Centers for Disease Control and Prevention



Supplementary informationforSchool dismissal as a pandemic influenza response:When, where and for how long?Timothy C. Germann, Hongjiang Gao, Manoj Gambhir, Andrew Plummer,Matthew Biggerstaff, Carrie Reed, and Amra Uzicanin1. Simulation Model and ParameterizationThe basic EpiCast model integrates four key elements: (a) Community-level transmission between people, through various contact groups (including households, workplaces, schools, and other settings); (b) Disease natural history model and parameters; (c) U.S. Census demographics (where people live) and workerflow data (where they work), data which was last available at tract-level resolution in the 2000 Census (281 million individuals); and (d) Bureau of Transportation Statistics long-distance travel data.EpiCast is constructed as a collection of interacting individual-based single-community models (such as those used in earlier modeling studies [e.g., Elv76, Hal02a, Hal02b, Lon04]), using U.S. Census demographics and worker-flow data to populate the communities with representative daytime and evening contact networks, and commuter flow between residential and workplace locations. Starting from a “baseline” model of pandemic spread in the absence of any mitigation measures, various pharmaceutical and non-pharmaceutical interventions can be modeled by how they affect person-to-person contact rates and/or the susceptibility, infectiousness, or disease course within individuals. Pharmaceutical measures include vaccines or antivirals, while school dismissal, liberal leave workplace policies, and other social distancing measures are among the most commonly proposed NPIs.In the remainder of this section, we will describe each of these aspects in greater detail. EpiCast model parameters can be divided into four classes which directly affect the spatio-temporal dynamics of the epidemiological simulation model: (i) person-to-person transmission; (ii) disease natural history; (iii) mitigation options, both pharmaceutical (vaccines and antivirals) and non-pharmaceutical (social distancing); and (iv) initial conditions, the when and where of index case(s). Parameter classes (i)-(iii) are discussed in this section, and the procedure by which they were varied to match the Pandemic Severity Assessment Framework (PSAF) and school dismissal options under consideration in subsection H. The initial conditions (iv) are discussed for the regional and US models in the main text. A. Community-level transmission modelThe basic building block for the EpiCast national simulation model is a discrete-time, stochastic simulation model of disease spread within a structured 2,000-person community. Such models have been developed and applied previously to both influenza [Elv76, Hal02a, Lon04] and smallpox [Hal02b]. The model population is stochastically generated to match census-based nationwide distributions of age, household size, and employment status. Each person in the population belongs to one of five age groups: preschool-age children (0-4 years), school-age children (5-18 years), young adults (19-29 years), adults (30-64 years), and older adults (64+ years). Households consist of one to seven persons, with either one or two adults, and are grouped randomly into clusters of four households each, and further grouped into one of four non-overlapping neighborhoods, each containing approximately 500 people. Every person also belongs to a set of close and casual contact (also referred to as “mixing”) groups, ranging from their household and household cluster (highest contact rates), to schools and workplaces, down to their neighborhood, and the entire community (with the lowest contact rates, representing occasional interactions in malls, supermarkets, and churches, for instance).All preschool-age children are assigned to either a neighborhood daycare center, with 14 children on average, or to one of several smaller neighborhood playgroups, each with four children. Depending on their age, school-age children may belong to one of two elementary school groups (each shared between two neighborhoods, with 79 students each on average), to a community-wide middle school group (128 students on average), or to a community-wide high school group (average 155 students). These school contact groups are, in general, not actual schools, but rather representative of the typical daily interactions a student may have with classmates and other peers. According to U.S. census data, 93% of children 5–18 years old attend school, so we allow the remaining 7% to mix in the household, household cluster, neighborhood, and community during the daytime. Working adults (restricted to those who are 19–64 years old in our model) belong to a work group of approximately 20 people. Although in reality many workplaces are larger than 20 people, we assume that workers make a contact of sufficient duration and/or closeness to transmit influenza virus with a subset of the entire workforce at that location.Transmission within each contact group i is described by a contact probability ci, which may depend on the age of both the infectious and susceptible persons (Table 1). This contact probability represents the likelihood (within each 12-hour time step of the simulation model) of having a contact of sufficient duration and closeness for transmission of an infectious dose of influenza virus to be possible between these two individuals in this social setting. These relative contact probabilities are all multiplied by an overall probability of transmission given contact (ptrans) to determine absolute transmission probabilities. Thus, ptrans is the key input parameter to control the overall transmissibility and, is thus, adjusted to obtain the desired the basic reproductive number R0, which is measured as a simulation output.At a finer scale, the individual contact probabilities ci (Table 1 of the supporting information for [Ger06]) were adjusted manually to give the desired age-stratified attack rates and infection sources (household, workplace, school, etc.), which are consistent with prior studies (e.g. Tecumseh [Mon85]), and the detailed EpiSimS simulation model [Eub04]. As described in the supporting information for [Ger06], the initial model was calibrated to give an age-dependent attack rate pattern between the historical 1957-8 “Asian” influenza A (H2N2) and 1968-9 “Hong Kong” influenza A (H3N2) pandemic strains, and transmission occurring primarily (~40%) in the household, followed by schools (~20%), and then evenly split (~10% each) between the remaining four contact groups (workgroups, household cluster, neighborhood, and community).B. Disease natural historyWe employ the same influenza natural history model as used previously [Lon05, Ger06], and based upon historical analysis of the 1957, 1968 pandemics and post-1968 influenza A [Kil75]. The main points are that the latent, incubation, and contagious period durations are each sampled from discrete distributions, with mean periods of 1.2, 1.9, and 4.1 days, respectively, leading to a mean serial interval (generation time) of 3.95 days, the sum of the mean incubation period and half of the average contagious period (Table S1). The contagious period includes both the slight difference between latent and incubation periods, as well as the standard post-incubation period, when symptoms appear in the 67% of infected people we assume will present with clinical symptoms. Any infectiousness that is not accompanied by overt symptoms (namely, the post-latent part of the incubation period, if any, and the 33% of infected people who remain asymptomatic, or at least subclinical) is assumed to be half as great as the infectiousness of symptomatic individuals, reducing the transmission probability by a factor of two. The 2:1 symptomatic/asymptomatic branching ratio has been used in several previous studies [Elv76, Hal02a, Lon04, Lon05, Ger06, Lee10, Mil13], although other modeling work has used a 1:1 branching ratio [Fer05, Bro11]. The 2:1 ratio received further support from a 2008 review of 56 volunteer challenge studies, which found a symptomatic infection frequency of 66.9% [Car08]. However, two independent community studies, the 1976-81 Tecumseh, Michigan [Mon85] and 2006-11 Flu Watch cohort study in England [Hay14], both found an inverse 1:3 ratio, with roughly 75% of infected people remaining asymptomatic. (Note that the latter study included the 2009 H1N1 pandemic season, finding no significant difference between the pandemic and seasonal influenza years.) Clearly, this symptomatic/asymptomatic branching ratio, as well as the relative infectiousness of asymptomatic individuals, remain key unknowns which will need to be estimated as rapidly as possible during an emerging influenza pandemic. To test the robustness of our conclusions to this assumed disease progression, we also utilize an alternative disease natural history with a shorter (~2.8 day) serial interval (Table S1), which has been used in other modeling work [Fer05, Fer06]. It has a comparable incubation time (~2 days, slightly longer than the 1.9 days for the previous model), based in part on the classic study on a delayed Alaska Airlines flight [Mos79], but an infectious period that is assumed to have a lognormal time dependence, with parameters fit via MCMC [Fer05] to a French household study of seasonal flu [Car02]. The average infectious period of 1.6 days leads to a serial interval more than a day shorter than the first model, only ~2.8 days as compared to ~4 days for the default model. In a household study [Cow09] during the 2007 interpandemic influenza season in Hong Kong, 122 household index cases were self-reported (and laboratory-confirmed), nose/throat swabs collected from 350 household contacts who were followed, 21 of whom had laboratory-confirmed influenza, and 14 of which were symptomatic (fortuitously exactly the 67% branching ratio). The clinical-onset serial interval of these 14 symptomatic contacts was fit to a Weibull model, with a mean of 3.6 days, and a 95% CI (2.9–4.3) almost exactly spanning the two disease natural histories considered here.C. Construction of the synthetic population from Census dataThe fundamental geographic unit in our model is the census tract, which is defined as a relatively stable geographic area with between 1,500 and 8,000 residents. In the 2000 Census, there were 65,443 census tracts containing 281,421,906 people in the United States (50 states and District of Columbia), corresponding to an average population of 4,300 residents per tract. We round off the population of each tract to the nearest 2,000 persons, and populate each tract with the appropriate number of 2,000-person communities, each with households, schools, and other mixing groups as described above. In addition, several urban tracts have little or no residential population, but a large daytime worker population. We model these by communities comprised solely of daytime work groups (in addition to the broad but weaker community-level mixing), with an average of five 20-person work groups per each such community (corresponding to the average number of work groups in the suburban community model). In this way, we are able to realistically differentiate primarily residential tracts (with few, if any, work groups) from primarily urban ones (including some with few or even zero households). Each of the 180,492 model communities making up the national model is stochastically generated in an independent manner, so that no two communities within the nation are exactly identical.Workplace tracts are chosen using the tract-to-tract worker flow data from the 2000 Census (the last decennial census to collect this detailed information), which also provides the total number of working (and conversely, of unemployed) adults in each tract. We note that this raw data refers to where individuals were working during the Census 2000 reference week (generally the last week of March 2000), which is why a significant number of people (1.13 million, or 0.9% of the total workforce) were reported as working at locations 100 miles or more from their residence. We assume that such travel does not occur on a daily basis and, instead, place these individuals in a workgroup in their home tract. A related issue is the workers who were sick, on vacation, or otherwise absent from work during the reference week, estimated at 2% of employed persons. Since both vacations and sick leave (withdrawal from workgroups) are included in the model, we compensate for these uncounted workers by multiplying each tract-to-tract worker flow total by 1.02.In the original EpiCast model [Ger06, Hal08], the nationally averaged household size and age distributions were used in constructing each ~2,000-person community. In order to improve the fidelity of the model population for the present study, specifically with regard to the diversity in school sizes across the United States, we utilize Census tract-level demographic information, specifically age and household size distributions, when creating the ~2,000-person model communities that comprise each Census tract. This has the effect of increasing school size heterogeneity. Whereas the synthetic population used in previous EpiCast simulation studies [Ger06, Hal08] had a narrow distribution of school sizes due to the stochastic population generation method (with national-average demographics used in each community), the use of local demographics broadens this distribution. If desired, a long tail could be added by merging the two elementary schools in each community, as shown in Fig. S1(a). Previous modeling studies [Str07] have demonstrated the impact that local demographic characteristics have on the resulting spread, in particular that the clinical attack rate is strongly correlated with both the average household size and proportion of school-age children within a community.As can be seen in Fig. S1(b), using either the national average of local Census tract demographics, the ordering of school mixing group sizes does not simply increase or decrease with increasing grade levels, but rather is: elementary school < high school < middle school. At first this appears somewhat curious, but in fact the average number of students per grade level follows this ordering. Moreover, the actual magnitudes (75, 188, and 172 mean students per grade level in elementary, middle, and high school, respectively) are comparable to the peaks of the distributions in Fig. S1(b), suggesting that one might roughly think of a school mixing group as a grade level. However, this is only a crude comparison; it is unknown to what extent the actual heterogeneity in school and classroom sizes and contact patterns affects model conclusions. Although to our knowledge no models of populations as large as the United States have explicitly included such data, it is worth noting that several detailed models with specific schools and workplaces have been constructed, including Portland, Oregon [Eub04, Bar05], Albany, Western Australia [Mil08], and the state of Pennsylvania [Bro11].D. Long-distance travel modelIn addition to the regular (daily) commuting patterns, it is also important to capture the infrequent and irregular long-distance travel, such as business trips or vacations. We base this component on the 1995 American Travel Survey data available from the U.S. Department of Transportation, Bureau of Transportation Statistics. The ~1 billion person-trips (defined as 100 miles or longer each way, within the U.S.) among the 263 million residents at that date leads to an average of 3.8 trips per person, which we allocate according to the age group-specific data. Based upon this data, we sample from a distribution between 0 and 11 nights, with an average trip duration of 4.3 nights. For the present implementation, each trip destination is a random neighborhood within a random community (including workgroup-only communities), which results in a simple “gravity” model with no distance information. The destination community determines what types of contacts the traveler may have, in addition to the broad (but low-level) neighborhood and community-level mixing groups. During the daytime, the traveler may interact with his or her peers in play, school, or work groups if such contact groups exist at the destination tract; and at nighttime, the traveler may interact with a household and household cluster if traveling to one of the 78% of communities that are residential. E. School Dismissal Policies and ImplementationThe three key school dismissal policy questions are when (trigger), where (geographic breadth), and how long (duration) schools should be closed. We assume that the diagnosis of school children within a community will trigger the dismissal of schools within that community, and possibly more broadly, depending upon the specific policy. We assume that symptomatic cases are correctly diagnosed at a fixed rate of 1, 5, 10, or 20%, and that only a single diagnosed child is sufficient to trigger intervention. For instance, with a 1% diagnosis ratio, a trigger of 10 diagnosed school children isn’t guaranteed to be reached until there are 1,000 symptomatic children, although on average, this trigger would be reached after 500 symptomatic. Thresholds of 5, 10, 20, and 100 symptomatic school children per (2,000-person) community thus correspond to diagnosis ratios of 20, 10, 5, and 1%, respectively, if dismissal is triggered upon the first diagnosed child. Since no other actions (e.g., therapeutic antivirals, isolation, or quarantine) are taken following diagnosis, the diagnosis ratio and trigger are not independent, and the number of symptomatic school children in any community used to trigger that community’s school dismissal (and possibly that of neighboring communities) is the direct model parameter.Upon dismissal, all schools, preschools, and playgroups within the affected community (or communities) are closed, so no transmission occurs within these mixing groups. The greatest uncertainty concerns how contact rates (within different mixing groups and ages) might change during a school dismissal. These will presumably vary with time (as a so-called “fear-based social distancing” gradually decays toward normal contact rate patterns), and with severity (e.g., the greater case fatality ratio of pandemic scenarios C and D are more likely to lead to a greater acceptance of, and compliance with, recommended social distancing measures). Social contact surveys [Eam12] and mathematical model-based analysis of virological data [Ear12] during the 2009 summer and fall holiday breaks suggest that there is a reduction of at least 40-50% in contact and transmission among school-age children during such scheduled dismissals; additional precautions during unscheduled dismissals due to a spreading pandemic would likely increase this even further. One attempt to gather such data in the US during the spring 2009 H1N1 outbreak [Mil10] was stymied by the late May closure period, just before final exams and the summer break, which prevented the collection of any baseline contact data. In the face of this limited amount of survey and field study data to constrain assumptions, we utilize a range of assumptions that spans previous modeling work, and which is consistent with these studies. In particular, contact probabilities within school, preschool, and playgroup mixing groups are set to zero, and other child contacts are modified as described in Table S1, for two scenarios representing pessimistic (greatest amount of contact during closure) and optimistic (least plausible amount of contact) assumptions. This range of behavioral pattern changes will be used to provide confidence bands on the effectiveness of school dismissal policies for each scenario considered in this study.F. Other social distancing measuresAs school closure is the primary focus of this study, no other generic or workplace social distancing (e.g., telecommuting or liberal leave policies) or quarantine will be included. However, as is normally done in models such as EpiCast, we always assume that people who become ill will be likely to stay home sick, even in the “baseline” (non-intervention) scenarios. Individuals who self-isolate will only interact with household members, at the usual household contact rate, and will otherwise withdraw from all other mixing groups until recovered. The age-dependent withdrawal probabilities, and distribution of the number of days of illness before withdrawal, are those from early influenza model of Elveback et al [Elv76] and used in subsequent work [Hal02a,Lon04,Lon05,Ger06,Hal08]: pre-school children (ages 0-4 years) withdraw with cumulative probabilities 30%, 70%, and 80%, respectively, 1, 2, and 3 days after the onset of symptoms; school-age children (5-18 years) withdraw with cumulative probabilities 20%, 60%, and 75%, respectively, 1, 2, and 3 days after the onset of symptoms; and adults (19+ years old) withdraw with cumulative probabilities 10%, 40%, and 50%, respectively, 1, 2, and 3 days after the onset of symptoms.G. Vaccine usageNon-pharmaceutical interventions (NPIs) such as school dismissal, the focus of this study, and other social distancing measures, are predicted to primarily delay a pandemic spread, not reduce its cumulative impact (unless they are never lifted, which is unlikely). Therefore, pharmaceutical interventions such as a dynamic vaccination program are required to bring the disease spread under control, and the goal of NPIs should be to buy time for this. The main two unknowns regarding vaccination regard timing (when a vaccine is first available, and at what rate it will be produced and distributed during an outbreak), and its efficacy. Rather than dealing with the specifics of any particular vaccine (including a multiple-dose regimen, partial efficacy between the administration of doses, and gradual accumulation of immunity), we simply combine the times for production, distribution, and acquisition of immunity, and refer only to the date at which vaccination becomes fully effective, which may be either before or after the outbreak begins. In this study, we distribute vaccines randomly throughout the entire eligible population, consisting of all individuals who have not been vaccinated and are not currently symptomatic.In past work, we have usually considered two alternative production scenarios, either assuming the early distribution of a low-efficacy (e.g., a poorly matched) vaccine or the delayed production of a higher-efficacy vaccine. The well-matched vaccine is assumed to require two doses, and to have a vaccine efficacy for susceptibility VEs = 0.70 (with a reduced VEs = 0.50 for the elderly, age 65+), and a vaccine efficacy for infectiousness VEi = 0.80 (for all age groups). The poorly matched vaccine has VEs = 0.30 and VEi = 0.50, for all age groups. The interested reader is referred to previous studies [Hal02a] for the evidence supporting these parameter sets. Here, since the focus is on the effects of potential school dismissal policies, we assume that only a well-matched vaccine is used, but that it will not be available until 6 months after the first US index case. From that day forward, we assume a constant production and distribution rate of 14 million doses per week nationwide, effectively unlimited for the duration of the flu season (typically 8 months, or 240 days), rather than explicitly modeling the two doses – with some fraction of patients rejecting the booster – and the gradual development of immunity over time, we simply confer immunity on 1 million randomly selected eligible individuals per day. This assumed production rate lies between that achieved (10 Md/week) in the early stages of the 2009 H1N1 response [CDC10,PCAST] and the stated goal of 30-60 Md/week [PCAST]. We note that our assumed VE is based on that achieved for seasonal flu vaccines [Lon00], which represents an upper bound on the efficacy likely to be achieved for novel pandemic strains, and there are great uncertainties about the timing and efficacy of any vaccination campaign. However, since the focus of the present study is on the reduced and delayed spread achievable by school dismissal, our intent is to postulate some plausible vaccine availability and efficacy that begins to slow the transmission in our simulations, rather than to completely omit such a possibility from our scenarios. Hence, alternative vaccine scenarios are not considered here. Instead, we will measure cumulative attack rates both before (on day 180) and after (on day 240) vaccine introduction to separate the effects of school dismissal alone from that coupled with a vaccination campaign.H. EpiCast model parameterizationIn order to represent each of the four quadrants of the pandemic severity assessment framework, characteristics based upon historical (1918, 1957, 1968, and 2009) and potential H5N1-like pandemics were developed. As listed in Table S1, these included age-specific attack rate patterns, R0, and case fatality ratios (CFR). The fitting procedure for each of these scenarios started from the baseline EpiCast model contact rates, adjusting them and the overall transmission factor ptrans to give age-specific and overall attack rates within 1% of the specified values for the single-community model. For instance, if the specified childhood attack rate in one scenario is greater than normal, the school contact rate is increased accordingly. Similarly, to increase the working-age adult attack rate, the relative amount of workplace transmission is raised. The resulting parameters were then refined, if necessary, in the regional model, giving the epidemic curves in Fig. S2, and parameter sets (overall transmission probability ptrans and the relative changes in the baseline contact probabilities) in Table S1 for both the longer and shorter generation time.One difficulty arose in this fitting procedure for the less transmissible scenarios, particularly scenario C and to a lesser extent A: the slow epidemic spread (small R0) from one household to another, and one community to another, extended over several months for the regional, and even longer for the national-level model. As seen in Fig. S2, this can lead to an epidemic that has not yet peaked within the 6-month window before vaccines become available, further reducing the spread and eventually quenching it as R drops below 1. 2. Sensitivity and Uncertainty AnalysesAs noted in the main text, the contact matrix elements, and particularly how they might be expected to change during a pandemic alert, have been developed in a data-poor environment. This is particularly true for the contact patterns during an imposed school dismissal, which is the focus of this project. An analysis of the parameter sensitivity and resulting uncertainty in model predictions is therefore needed, but to our knowledge has not even been performed for the simple Elveback/Longini community model. As part of this project, we have begun such a sensitivity and uncertainty analysis, although further such work is clearly desirable. To do this, we follow the approach taken by Blower and Dowlatabadi [Blo94], who presented an uncertainty and sensitivity analysis method and applied it to a deterministic HIV disease transmission model.As a first step towards understanding the sensitivity of outcomes towards the different mixing group contact rates, we consider the baseline single-community model, with the initial contact matrix (and ptrans corresponding to R0 = 1.6) from [Ger06]. Attributing the source of each new infection to a single contact group for this baseline scenario, and a modified scenario in which school transmission is eliminated, we obtain the results shown in Fig. S3. Household transmission dominates (39.5%), followed by the age-appropriate daytime mixing group (school or work) (30.2%) and non-specific contact settings (30.3%). This is consistent with the pattern used in other modeling work, perhaps with household transmission slightly on the high side. For instance, Lee et al [Lee10] assert a “30–70 rule developed by Ferguson et al, which posits that 70% of all transmission occurred outside the household (which includes 33% that occurred in the general community and 37% in schools and workplaces).” (This is of course by design, as such an attribution distribution was used to develop the individual contact probabilities, as described above in Section 1A.) Closing schools and keeping children at home only further increases the role of household transmission, increasing its responsibility to over half of all cases. Similar infection source distributions are seen for the five pandemic scenarios of the current study, as shown in Fig. S3.As a second step, we perform an uncertainty/sensitivity analysis of this baseline single-community model to the assumed contact rates (for the R0 = 1.6 model from [Ger06], since the contact rate modifications in Table S1 for scenarios A-D do not significantly change the conclusions, as Fig. S3 indicates). Uniform distributions varying the default contact probabilities by +/-50% for six contact group categories are considered: households, household clusters, neighborhoods, communities, schools, and workplaces. Latin hypercube sampling is used to generate 1000 parameter sets, from which the 2000-person single-community model is run with 10 communities for each parameter set, and the results averaged.The results are summarized as a “tornado plot” of partial rank correlation coefficients (PRCCs) between the 6 independent input variables (scaling factors for these 6 contact groups) and the number of days between the index case and peak incidence in Fig. S4. In addition, scatter plots illustrating the dependence of cumulative attack rate (not shown) demonstrate that the assumed household, school, and workplace contacts (in that order) have the greatest impact on the resulting attack rate, as expected, with non-specific community transmission (which contributes to roughly a quarter of all cases, as seen in Fig. S3) following closely behind.The relationship between these contact matrix elements and the epidemic timing is even more interesting. The PRCC, which measures the sensitivity of output variables to inputs, indicates that school transmission has by far the largest impact on the number of days from initial outbreak to peak incidence, followed by household transmission. (The negative values simply indicate that for increasing contact rates, the time to peak incidence decreases. It is curious that the workplace contact rate PRCC has the opposite sign, but its small magnitude may indicate that this is merely a statistical fluke.) Consistent with other observations, this clearly shows that reducing school transmission is the greatest lever for slowing disease spread, and buying time for a vaccine to be developed and distributed.3. Epidemic Source ConsiderationsFor national-scale simulations, the manner of introduction of a pandemic influenza strain into the continental United States must be considered. In particular, such a human-transmissible strain may emerge either domestically or overseas, in both cases most likely in a rural area. A domestic strain, such as one that might emerge from a poultry farm in the DelMarVa (Delaware-Maryland-Virginia) peninsula, or a hog farm in the Midwest, would likely take longer to become established than a foreign strain which was already spreading from human to human before arriving in the U.S.; the only difference is whether these initial generations, as the epidemic slowly spreads through the more dispersed rural population before reaching a dense urban population where it can thrive, are taking place in (for instance) rural Thailand or rural Arkansas.In the case of a foreign strain, it would likely be introduced into the United States through arriving international air passengers, potentially preferentially those arriving on the East Coast (e.g. from Africa) or West Coast (e.g. from Asia); or overland from Mexico or Canada. In our previous work, we found that introduction via air travel into major metropolitan areas (at a constant rate of 1-10 potential infecteds per 10,000 daily passengers at each of the 14 largest U.S. international airports), or point source introductions into large cities (either New York or Los Angeles), result in nearly identical national-level incidence rates, with only a difference in the rapid spatiotemporal spread [Ger06]. Consequently, we model the scenario of an overseas emergence of a pandemic strain by the introduction via arriving international air passengers (2 per 10,000), a rate comparable with that used for the regional model.For the emergence of a domestic strain, it is important to realize that the predicted distribution of attack rates will be bimodal, with a finite probability of extinction. The mean predicted attack rate (or more general consequences) thus should only represent simulations where the outbreak was NOT extinguished, increasing the number of index cases (>1) if necessary to reduce the probability of extinction. Shown in Figs. S7 and S8, respectively, are epidemic curves and snapshots of the incidence 120 days after introduction of the index case, for either the overseas or domestic emergence, represented by the introduction of 10 infected individuals into Sussex County, Delaware, a large poultry-farming region on the DelMarVa peninsula. Two stochastic realizations of the latter scenario are shown, to indicate the variability. Depending on how “close” the nearest urban area is, a rural domestic emergence may lag 1-3 months behind urban introductions, in which case the rural-to-urban migration has already occurred overseas. Based on these observations, we choose to focus on the overseas introduction scenario for this study. The combination of urban and rural regions of the U.S. also affects its subsequent spread following introduction, as illustrated in Fig. S9, which compares the baseline epidemic curves for several pandemic scenarios, both in the regional and national models. Two key points are worth noting. First, the national-level model involves a simultaneous introduction into 13 urban areas (two of the largest airports, JFK and Newark, are both near New York City), and thus averages out the stochastic variability of a single urban introduction, such as in the regional model. For this reason, the stochastic variability of the national-level model in Fig. S9 is much less than that of the regional model (for which 5 realizations are shown for each scenario), and in fact is comparable to the thickness of the US curves, which is why only one realization is shown for each scenario. Second, the outbreaks around these initial 13 urban areas are followed by a spread to surrounding rural and urban regions, reversing the rural-to-urban spread of a rural domestic emergence. For this reason, a national-level outbreak naturally extends over a longer duration (by several weeks to a month or more) than a localized regional outbreak.4. Remarks on Pandemic Scenario CThis delayed spread impacts our study of pandemic scenario C, which for the regional model is only reaching its epidemic peak around 6 months after the first introduction. Since the assumed availability of an effective vaccine at this point in time begins to slow and ultimately stamp out the outbreak, the slower spread of scenario C exhibits a sharply reduced cumulative attack rate for the national model. This is also illustrated in Fig. S10, which shows the epidemic curves and number of schools closed for pandemic scenario C, using the same dismissal trigger and duration as used for pandemic scenarios B2 and D in Figs. 3 and S6. We note that the baseline (no dismissal) epidemic curve peaks after 180 days, when it begins to be quenched by vaccine availability. With an individual school dismissal policy, few schools reach the threshold to trigger closure. On the other hand, more proactive school dismissals, such as those over multi-county regions, will still be triggered since it is more likely that at least one school in the region will reach the threshold, although this is often early in the outbreak and well before the epidemic peak.References[Bar05]C. L. Barrett, S. G. Eubank, and J. P. Smith, If smallpox strikes Portland ..., Sci Am. 292(3):42-9 (2005).[Blo94]S. M. Blower and H. Dowlatabadi, Sensitivity and Uncertainty Analysis of Complex-Models of Disease Transmission – an HIV Model, as an Example, International Statistical Review 62, 229-243 (1994).[Bro11]S. T. Brown, J. H. Y. Tai, R. R. Bailey, P. C. Cooley, W. D. Wheaton, M. A. Potter, R. E. Voorhees, M. LeJeune, J. J. Grefenstette, D. S. Burke, S. M. McGlone, and B. Y. Lee, Would school closure for the 2009 H1N1 influenza epidemic have been worth the cost?: a computational simulation of Pennsylvania, BMC Public Health 11, 353 (2011). [Car02]F. Carrat, C. Sahler, S. Rogez, M. Leruez-Ville, F. Freymuth, C. Le Gales, M. Bungener, B. Housset, M. Nicolas, and C. Rouzioux, Influenza burden of illness: estimates from a national prospective survey of household contacts in France, Arch. Intern. Med. 162, 1842-8 (2002). doi:10.1001/archinte.162.16.1842[Car08]F. Carrat, E. Vergu, N. M. Ferguson, M. Lemaitre, S. Cauchemez, S. Leach, and A.-J. Valleron, Time Lines of Infection and Disease in Human Influenza: A Review of Volunteer Challenge Studies, Am. J. Epidemiol. 167, 775 (2008). doi: 10.1093/aje/kwm375[Cow09]B. J. Cowling, V. J. Fang, S. Riley, J. S. Malik Peiris, and G. M. Leung, Estimation of the serial interval of influenza, Epidemiology 20, 344 (2009).[Eam12]K. T. Eames, N. L. Tilston, E. Brooks-Pollock, and W. J. Edmunds, Measured dynamic social contact patterns explain the spread of H1N1v influenza, PLOS Comput Biol. 8, e1002425 (2012). doi:10.1016/S1473-3099(09)70176-8.[Ear12]D. J. D. Earn, D. He, M. B. Loeb, K. Fonseca, B. E. Lee, and J. Dushoff, Effects of School Closure on Incidence of Pandemic Influenza in Alberta, Canada, Ann. Intern. Med.156, 173 (2012).[Elv76]L. R. Elveback, J. P. Fox, E. Ackerman, A. Langworthy, M. Boyd, and L. Gatewood, An influenza simulation model for immunization studies, Am. J. Epidemiol. 103, 152 (1976).[Eub04]S. Eubank, H. Guclu, V. S. Anil Kumar, M. V. Marathe, A. Srinivasan, Z. Toroczkai, and N. Wang, Modelling disease outbreaks in realistic urban social networks, Nature 429, 180-184 (2004). doi: 10.1038/nature02541.[Fer05]N. M. Ferguson, D. A. Cummings, S. Cauchemez, C. Fraser, S. Riley, A. Meeyai, S. Iamsirithaworn, and D. S. Burke, Strategies for containing an emerging influenza pandemic in Southeast Asia, Nature 437, 209 (2005).[Fer06]N. M. Ferguson, D. A. T. Cummings, C. Fraser, J. C. Cajka, P. C. Cooley, and D. S. Burke, Strategies for mitigating an influenza pandemic, Nature 442, 448 (2006). doi: 10.1038/nature04795.[Ger06]T. C. Germann, K. Kadau, I. M. Longini, and C. A. Macken, Mitigation Strategies for Pandemic Influenza in the United States, Proc. Natl. Acad. Sci. (USA) 103, 5935-5940 (2006).[Hal02a]M. E. Halloran, I. M. Longini, D. M. Cowart, and A. Nizam, Community interventions and the epidemic prevention potential, Vaccine 20, 3254 (2002).[Hal02b]M. E. Halloran, I. M. Longini, A. Nizam, and Y. Yang, Containing Bioterrorist Smallpox, Science 298, 1428 (2002).[Hal08]M. E. Halloran, N. M. Ferguson, S. Eubank, I. M. Longini, D. A. T. Cummings, B. Lewis, S. Xu, C. Fraser, A. Vullikanti, T. C. Germann, D. Wagener, R. Beckman, K. Kadau, C. Barrett, C. A. Macken, D. S. Burke, and P. Cooley, Modeling Targeted Layered Containment of an Influenza Pandemic in the USA, Proc. Natl. Acad. Sci. (USA) 105, 4639-4644 (2008).[Hay14]A. C. Hayward, E. B. Fragaszy, A. Bermingham, et al. on behalf of the Flu Watch Group, Comparative community burden and severity of seasonal and pandemic influenza: results of the Flu Watch cohort study, Lancet Respir. Med. (2014). (14)70034-7 [Kil75]E. D. Kilbourne, The Influenza Viruses and Influenza (Academic, New York, 1975).[Lee10]B. Y. Lee et al, Simulating School Closure Strategies to Mitigate an Influenza Epidemic, J. Pub. Health Manag. Pract. 16(3), 252-261 (2010).[CDC10]Centers for Disease Control and Prevention, 2009 H1N1 Vaccine Doses Allocated, Ordered, and Shipped by Project Area (2010). [PCAST]President’s Council of Advisors on Science and Technology, Report To The President On Reengineering The Influenza Vaccine Production Enterprise To Meet The Challenges Of Pandemic Influenza (2010). [Lon00]I. M. Longini, M. E. Halloran, A. Nizam, et al., Estimation of the efficacy of live, attenuated influenza vaccine from a two-year, multi-center vaccine trial: implications for influenza epidemic control, Vaccine 18, 1902–9 (2000).[Lon04]I. M. Longini, M. E. Halloran, A. Nizam, and Y. Yang, Containing Pandemic Influenza with Antiviral Agents, Am. J. Epidemiol. 159, 623 (2004).[Lon05]I. M. Longini, A. Nizam, S. Xu, K. Ungchusak, W. Hanshaoworakul, D. A. T. Cummings, and M. E. Halloran, Containing Pandemic Influenza at the Source, Science 309, 1083 (2005). doi: 10.1126/science.1115717[Mil08]G. J. Milne, J. K. Kelso, H. A. Kelly, S. T. Huband, and J. McVernon, 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 3, e4005 (2008).[Mil10]J. C. Miller, L. Danon, J. J. O'Hagan, E. Goldstein, M. Lajous, and M. Lipsitch, Student Behavior during a School Closure Caused by Pandemic Influenza A/H1N1, PLoS ONE 5, e10425 (2010). doi:10.1371/journal.pone.0010425[Mil13]G. J. Milne, N. Halder, and J. K. Kelso, The Cost Effectiveness of Pandemic Influenza Interventions: A Pandemic Severity Based Analysis, PLoS ONE 8, e61504 (2013). doi:10.1371/journal.pone.0061504[Mon85]A. S. Monto, J. S. Koopman, and I. M. Longini, Tecumseh study of illness. XIII. Influenza infection and disease, 1976–1981, Am. J. Epidemiol. 121, 811–22 (1985).[Mos79]M. R. Moser, T. R. Bender, H. S. Margolis, G. R. Noble, A. P. Kendal, and D. G. Ritter, An outbreak of influenza aboard a commercial airliner, Am. J. Epidemiol. 110, 1-6 (1979).[Str07]P. D. Stroud, S. Y. Del Valle, S. J. Sydoriak, J. Riese, and S. Mniszewski, Spatial Dynamics of Pandemic Influenza in a Massive Artificial Society, J. Artif. Soc. Soc. Simulat. 10, 9 (2007). [Lon04]I. M. Longini, M. E. Halloran, A. Nizam, and Y. Yang, Containing Pandemic Influenza with Antiviral Agents, Am. J. Epidemiol. 159, 623 (2004).Figure CaptionsFigure S1. (a, Top) Elementary school mixing group size distributions using the national average age and household size distributions, the distributions within each census tract, and the census track distributions with a single elementary school (rather than two) in half of the model communities, introducing a long tail in mixing group size distributions. (b, Bottom) Elementary, Middle, and High school mixing group size distributions using either national average or tract-specific household size and age distributions.Figure S2. Epidemic curves for five stochastic realizations of each baseline pandemic scenario for the regional model. (Top) Shorter serial interval; (Bottom) longer serial interval.Figure S3. Infection source attribution for original EpiCast contact matrix (R0 = 1.6 baseline scenario) [Ger06] and the five pandemic scenarios of the current study, with schools either open (top) or closed (bottom).Figure S4: Partial rank correlation coefficient analysis of sensitivity of the epidemic timing with respect to contact rates in different settings: schools (elementary, middle, and high), households, workplaces, household clusters (immediate neighbors), neighborhood, and community. Figure S5: US model predictions of the number of (symptomatic) influenza cases averted by school dismissal strategies, for the longer serial interval. School dismissal is activated when one symptomatic child is diagnosed at an assumed diagnosis ratio of 5%, 10%, or 20%. Two alternative assumptions for contact rates during school dismissal are considered: “worst-case” (filled bars) and “best-case” (extensions).Figure S6: US model results for pandemic scenarios B2 (left panels) and D (right panels) with school dismissal activated when 20 children are symptomatic (closure upon first diagnosed at a 5% diagnosis ratio) and a four-week duration, for the shorter serial interval. (Top) Epidemic curves. (Bottom) Number of schools closed at any time during the outbreak. The “best-case” assumption for contact rates during school dismissal is used (contact rates involving children in households are unchanged, and child-related contacts outside the home are reduced by 50%). Beginning on day 180, one million people per day are vaccinated (see text and SI for details). Note that for scenario D, the multi-county and state-wide dismissals are virtually indistinguishable, particularly for the epidemic curves.Figure S7: Epidemic curves for the baseline (no intervention) spread of pandemic influenza within the continental US, either via arriving international air passengers (solid curves) or via a point source introduction of 10 infected individuals into Sussex County, Delaware, a poultry farming region on the DelMaVa peninsula (dashed curves, with two realizations for scenario D which differ only in the initial random number generator seed).Figure S8: Prevalence 120 days after introduction of a novel pandemic strain (scenario D) into the continental US, either via arriving international air passengers (top) or via a point source introduction of 10 infected individuals into the DelMaVa peninsula (bottom two panels, differing only via the random number generator seed). Each census tract is shown as a dot colored according to the current prevalence, on a logarithmic scale from white for 0.03% or fewer ill people per capita, to red for 3% or greater.?Figure S9: Epidemic curves for baseline pandemic scenarios B1, B2, C, and D, comparing the US model (bold curves from a single realization) with 5 stochastic realizations of each baseline pandemic scenario for the regional model (thin curves).Figure S10: U.S. model results for pandemic scenario C with school dismissal activated when 20 children are symptomatic (closure upon first diagnosed at a 5% diagnosis ratio) and a 4-week duration, for the shorter serial interval. (Top) Epidemic curves. (Bottom) Number of schools closed at any time during the outbreak. The “worst-case” assumption for contact rates during school dismissal is used (contact rates involving children in households are doubled, and child-related contacts outside the home are reduced by 30%). Beginning on day 180, 1 million people per day are vaccinated (see text and SI for details).Table CaptionsTable S1: Summary of specific EpiCast model parameters for this study (see [Ger06] for further details).Table S2: Cumulative attack rates (AR) with the reduction from the baseline scenario in parentheses (?) using the regional model, with the shorter (mean ~2.8 day) serial interval and worst-case contact pattern change (a 30% reduction in non-household contacts and doubling of child-related household contacts) during school dismissal.Table S3: Cumulative attack rates (AR) with the reduction from the baseline scenario in parentheses (?) using the regional model, with the longer (mean ~4.0 day) serial interval and best-case contact pattern change (a 50% reduction in non-household) during school dismissal.Table S4: Cumulative attack rates (AR) with the reduction from the baseline scenario in parentheses (?) using the regional model, with the longer (mean ~4.0 day) serial interval and worst-case contact pattern change (a 30% reduction in non-household contacts and doubling of child-related household contacts) during school dismissal.Table S5: Cumulative attack rates (AR) through 180 days with the reduction from the baseline scenario in parentheses (?) using the regional model, with the shorter (mean ~2.8 day) serial interval and worst-case contact pattern change (a 30% reduction in non-household contacts and doubling of child-related household contacts) during school dismissal.Table S6: Cumulative attack rates (AR) through 180 days with the reduction from the baseline scenario in parentheses (?) using the regional model, with the shorter (mean ~2.8 day) serial interval and best-case contact pattern change (a 50% reduction in non-household) during school dismissal.Table S7: Cumulative attack rates (AR) through 180 days with the reduction from the baseline scenario in parentheses (?) using the regional model, with the longer (mean ~4.0 day) serial interval and worst-case contact pattern change (a 30% reduction in non-household contacts and doubling of child-related household contacts) during school dismissal.Table S8: Cumulative attack rates (AR) through 180 days with the reduction from the baseline scenario in parentheses (?) using the regional model, with the longer (mean ~4.0 day) serial interval and best-case contact pattern change (a 50% reduction in non-household) during school dismissal.Figure S1. (a, Top) Elementary school mixing group size distributions using the national average age and household size distributions, the distributions within each census tract, and the census track distributions with a single elementary school (rather than two) in half of the model communities, introducing a long tail in mixing group size distributions. (b, Bottom) Elementary, Middle, and High school mixing group size distributions using either national average or tract-specific household size and age distributions.Figure S2. Epidemic curves for five stochastic realizations of each baseline pandemic scenario for the regional model. (Top) Shorter serial interval; (Bottom) longer serial interval.Figure S3. Infection source attribution for original EpiCast contact matrix (R0 = 1.6 baseline scenario) [Ger06] and the five pandemic scenarios of the current study, with schools either open (top) or closed (bottom).Figure S4: Partial rank correlation coefficient analysis of sensitivity of the epidemic timing with respect to contact rates in different settings: schools (elementary, middle, and high), households, workplaces, household clusters (immediate neighbors), neighborhood, and community.Figure S5: US model predictions of the number of (symptomatic) influenza cases averted by school dismissal strategies, for the longer serial interval. School dismissal is activated when one symptomatic child is diagnosed at an assumed diagnosis ratio of 5%, 10%, or 20%. Two alternative assumptions for contact rates during school dismissal are considered: “worst-case” (filled bars) and “best-case” (extensions).Figure S6: US model results for pandemic scenarios B2 (left panels) and D (right panels) with school dismissal activated when 20 children are symptomatic (closure upon first diagnosed at a 5% diagnosis ratio) and a four-week duration, for the shorter serial interval. (Top) Epidemic curves. (Bottom) Number of schools closed at any time during the outbreak. The “best-case” assumption for contact rates during school dismissal is used (contact rates involving children in households are unchanged, and child-related contacts outside the home are reduced by 50%). Beginning on day 180, one million people per day are vaccinated (see text and SI for details). Note that for scenario D, the multi-county and state-wide dismissals are virtually indistinguishable, particularly for the epidemic curves.Figure S7: Epidemic curves for the baseline (no intervention) spread of pandemic influenza within the continental US, either via arriving international air passengers (solid curves) or via a point source introduction of 10 infected individuals into Sussex County, Delaware, a poultry farming region on the DelMaVa peninsula (dashed curves, with two realizations for scenario D which differ only in the initial random number generator seed).Figure S8: Prevalence 120 days after introduction of a novel pandemic strain (scenario D) into the continental US, either via arriving international air passengers (top) or via a point source introduction of 10 infected individuals into the DelMaVa peninsula (bottom two panels, differing only via the random number generator seed). Each census tract is shown as a dot colored according to the current prevalence, on a logarithmic scale from white for 0.03% or fewer ill people per capita, to red for 3% or greater.Figure S9: Epidemic curves for baseline pandemic scenarios B1, B2, C, and D, comparing the US model (bold curves from a single realization) with 5 stochastic realizations of each baseline pandemic scenario for the regional model (thin curves).Figure S10: U.S. model results for pandemic scenario C with school dismissal activated when 20 children are symptomatic (closure upon first diagnosed at a 5% diagnosis ratio) and a 4-week duration, for the shorter serial interval. (Top) Epidemic curves. (Bottom) Number of schools closed at any time during the outbreak. The “worst-case” assumption for contact rates during school dismissal is used (contact rates involving children in households are doubled, and child-related contacts outside the home are reduced by 30%). Beginning on day 180, 1 million people per day are vaccinated (see text and SI for details).Table S1: Summary of specific EpiCast model parameters for this study (see [Ger06] for further details).General ParametersDisease natural history parameter distributionsIncubation periodLong SIShort SIInfectious periodLong SIShort SI1 day30%16%1 day70%2 days50%70%2 days15%3 days20%14%3 days30%6%4 days40%4%5 days20%3%Mean incubation1.9 days1.98 days6 days10%2%Mean latent1.2 days1.98 daysMean infectious4.1 days1.61 daysContact changes during dismissalWorstBestIncreased home contact during school dismissal: household contact probabilities involving children (child-child and child-adult) are increased by the following amount:100%0%Reduced contact outside household during school dismissal: contact probabilities within school, preschool, and playgroup mixing groups are set to zero, and other child-related contacts (neighborhood cluster, neighborhood, and community mixing groups) are reduced by the following amount:30%50%Scenario-specific parametersAge-specific attack ratesAB1B2CD Child (0-18 years)32%39%50%18%54% Adult (19-64 years)15%18%23%8%25% Elderly (65+ years)7%8%11%4%12% Overall18%22%28%10%30%R01.31.51.81.32.0Case fatality ratio0.03%0.05%0.1%1%≥1%Relative contact rates (compared with Ger06) for the longer serial intervalTransmission probability given contact0.0930.1010.1050.090.115Relative high school contact1.01.01.00.51.0Relative workplace contact1.01.00.61.01.0Relative community contact1.61.01.11.41.1Relative contact rates (compared with Ger06) for the shorter serial intervalTransmission probability given contact0.230.230.270.2150.245Relative elementary school contact0.80.831.00.831.33Relative middle school contact0.80.830.830.671.33Relative high school contact0.630.830.830.671.0Relative workplace contact1.01.00.51.01.0Relative community contact1.311.721.521.721.38Table S2: Cumulative attack rates (AR) with the reduction from the baseline scenario in parentheses (?) using the regional model, with the shorter (mean ~2.8 day) serial interval and worst-case contact pattern change (a 30% reduction in non-household contacts and doubling of child-related household contacts) during school dismissal.Pandemic scenarioBaselineClinical Attack Rate TriggerCommunity School DismissalRegional School DismissalDiag. ratioDuration Duration 1 wk2 wks4 wks8 wks12 wks1 wk2 wks4 wks8 wks12 wksAR(δ)AR(δ)AR(δ)AR(δ)AR(δ) AR(δ) AR(δ) AR(δ) AR(δ) AR(δ)A(2009-like)18.4%1%18.4 (0.0)18.4 (0.0)18.4 (0.0)18.4 (0.0)18.4 (0.0)18.4 (0.0)18.4 (0.0)18.4 (0.0)18.4 (0.0)18.4 (0.0)5%17.8 (0.6)17.4 (1.1)17.2 (1.3)17.1 (1.4)17.1 (1.3)18.2 (0.2)18.0 (0.4)16.9 (1.5)11.9 (6.5)5.7 (12.7)10%16.5 (2.0)15.2 (3.2)14.2 (4.2)13.5 (4.9)13.2 (5.2)18.2 (0.2)17.9 (0.6)16.8 (1.7)12.4 (6.0)5.9 (12.6)20%15.3 (3.1)13.0 (5.4)10.6 (7.8)8.7 (9.7)8.1 (10.4)18.2 (0.2)17.9 (0.5)17.0 (1.4)12.7 (5.7)7.5 (10.9)B1(1968-like)22.6%1%22.6 (0.0)22.6 (0.0)22.6 (0.0)22.6 (0.0)22.6 (0.0)22.6 (0.0)22.5 (0.0)22.6 (0.0)22.6 (0.0)22.6 (0.0)5%21.9 (0.7)21.4 (1.2)21.2 (1.4)20.9 (1.7)20.9 (1.7)22.6 (0.0)22.4 (0.2)22.3 (0.3)20.6 (2.0)15.1 (7.5)10%20.9 (1.7)19.8 (2.8)18.5 (4.1)17.0 (5.6)16.5 (6.1)22.5 (0.0)22.5 (0.1)22.3 (0.3)20.7 (1.8)15.1 (7.5)20%20.7 (1.9)19.2 (3.3)16.8 (5.8)13.6 (9.0)11.8 (10.8)22.5 (0.0)22.5 (0.1)22.4 (0.2)20.8 (1.8)16.5 (6.1)B2(1957-like)28.3%1%28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)5%26.9 (1.4)26.0 (2.3)25.1 (3.2)24.5 (3.8)24.4 (3.9)28.3 (0.0)28.3 (0.0)28.3 (0.1)28.0 (0.3)25.6 (2.7)10%26.5 (1.8)25.3 (3.1)23.2 (5.1)19.9 (8.4)18.5 (9.9)28.3 (0.0)28.3 (0.0)28.2 (0.1)28.0 (0.3)25.9 (2.5)20%27.1 (1.3)26.1 (2.2)24.2 (4.1)18.7 (9.7)14.3 (14.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.1 (0.2)26.3 (2.0)C(H5N1-like)12.1%1%12.1 (0.0)12.1 (0.0)12.1 (0.0)12.1 (0.0)12.1 (0.0)12.1 (0.0)12.1 (0.0)12.1 (0.0)12.1 (0.0)12.1 (0.0)5%11.9 (0.2)11.6 (0.5)11.4 (0.7)11.3 (0.8)11.5 (0.611.5 (0.5)10.6 (1.5)8.9 (3.2)4.7 (7.4)1.8 (10.3)10%10.6 (1.5)9.4 (2.7)9.0 (3.1)8.8 (3.3)8.8 (3.3)11.6 (0.5)11.1 (1.0)9.3 (2.8)5.3 (6.8)2.2 (9.9)20%8.9 (3.2)6.5 (5.6)5.4 (6.7)4.9 (7.2)4.6 (7.5)11.8 (0.3)11.6 (0.5)9.5 (2.6)5.6 (6.5)2.4 (9.6)D(1918-like)30.1%1%30.1 (0.0)30.1 (0.0)30.1 (0.0)30.1 (0.0)30.1 (0.0)29.9 (0.2)29.8 (0.3)29.7 (0.4)28.3 (1.8)27.4 (2.7)5%28.4 (1.7)27.3 (2.8)26.1 (4.1)24.9 (5.2)24.6 (5.5)30.1 (0.0)30.1 (0.0)30.1 (0.0)30.1 (0.0)29.7 (0.4)10%28.6 (1.5)27.5 (2.6)25.8 (4.3)21.5 (8.6)19.1 (11.0)30.1 (0.0)30.1 (0.0)30.1 (0.0)30.1 (0.0)29.8 (0.3)20%29.2 (0.9)28.6 (1.5)27.7 (2.4)23.5 (6.6)16.6 (13.5)30.1 (0.0)30.1 (0.0)30.1 (0.0)30.1 (0.0)29.9 (0.2)Table S3: Cumulative attack rates (AR) with the reduction from the baseline scenario in parentheses (?) using the regional model, with the longer (mean ~4.0 day) serial interval and best-case contact pattern change (a 50% reduction in non-household) during school dismissal.Pandemic scenarioBaselineClinical Attack Rate TriggerCommunity School DismissalRegional School DismissalDiag. ratioDuration Duration 1 wk2 wks4 wks8 wks12 wks1 wk2 wks4 wks8 wks12 wksAR(δ)AR(δ)AR(δ)AR(δ)AR(δ) AR(δ) AR(δ) AR(δ) AR(δ) AR(δ)A(2009-like)19.1%1%19.1 (0.0)19.1 (0.0)19.1 (0.0)19.1 (0.0)19.0 (0.0)17.7 (1.4)16.0 (3.0)12.1 (7.0)5.6 (13.4)3.8 (15.3)5%16.8 (2.3)15.4 (3.6)14.3 (4.8)13.5 (5.6)13.3 (5.8)18.7 (0.4)17.9 (1.2)15.7 (3.4)9.5 (9.5)4.1 (14.9)10%14.8 (4.3)12.3 (6.8)9.7 (9.4)8.3 (10.8)7.8 (11.3)18.6 (0.5)18.1 (1.0)16.2 (3.4)10.4 (8.6)4.2 (14.9)20%14.0 (5.1)9.6 (9.4)5.7 (13.3)4.2 (14.9)3.6 (15.5)18.6 (0.4)18.1 (1.0)16.5 (2.8)11.0 (8.0)5.6 (13.5)B1(1968-like)22.2%1%22.2 (0.0)22.2 (0.0)22.2 (0.0)22.1 (0.0)22.2 (0.0)21.6 (0.6)20.8 (1.3)18.1 (4.1)8.6 (13.5)3.0 (19.2)5%19.8 (2.4)18.3 (3.9)16.7 (5.5)15.3 (6.8)14.8 (7.4)22.0 (0.1)21.5 (0.7)19.7 (2.5)13.6 (8.5)6.1 (16.1)10%18.7 (3.5)15.7 (6.5)12.4 (9.7)10.0 (12.2)8.9 (13.3)22.0 (0.2)21.6 (0.6)20.5 (1.7)15.2 (7.0)7.5 (14.7)20%18.2 (3.9)13.9 (8.2)8.8 (13.4)5.7 (16.4)4.6 (17.5)22.0 (0.2)21.7 (0.5)20.2 (1.9)15.6 (6.6)8.5 (13.7)B2(1957-like)28.2%1%28.0 (0.2)27.8 (0.3)27.8 (0.3)27.8 (0.3)27.8 (0.4)28.2 (0.0)28.1 (0.0)27.9 (0.2)24.2 (3.9)18.6 (9.5)5%25.6 (2.6)23.8 (4.4)21.0 (7.1)16.3 (11.9)13.7 (14.5)28.2 (0.0)28.2 (0.0)28.0 (0.2)25.9 (2.2)18.6 (9.5)10%26.3 (1.9)24.3 (3.8)19.9 (8.3)13.3 (14.9)9.2 (18.9)28.2 (0.0)28.1 (0.0)28.0 (0.2)26.2 (2.0)19.5 (8.7)20%26.9 (1.3)24.4 (3.8)17.0 (11.1)9.3 (18.9)5.8 (22.4)28.2 (0.0)28.2 (0.0)28.1 (0.1)26.5 (1.6)19.6 (8.6)C(H5N1-like)11.6%1%11.6 (0.0)11.6 (0.0)11.6 (0.0)11.6 (0.0)11.6 (0.0)10.3 (1.3)9.4 (2.2)8.2 (3.4)7.5 (4.1)7.4 (4.2)5%9.6 (2.0)8.7 (2.9)7.9 (3.7)7.8 (3.8)7.7 (3.9)10.6 (1.0)9.5 (2.1)6.7 (4.9)3.1 (8.5)1.2 (10.4)10%7.4 (4.2)5.3 (6.3)3.9 (7.7)3.8 (7.8)3.7 (7.9)11.2 (0.4)9.6 (2.0)7.8 (3.8)4.0 (7.6)1.4 (10.2)20%5.6 (6.0)3.1 (8.5)1.7 (9.9)1.4 (10.2)1.2 (10.4)10.9 (0.7)10.1 (1.5)8.4 (3.2)4.5 (7.1)1.8 (9.8)D(1918-like)29.8%1%29.7 (0.1)29.7 (0.1)29.7 (0.1)29.7 (0.1)29.6 (0.2)29.8 (0.0)29.8 (0.0)29.6 (0.2)26.9 (2.9)16.6 (13.2)5%27.4 (2.4)26.0 (3.8)23.6 (6.2)19.0 (10.8)16.4 (13.4)29.8 (0.0)29.8 (0.0)29.7 (0.1)28.1 (1.7)21.2 (8.6)10%28.1 (1.7)26.7 (3.1)23.8 (6.0)17.0 (12.8)11.7 (18.1)29.8 (0.0)29.8 (0.0)29.7 (0.1)28.5 (1.3)22.7 (7.1)20%28.7 (1.1)27.3 (2.5)22.9 (6.9)14.6 (15.2)8.7 (21.1)29.8 (0.0)29.8 (0.0)29.7 (0.1)28.7 (1.1)23.4 (6.4)Table S4: Cumulative attack rates (AR) with the reduction from the baseline scenario in parentheses (?) using the regional model, with the longer (mean ~4.0 day) serial interval and worst-case contact pattern change (a 30% reduction in non-household contacts and doubling of child-related household contacts) during school dismissal.Pandemic scenarioBaselineClinical Attack Rate TriggerCommunity School DismissalRegional School DismissalDiag. ratioDuration Duration 1 wk2 wks4 wks8 wks12 wks1 wk2 wks4 wks8 wks12 wksAR(δ)AR(δ)AR(δ)AR(δ)AR(δ) AR(δ) AR(δ) AR(δ) AR(δ) AR(δ)A(2009-like)19.1%1%19.1 (0.0)19.0 (0.0)19.1 (0.0)19.0 (0.0)19.0 (0.0)18.9 (0.2)18.1 (1.0)16.4 (2.7)12.1 (6.9)8.8 (10.3)5%18.2 (0.9)17.5 (1.6)16.7 (2.4)16.2 (2.9)16.1 (3.0)19.0 (0.1)18.7 (0.4)17.6 (1.4)14.4 (4.7)10.0 (9.1)10%17.5 (1.5)15.8 (3.2)14.0 (5.1)12.6 (6.5)12.1 (7.0)19.0 (0.1)18.6 (0.5)17.8 (1.2)14.8 (4.3)10.6 (8.4)20%17.0 (2.1)14.5 (4.5)11.5 (7.6)9.3 (9.8)8.2 (10.8)19.0 (0.1)18.7 (0.4)17.9 (1.2)15.1 (3.9)10.8 (8.2)B1(1968-like)22.2%1%22.2 (0.0)22.2 (0.0)22.2 (0.0)22.2 (0.0)22.2 (0.0)22.0 (0.1)21.7 (0.5)20.8 (1.4)16.4 (5.7)10.4 (11.7)5%21.2 (1.0)20.2 (1.9)19.1 (3.1)18.3 (3.8)18.0 (4.2)22.2 (0.0)21.9 (0.3)21.3 (0.8)18.8 (3.4)14.0 (8.2)10%20.8 (1.4)19.2 (2.9)17.0 (5.1)14.5 (7.6)13.6 (8.6)22.1 (0.0)21.9 (0.3)21.3 (0.8)19.1 (3.1)14.6 (7.6)20%20.8 (1.4)18.7 (3.4)15.2 (6.9)11.4 (10.8)9.7 (12.4)22.1 (0.1)21.9 (0.2)21.5 (0.7)19.1 (3.1)14.6 (7.6)B2(1957-like)28.2%1%28.1 (0.0)28.1 (0.1)28.1 (0.1)28.1 (0.1)28.1 (0.1)28.2 (0.0)28.2 (0.0)28.1 (0.1)27.2 (0.9)22.1 (6.1)5%26.9 (1.2)25.6 (2.5)23.6 (4.5)20.3 (7.9)18.7 (9.5)28.2 (0.0)28.2 (0.0)28.1 (0.0)27.5 (0.7)23.2 (5.0)10%27.3 (0.9)26.2 (1.9)24.0 (4.2)20.3 (7.9)18.7 (9.5)28.2 (0.0)28.2 (0.0)28.1 (0.0)27.6 (0.6)23.5 (4.6)20%27.6 (0.6)26.7 (1.5)23.2 (4.9)16.0 (12.2)11.2 (17.0)28.2 (0.0)28.2 (0.0)28.1 (0.1)27.5 (0.6)24.1 (4.0)C(H5N1-like)11.6%1%11.6 (0.0)11.6 (0.0)11.6 (0.0)11.6 (0.0)11.6 (0.0)11.3 (0.3)10.7 (0.9)9.8 (1.8)8.8 (2.8)8.5 (3.1)5%11.1 (0.5)10.2 (1.4)9.8 (1.8)9.7 (1.9)9.8 (1.8)11.7 (0.0)11.1 (0.5)9.2 (2.4)5.6 (6.0)2.6 (9.0)10%9.8 (1.8)8.1 (3.5)6.8 (4.8)6.2 (5.4)6.2 (5.4)10.9 (0.7)10.8 (0.8)9.8 (1.8)6.2 (5.4)3.2 (8.4)20%8.4 (3.2)6.6 (5.0)4.4 (7.2)3.5 (8.1)3.3 (8.3)11.6 (0.0)11.0 (0.6)9.6 (2.0)7.0 (4.6)3.8 (7.8)D(1918-like)29.8%1%29.8 (0.0)29.8 (0.0)29.8 (0.0)29.8 (0.0)29.8 (0.0)29.8 (0.0)29.8 (0.0)29.8 (0.0)29.3 (0.5)26.9 (2.9)5%28.8 (1.0)27.7 (2.1)25.9 (3.9)23.1 (6.7)21.9 (7.9)29.8 (0.0)29.8 (0.0)29.8 (0.0)29.4 (0.4)27.5 (2.3)10%29.0 (0.8)28.1 (1.7)26.3 (3.5)21.5 (8.3)17.9 (11.9)29.8 (0.0)29.8 (0.0)29.7 (0.1)29.5 (0.3)27.8 (2.0)20%29.3 (0.5)28.7 (1.1)26.9 (2.9)21.4 (8.4)15.8 (14.0)29.8 (0.0)29.8 (0.0)29.7 (0.1)29.6 (0.2)27.8 (2.0)Table S5: Cumulative attack rates (AR) through 180 days with the reduction from the baseline scenario in parentheses (?) using the regional model, with the shorter (mean ~2.8 day) serial interval and worst-case contact pattern change (a 30% reduction in non-household contacts and doubling of child-related household contacts) during school dismissal.Pandemic scenarioBaselineClinical Attack Rate TriggerCommunity School DismissalRegional School DismissalDiag. ratioDuration Duration 1 wk2 wks4 wks8 wks12 wks1 wk2 wks4 wks8 wks12 wksAR(δ)AR(δ)AR(δ)AR(δ)AR(δ) AR(δ) AR(δ) AR(δ) AR(δ) AR(δ)A(2009-like)17.3%1%17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)5%16.5 (0.9)15.9 (1.4)15.7 (1.6)15.6 (1.7)15.7 (1.7)16.8 (0.5)16.2 (1.1)13.3 (4.1)5.4 (12.0)1.5 (15.9)10%14.7 (2.7)13.2 (4.2)12.2 (5.1)11.8 (5.6)11.7 (5.6)16.7 (0.6)15.7 (1.6)13.0 (4.3)5.9 (11.4)1.5 (15.8)20%12.6 (4.8)9.8 (7.5)7.7 (9.7)6.5 (10.8)6.2 (11.1)16.7 (0.7)16.0 (1.4)13.6 (3.7)6.2 (11.1)2.2 (15.1)B1(1968-like)22.3%1%22.3 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)22.4 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)5%21.6 (0.8)21.1 (1.3)20.8 (1.5)20.6 (1.7)20.6 (1.7)22.2 (0.1)21.9 (0.4)21.1 (1.2)15.6 (6.7)5.5 (16.8)10%20.4 (2.0)18.8 (3.5)17.4 (5.0)16.2 (6.1)15.9 (6.4)22.2 (0.1)22.0 (0.3)21.2 (1.2)15.9 (6.5)5.5 (16.8)20%19.5 (2.8)17.4 (5.0)14.0 (8.4)11.5 (10.8)10.6 (11.7)22.2 (0.1)22.1 (0.2)21.5 (0.8)16.1 (6.2)7.1 (15.3)B2(1957-like)28.3%1%28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)5%26.9 (1.4)26.0 (2.4)25.1 (3.3)24.5 (3.8)24.4 (3.9)28.3 (0.0)28.3 (0.0)28.2 (0.1)27.0 (1.3)17.3 (11.1)10%26.5 (1.8)25.1 (3.2)22.6 (5.7)19.3 (9.0)18.3 (10.0)28.3 (0.0)28.3 (0.0)28.2 (0.1)27.1 (1.2)18.0 (10.3)20%27.0 (1.3)25.7 (2.6)22.6 (5.7)15.9 (12.4)13.2 (15.1)28.2 (0.1)28.3 (0.0)28.2 (0.1)27.4 (0.9)19.4 (8.9)C(H5N1-like)8.9%1%8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)5%8.7 (0.2)8.5 (0.5)8.3 (0.7)8.2 (0.8)8.3 (0.6)8.2 (0.7)6.8 (2.1)4.8 (4.1)1.8 (7.2)0.6 (8.4)10%7.6 (1.3)6.4 (2.6)6.2 (2.8)6.1 (2.9)6.1 (2.9)8.2 (0.7)7.5 (1.5)5.3 (3.7)2.1 (6.9)0.7 (8.3)20%5.9 (3.0)3.9 (5.0)3.3 (5.6)3.0 (5.9)2.9 (6.0)8.4 (0.5)8.2 (0.8)5.5 (3.4)2.3 (6.7)0.8 (8.2)D(1918-like)30.1%1%30.1 (0.0)30.1 (0.0)30.1 (0.0)30.1 (0.0)30.1 (0.0)29.9 (0.2)29.8 (0.3)29.7 (0.4)28.1 (2.0)25.7 (4.4)5%28.4 (1.7)27.2 (2.8)26.0 (4.1)24.8 (5.3)24.6 (5.5)30.1 (0.0)30.1 (0.0)30.1 (0.0)30.0 (0.1)27.3 (2.8)10%28.5 (1.5)27.4 (2.7)25.5 (4.6)20.7 (9.4)18.9 (11.2)30.1 (0.0)30.1 (0.0)30.1 (0.0)30.0 (0.1)27.9 (2.2)20%29.2 (0.9)28.5 (1.6)27.2 (2.9)20.3 (9.8)14.8 (15.3)30.1 (0.0)30.1 (0.0)30.1 (0.0)30.0 (0.1)28.5 (1.6)Table S6: Cumulative attack rates (AR) through 180 days with the reduction from the baseline scenario in parentheses (?) using the regional model, with the shorter (mean ~2.8 day) serial interval and best-case contact pattern change (a 50% reduction in non-household) during school dismissal.Pandemic scenarioBaselineClinical Attack Rate TriggerCommunity School DismissalRegional School DismissalDiag. ratioDuration Duration 1 wk2 wks4 wks8 wks12 wks1 wk2 wks4 wks8 wks12 wksAR(δ)AR(δ)AR(δ)AR(δ)AR(δ) AR(δ) AR(δ) AR(δ) AR(δ) AR(δ)A(2009-like)17.3%1%17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)17.3 (0.0)5%15.5 (1.8)15.1 (2.3)14.7 (2.6)14.6 (2.7)14.6 (2.8)16.0 (1.3)14.3 (3.0)9.4 (7.9)2.6 (14.7)0.6 (16.7)10%12.0 (5.3)10.1 (7.2)9.2 (8.1)8.8 (8.5)8.6 (8.7)16.3 (1.0)14.5 (2.8)9.6 (7.7)3.7 (13.6)1.0 (16.3)20%9.1 (8.3)5.1 (12.2)3.5 (13.8)3.2 (14.1)3.0 (14.3)15.9 (1.4)13.9 (3.4)11.2 (6.2)4.2 (13.1)1.1 (16.2)B1(1968-like)22.3%1%22.3 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)22.3 (0.0)5%20.7 (1.7)20.1 (2.2)19.6 (2.7)19.3 (3.0)19.1 (3.2)22.0 (0.3)21.2 (1.1)18.7 (3.6)6.4 (15.9)1.4 (20.9)10%18.3 (4.0)16.2 (6.1)14.3 (8.0)12.7 (9.6)12.3 (10.0)22.0 (0.3)21.5 (0.8)19.0 (3.3)8.3 (14.1)1.9 (20.4)20%16.6 (5.7)12.1 (10.2)8.6 (13.7)6.7 (15.7)6.1 (16.2)22.1 (0.3)21.6 (0.8)20.0 (2.4)9.9 (12.4)2.3 (20.0)B2(1957-like)28.3%1%28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)28.3 (0.0)5%25.5 (2.8)24.3 (4.0)23.1 (5.2)21.8 (6.6)21.4 (6.9)28.2 (0.1)28.2 (0.1)27.9 (0.4)22.5 (5.9)5.7 (22.6)10%25.0 (3.3)23.1 (5.3)19.6 (8.7)15.0 (13.3)13.5 (14.9)28.2 (0.1)28.3 (0.1)28.0 (0.3)23.1 (5.2)6.7 (21.6)20%25.9 (2.4)23.1 (5.2)17.1 (11.3)10.1 (18.2)7.8 (20.5)28.3 (0.0)28.2 (0.1)28.1 (0.2)24.5 (3.8)9.5 (18.8)C(H5N1-like)8.9%1%8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)8.9 (0.0)5%8.4 (0.5)7.9 (1.0)7.9 (1.1)8.0 (0.9)8.1 (0.9)7.3 (1.6)5.5 (3.5)2.7 (6.3)0.9 (8.0)0.3 (8.7)10%6.1 (2.8)4.9 (4.0)4.4 (4.5)4.4 (4.5)4.5 (4.4)7.7 (1.3)6.1 (2.9)3.4 (5.5)1.3 (7.6)0.4 (8.6)20%3.8 (5.1)2.1 (6.8)1.7 (7.2)1.6 (7.4)1.5 (7.5)8.0 (0.9)5.9 (3.0)4.2 (4.8)1.5 (7.4)0.5 (8.4)D(1918-like)30.1%1%30.1 (0.0)30.1 (0.0)30.1 (0.0)30.1 (0.0)30.1 (0.0)29.7 (0.4)29.5 (0.6)28.9 (1.2)26.0 (4.1)21.8 (8.3)5%26.7 (3.3)25.4 (4.7)23.7 (6.4)21.2 (8.9)20.5 (9.6)30.1 (0.0)30.1 (0.0)30.1 (0.0)29.3 (0.8)14.4 (15.7)10%27.3 (2.8)26.1 (4.0)23.9 (6.1)16.5 (13.6)13.3 (16.7)30.1 (0.0)30.1 (0.0)30.1 (0.0)29.5 (0.6)17.7 (12.4)20%28.5 (1.6)27.6 (2.5)25.1 (4.9)14.9 (15.2)8.9 (21.2)30.1 (0.0)30.1 (0.0)30.1 (0.0)29.7 (0.4)20.3 (9.8)Table S7: Cumulative attack rates (AR) through 180 days with the reduction from the baseline scenario in parentheses (?) using the regional model, with the longer (mean ~4.0 day) serial interval and worst-case contact pattern change (a 30% reduction in non-household contacts and doubling of child-related household contacts) during school dismissal.Pandemic scenarioBaselineClinical Attack Rate TriggerCommunity School DismissalRegional School DismissalDiag. ratioDuration Duration 1 wk2 wks4 wks8 wks12 wks1 wk2 wks4 wks8 wks12 wksAR(δ)AR(δ)AR(δ)AR(δ)AR(δ) AR(δ) AR(δ) AR(δ) AR(δ) AR(δ)A(2009-like)16.3%1%16.3 (0.0)16.3 (0.0)16.3 (0.0)16.2 (0.1)16.2 (0.1)15.8 (0.5)14.5 (1.8)11.9 (4.4)7.7 (8.6)6.5 (9.8)5%15.1 (1.2)14.3 (2.0)13.5 (2.8)13.3 (3.0)13.2 (3.1)16.1 (0.2)15.1 (1.2)12.8 (3.5)7.2 (9.1)3.3 (13.0)10%14.0 (2.3)11.9 (4.4)10.1 (6.2)9.4 (6.9)9.2 (7.1)15.9 (0.4)15.0 (1.3)13.1 (3.2)7.8 (8.5)3.6 (12.7)20%12.9 (3.4)9.5 (6.8)6.9 (9.4)5.8 (10.5)5.4 (10.9)16.0 (0.3)15.1 (1.2)13.3 (3.0)8.2 (8.1)3.9 (12.4)B1(1968-like)20.8%1%20.8 (0.0)20.8 (0.0)20.9 (0.0)20.9 (0.0)20.9 (0.0)20.6 (0.3)19.6 (1.2)17.2 (3.6)9.5 (11.3)5.3 (15.6)5%19.5 (1.3)18.1 (2.7)16.9 (3.9)16.5 (4.3)16.3 (4.6)20.6 (0.2)19.8 (1.1)18.0 (2.8)11.5 (9.3)4.9 (16.0)10%18.5 (2.3)16.3 (4.5)13.5 (7.3)11.8 (9.0)11.4 (9.4)20.6 (0.2)19.7 (1.1)17.9 (2.9)12.2 (8.6)5.6 (15.3)20%17.9 (2.9)14.4 (6.4)10.0 (10.8)7.4 (13.4)6.8 (14.0)20.6 (0.2)20.0 (0.8)18.5 (2.3)12.3 (8.5)5.6 (15.3)B2(1957-like)28.1%1%28.1 (0.0)28.0 (0.1)28.0 (0.1)28.0 (0.1)28.0 (0.1)28.1 (0.0)28.1 (0.1)27.8 (0.3)23.9 (4.2)9.5 (18.6)5%26.8 (1.4)25.1 (3.0)22.5 (5.6)19.2 (8.9)18.3 (9.8)28.1 (0.0)28.1 (0.0)27.8 (0.3)24.9 (3.2)11.6 (16.5)10%27.0 (1.1)25.3 (2.8)21.2 (6.9)15.0 (13.2)13.0 (15.1)28.1 (0.0)28.1 (0.0)27.8 (0.3)25.2 (2.9)12.4 (15.7)20%27.2 (0.9)25.1 (3.0)18.0 (10.1)10.4 (17.7)8.3 (19.9)28.1 (0.0)28.1 (0.0)27.8 (0.3)25.2 (2.9)13.8 (14.3)C(H5N1-like)7.6%1%7.6 (0.0)7.6 (0.0)7.6 (0.0)7.6 (0.0)7.6 (0.0)7.4 (0.2)7.0 (0.6)6.5 (1.0)6.2 (1.4)6.1 (1.4)5%7.2 (0.3)6.4 (1.1)6.2 (1.4)6.1 (1.4)6.2 (1.4)7.6 (0.0)6.9 (0.6)4.9 (2.7)2.1 (5.4)0.8 (6.8)10%6.0 (1.6)4.7 (2.9)3.8 (3.7)3.5 (4.0)3.6 (4.0)6.8 (0.8)6.5 (1.1)5.3 (2.2)2.5 (5.1)1.0 (6.5)20%4.7 (2.8)3.6 (4.0)2.3 (5.2)1.9 (5.6)1.8 (5.7)7.6 (0.0)6.8 (0.8)5.1 (2.4)3.0 (4.6)1.2 (6.3)D(1918-like)29.7%1%29.8 (0.0)29.7 (0.0)29.7 (0.0)29.7 (0.0)29.7 (0.0)29.7 (0.0)29.7 (0.0)29.5 (0.2)27.7 (2.1)18.2 (11.6)5%28.6 (1.1)27.4 (2.4)25.1 (4.6)22.4 (7.4)21.6 (8.2)29.8 (0.0)29.7 (0.0)29.6 (0.1)28.0 (1.7)19.7 (10.0)10%28.8 (0.9)27.6 (2.2)24.6 (5.1)18.9 (10.9)16.7 (13.0)29.7 (0.0)29.7 (0.0)29.5 (0.2)28.1 (1.6)21.0 (8.7)20%29.1 (0.7)27.9 (1.8)24.1 (5.6)16.2 (13.6)12.8 (17.0)29.7 (0.0)29.7 (0.0)29.5 (0.2)28.1 (1.6)21.1 (8.6)Table S8: Cumulative attack rates (AR) through 180 days with the reduction from the baseline scenario in parentheses (?) using the regional model, with the longer (mean ~4.0 day) serial interval and best-case contact pattern change (a 50% reduction in non-household) during school dismissal.Pandemic scenarioBaselineClinical Attack Rate TriggerCommunity School DismissalRegional School DismissalDiag. ratioDuration Duration 1 wk2 wks4 wks8 wks12 wks1 wk2 wks4 wks8 wks12 wksAR(δ)AR(δ)AR(δ)AR(δ)AR(δ) AR(δ) AR(δ) AR(δ) AR(δ) AR(δ)A(2009-like)16.3%1%16.2 (0.1)16.3 (0.0)16.3 (0.0)16.2 (0.1)16.2 (0.1)13.8 (2.5)11.3 (5.0)6.9 (9.4)3.8 (12.5)3.5 (12.8)5%13.2 (3.1)11.7 (4.6)10.6 (5.7)10.3 (6.0)10.2 (6.1)15.2 (1.1)13.3 (3.0)9.1 (7.2)3.0 (13.3)0.8 (15.5)10%10.4 (5.9)7.9 (8.4)5.9 (10.4)5.4 (10.9)5.2 (11.1)14.9 (1.4)13.8 (2.5)10.0 (6.3)3.5 (12.8)0.9 (15.4)20%9.0 (7.3)5.2 (11.1)2.8 (13.5)2.2 (14.1)2.1 (14.2)15.0 (1.3)13.7 (2.6)10.3 (6.0)4.0 (12.3)1.2 (15.1)B1(1968-like)20.8%1%20.9 (0.0)20.9 (0.0)20.9 (0.0)20.8 (0.0)20.8 (0.0)19.3 (1.5)17.3 (3.5)11.4 (9.4)3.1 (17.7)1.8 (19.0)5%17.5 (3.3)15.4 (5.5)13.7 (7.1)13.0 (7.8)12.9 (8.0)20.2 (0.6)18.5 (2.3)13.7 (7.1)4.7 (16.1)1.1 (19.8)10%15.3 (5.5)11.1 (9.7)8.2 (12.6)7.1 (13.7)6.8 (14.0)20.1 (0.7)18.7 (2.1)15.6 (5.2)6.1 (14.7)1.4 (19.4)20%13.5 (7.3)8.2 (12.6)4.3 (16.5)3.2 (17.7)2.9 (17.9)20.1 (0.8)19.2 (1.7)15.0 (5.8)6.7 (14.2)1.7 (19.1)B2(1957-like)28.1%1%27.9 (0.2)27.8 (0.3)27.8 (0.3)27.8 (0.4)27.7 (0.4)28.1 (0.1)27.9 (0.2)26.9 (1.2)14.2 (14.0)2.3 (25.8)5%25.1 (3.1)22.4 (5.7)18.3 (9.8)14.2 (13.9)12.9 (15.2)28.1 (0.1)28.0 (0.1)27.3 (0.8)19.0 (9.1)5.1 (23.0)10%25.5 (2.6)21.5 (6.6)14.3 (13.8)9.0 (19.1)7.4 (20.7)28.1 (0.1)28.0 (0.2)27.3 (0.8)19.9 (8.2)5.9 (22.2)20%25.8 (2.3)19.8 (8.3)9.4 (18.7)4.7 (23.4)3.5 (24.6)28.1 (0.0)28.0 (0.1)27.5 (0.6)21.0 (7.1)6.2 (21.9)C(H5N1-like)7.6%1%7.6 (0.0)7.6 (0.0)7.6 (0.0)7.6 (0.0)7.6 (0.0)6.7 (0.8)6.2 (1.3)5.8 (1.8)5.6 (1.9)5.6 (1.9)5%5.9 (1.7)5.2 (2.4)4.6 (2.9)4.6 (3.0)4.5 (3.0)6.2 (1.3)5.2 (2.4)2.8 (4.8)0.9 (6.6)0.3 (7.2)10%4.1 (3.4)2.7 (4.8)2.1 (5.5)2.1 (5.5)2.0 (5.5)7.0 (0.6)5.2 (2.4)3.6 (4.0)1.3 (6.2)0.4 (7.2)20%2.9 (4.6)1.6 (5.9)0.9 (6.6)0.7 (6.8)0.7 (6.8)6.6 (0.9)5.8 (1.8)4.1 (3.5)1.5 (6.0)0.5 (7.1)D(1918-like)29.7%1%29.7 (0.1)29.6 (0.1)29.6 (0.1)29.6 (0.1)29.6 (0.2)29.7 (0.0)29.6 (0.1)29.0 (0.8)17.7 (12.0)3.0 (26.7)5%27.1 (2.7)25.0 (4.7)21.6 (8.1)17.0 (12.7)15.7 (14.0)29.7 (0.0)29.6 (0.1)29.1 (0.6)22.1 (7.7)6.3 (23.5)10%27.5 (2.2)25.0 (4.8)19.3 (10.4)12.2 (17.5)9.8 (19.9)29.7 (0.0)29.6 (0.1)29.2 (0.6)23.6 (6.1)8.4 (21.3)20%28.0 (1.7)24.6 (5.2)15.7 (14.1)7.9 (21.8)5.7 (24.0)29.7 (0.0)29.7 (0.1)29.3 (0.5)24.5 (5.2)9.3 (20.4) ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download