xu S, Fraser C, Vullikanti A, Germann TC: Modeling targeted layered containment of an influenza pandemic in the United States. PNAS. 2008, 105: 4639-10.1073/pnas.0706849105.PubMed CentralView ArticlePubMed
- Kelso JK, Milne GJ, Kelly HA: Simulation suggests that rapid activation of social distancing can arrest epidemic development due to a novel strain of influenza. BMC Public Health. 2009, 9: 117-10.1186/1471-2458-9-117.PubMed CentralView ArticlePubMed
- Sadique M, Adams E, Edmunds W: Estimating the costs of school closure for mitigating an influenza pandemic. BMC Public Health. 2008, 8: 135-10.1186/1471-2458-8-135.PubMed CentralView ArticlePubMed
- Milne G, Kelso J, Kelly H: Strategies for mitigating an influenza pandemic with pre-pandemic H5N1 vaccines. J R Soc Interface. 2010, 7: 573-586. 10.1098/rsif.2009.0312.PubMed CentralView ArticlePubMed
- Halder N, Kelso J, Milne G: Analysis of the effectiveness of interventions used during the 2009 H1N1 influenza pandemic. BMC Public Health. 2010, 10: 168-10.1186/1471-2458-10-168.PubMed CentralView ArticlePubMed
- Carrat F, Vergu E, Ferguson NM, Lemaitre M, Cauchemez S, Leach S, Valleron A-J: Time lines of infection and disease in human influenza: a review of volunteer challenge studies. Am J Epidemiol. 2008, 167: 775-785. 10.1093/aje/kwm375.View ArticlePubMed
- Fox JP, Hall CE, Cooney MK, Foy HM: Influenzavirus infections in Seattle families, 1975-1979. Am J Epidemiol. 1982, 116: 212-227.PubMed
- Monto A, Koopman J, Longini IMJ: Tecumseh study of illnes. XIII. Influenza infection and disease, 1976-1981. Am J Epidemiol. 1985, 121: 811-822.PubMed
- Yang Y, Longini IM, Halloran ME: Design and evaluation of prophylactic interventions using infectious disease incidence data from close contact groups. Appl Statist. 2006, 55: 317-330.
- Australian Burea of Statistics: Census Data Online. [http://abs.gov.au/websitedbs/D3310114.nsf/home/Census+data]
- Colizza V, Barrat A, Barthelemy M, Valleron A-J, Vespignani A: Modeling the worldwide spread of pandemic influenza: baseline case and containment interventions. PLoS Medicine. 2007, 4: e16-10.1371/journal.pmed.0040013.View Article
- Ciofi degli Atti ML, Merler S, Rizzo C, Ajelli M, Massari M, Manfredi P, Furlanello C, Tomba GS, Iannelli M: Mitigation measures for pandemic influenza in Italy: an individual based model considering different scenarios. PLoS ONE. 2008, 3: 1-11. 10.1371/journal.pone.0001790.View Article
- Rvachev LA, Longini IM: A mathematical model for the global spread of influenza. Math Biosci. 1985, 3-22. 10.1016/0025-5564(85)90064-1.
- Arinaminphathy N, McLean AR: Antiviral treatment for the control of pandemic influenza: some logistical constraints. J R Soc Interface. 2007, 5: 545-553. 10.1098/rsif.2007.1152.View Article
- McCaw JM, McVernon J: Prophylaxis or treatment? Optimal use of an antiviral stockpile during an influenza pandemic. Math Biosci. 2007, 209: 336-360. 10.1016/j.mbs.2007.02.003.View ArticlePubMed
- Chin J, Koh G, Lee D-Y: How necessary is a fast testkit for mitigation of pandemic flu?. J R Soc Interface. 2009.
- Coburn B, Wagner B, Blower S: Modeling influenza epidemics and pandemics: insights into the future of swine flu (H1N1). BMC Medicine. 2009, 7: 30-10.1186/1741-7015-7-30.PubMed CentralView ArticlePubMed
- Davey VJ, Glass RJ: Rescinding Community Mitigation Strategies in an Influenza Pandemic. Emerg Infect Dis. 2008, 14: 365-371. 10.3201/eid1403.070673.PubMed CentralView ArticlePubMed
- Arino J, Bowman CS, Moghadas SM: Antiviral resistance during pandemic influenza: implications for stockpiling and drug use. BMC Infect Dis. 2009, 9: 8-10.1186/1471-2334-9-8.PubMed CentralView ArticlePubMed
- Eichner M, Schwehm M, Duerr H, Witschi M, Koch D, Brockmann S, Vidondo B: Antiviral prophylaxis during pandemic influenza may increase drug resistance. BMC Infect Dis. 2009, 9: 4-10.1186/1471-2334-9-4.PubMed CentralView ArticlePubMed
- Nishiura H, Chowell G, Safan M, Castillo-Chavez C: Pros and cons of estimating the reproduction number from early epidemic growth rate of influenza A(H1N1) 2009. Theoretical Biology and Medical Modelling. 2010, 7: 10.1186/1742-4682-7-1.
- US Center for Disease Control: Questions & Answers 2009 H1N1 Flu (Swine Flu) and You. [http://www.cdc.gov/h1n1flu/qa.htm]
- Nishiura H, Wilson N, Baker M: Estimating the reproduction number of the novel influenza A virus (H1N1) in a Southern Hemisphere setting: preliminary estimate in New Zealand. Journal of the New Zealand Medical Association. 2009, 122: 73-77.
- Fraser C, Christl DA, Cauchemez S, Hanage WP, Van Kerkhove MD, Hollingsworth TD, Griffin J, Baggaley RF, Jenkins HE, Lyons EJ, et al: Pandemic potential of a strain of influenza A (H1N1): early findings. Science. 2009, 324: 1557-1561. 10.1126/science.1176062.PubMed CentralView ArticlePubMed
- McBryde E, Bergeri I, van Gemert C, Rotty J, Headley E, Simpson K, Lester R, Hellard M, Fielding J: Early transmission characteristics of influenza A (H1N1) v in Australia: Victorian state, 16 May-3 June 2009. Euro Surveill. 2009, 14: pii = 19363
- Munayco C, Gomez J, Laguna-Torres V, Arrasco J, Kochel T, Fiestas V, Garcia J, Perez J, Torres I, Condori F: Epidemiological and transmissibility analysis of influenza A (H1N1) v in a southern hemisphere setting: Peru. Euro Surveill. 2009, 14: pii = 19299
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2334/10/221/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.