Nonpharmaceutical Measures for Pandemic Influenza in Nonhealthcare Settings—Social Distancing Measures

Influenza virus infections are believed to spread mostly by close contact in the community. Social distancing measures are essential components of the public health response to influenza pandemics. The objective of these mitigation measures is to reduce transmission, thereby delaying the epidemic peak, reducing the size of the epidemic peak, and spreading cases over a longer time to relieve pressure on the healthcare system. We conducted systematic reviews of the evidence base for effectiveness of multiple mitigation measures: isolating ill persons, contact tracing, quarantining exposed persons, school closures, workplace measures/closures, and avoiding crowding. Evidence supporting the effectiveness of these measures was obtained largely from observational studies and simulation studies. Voluntary isolation at home might be a more feasible social distancing measure, and pandemic plans should consider how to facilitate this measure. More drastic social distancing measures might be reserved for severe pandemics.


Findings
The initial database search yielded 588 articles, of which 70 were selected for full-text screening based on their title and abstract contents. Of these, 56 articles were excluded; main reasons for exclusion of relevant articles include absence of discussion on effectiveness of isolation and focus on healthcare setting. One other study for inclusion was identified through snowball searches. The study selection process is detailed in Appendix  Table 3) (3)(4)(5)(6). The remaining 11 are simulation studies (Appendix Table 4 (7)(8)(9)(10)(11)(12)(13)(14)(15)(16). Isolation was implemented in the outbreaks as a combination with various other interventions such as antiviral prophylaxis and use of a face mask. Isolation was also studied as a single intervention or combined with other interventions in the 11 simulation studies. It is of note that the simulation studies were conducted based on a wide range of assumptions, for example asymptomatic fraction and contact rate reduction brought forth by isolation, hence providing wide-ranging insights on effectiveness of isolation in different scenarios. These included studies focused mostly on reduction of attack rate, epidemic size, transmissibility, and delay in epidemic peak as outcomes-of-interest. All but one study suggested favorable impact of isolation, or combination of isolation with other interventions.

Reduction of Impact
Eight studies suggested decrease in attack rate (AR) brought about by implementation of case isolation (3,(6)(7)(8)(10)(11)(12)14). An individual-based simulation model for Great Britain and the United States suggested rapid isolation could reduce the cumulative clinical attack rate from 34% to 27% for a pandemic with R0 2.0, assuming uniform reductions in contact rates in schools, workplaces and households (7). Kelso et al. reported similar findings, in which case isolation alone is able to prevent an epidemic (<10% infected) in a 30,000 persons community with R0 1.5, when 90% of cases are isolated and such measure is implemented within 3 weeks from the introduction of an initial case (11). Although isolation alone has been suggested to be more impactful than other interventions, combination with other interventions further improved the effectiveness (10)(11)(12)14). In addition, increase in isolation rate is quasi-linearly correlated with decrease in attack rate of influenza (8).
A reduction in the cumulative incidence of infections due to an isolation policy was also recorded during an influenza A(H1N1)pdm09 outbreak on a navy ship (6). A combination of isolating cases of influenza-like illness (ILI), use of masks and hand sanitizers was implemented.
The clinical attack rate in the outbreak was 23.9%, a significant reduction from the 97% projected in the absence of any intervention. This also corresponded to a reduction in the effective reproduction number (R) from 1.55 to 0.7 with the intervention. Chu et al. reported similar findings in an outbreak in a physical training camp, in which the final AR recorded was ≈25% of the projected AR of 81% in absence of previous exposure, immunity, and any interventions. In the 1918-19 pandemic, excess death rates due to pneumonia and influenza decreased in New York City and Denver after isolation and quarantine were implemented (5).
Conversely, Fraser et al. discussed the difficulty in controlling influenza even with high level of case isolation combined with contact tracing and quarantine, due to the high proportion of asymptomatic transmission of influenza (9). The probability of self-isolation without increased public health effort by persons in the community have also been suggested to be high, at 50% and 90% for adult and children respectively (11).

Delay of Epidemic Peak
The study of Flauhault et al. suggested that case isolation would have the strongest impact on global spread of a pandemic involving 52 cities compared with air travel restrictions and antiviral treatment, such that isolation of 40% of cases would delay the epidemic by 83 days compared with absence of any intervention (8). A combination of isolation of 10% of symptomatic cases with 60% reduction in air traffic on the other hand would delay the start of epidemics in each city by an average of 19 days with considerable case reduction (8). The study of Wang et al. study showed similar effect albeit focusing on arrival time of influenza pandemic, in which isolation of a moderate proportion of cases delayed the arrival of the pandemic in a subpopulation for about a month, in the circumstance of high compliance and early implementation (13). Delay in response will reduce the effectiveness. Combined intervention with quarantine, school closure, community contact reduction, and personal protective measures further augmented the effect (12).

Reduction in Transmissibility
Zhang et al. showed in their simulation studies that isolation of cases can reduce household reproduction number to below one, and compensate delay in antiviral drug distribution by 1 to 2 days. Compliance for isolation has to be much higher to offset longer delays (15,16). An outbreak in an elderly home in France reported an abrupt cessation of outbreak after case isolation, antiviral treatment and prophylaxis were implemented (4). Reduction in reproduction number was also recorded in the navy ship outbreak previously described, by 54% from 1.55 to 0.7 with a combination of interventions (6). The projected reproduction number without isolation of cases was 4.5.
Appendix Figure 1. Flowchart of literature search and study selection for isolation. Isolation of symptomatic persons contact-tracing and quarantine of some persons who were infected before symptomatic persons were isolated; Interventions were implemented without delay. Efficacy of isolation considered were 75%, 90%, and 100%; contact tracing and isolation were assumed to be fully effective.

Appendix
Not available Control of influenza is challenging even at high level (90%) of quarantine and contact tracing, due to the considerable proportion of pre-symptomatic transmission.
Halloran ME, 2008 Contact tracing is the identification and follow-up of persons who may have come into contact with an infected person (18). Although contact tracing is often coupled with quarantine or provision of antiviral prophylaxis to exposed contacts, the term contact tracing does not involve these processes.  Table 5).

Findings
The initial database search yielded 1188 articles, of which 75 were selected for full-text screening based on their title and abstract contents. Of these, 71 articles were excluded; the main reasons for exclusion of these articles include absence of discussion on effectiveness of contact tracing and irrelevance. The study selection process is detailed in Appendix Figure 2.
All 4 studies were simulation studies (9,14,19,20). None studied contact tracing as a single intervention; instead, this measure was studied in combination with other interventions, such as quarantine, and isolation and provision of antiviral drugs (Appendix Table 6). Such combinations of interventions have been suggested to reduce transmission and delay the epidemic peak (9,14,20).

Reduction of Impact
Wu et al. estimated in their simulation model of an influenza pandemic with a reproductive number (R0) of 1.8 that the combination of contact tracing, quarantine, isolation and antivirals can reduce the infection attack rate from the baseline of 74% to 34% (14). However, the addition of contact tracing on top of quarantine and isolation measures was suggested to provide only modest benefit, while at the same time greatly increasing the proportion of quarantined persons. Conversely, Fraser et al. suggested that it would be difficult to control influenza even with 90% contact tracing and quarantine, due to the high level of presymptomatic or asymptomatic transmission in influenza (9).

Delay of Epidemic Peak
In an epidemic of R0 1.58 in the population structure of Germany, a combination of isolation, treatment of cases, contact tracing, quarantine and postexposure prophylaxis for both community and household contacts, in addition to some household-focused measures, have been estimated to delay the epidemic peak for up to 6 weeks, assuming a case detection rate of 10%-30% (20). The authors assumed that the above combination of measures would be 75% effective in reducing secondary cases, and household-focused measures would be 50% effective.

Reduction in Transmissibility
Peak et al. compared the combination of contact tracing with quarantine or symptom monitoring in the early phase of an epidemic with an R0 of 1.54 (19). The study suggested that contact tracing combined with quarantine was more effective than a combination with symptom monitoring in reducing transmission. 1. 80 (1) Model based on population structure of Hong Kong (i.e., household sizes and average number of children in households) (2); 1.5 infected persons introduced each day per 100,000 persons for a year (3); 70% of transmission occur outside household (e.g., in schools and workplaces)

Appendix
Combination of contact tracing with other interventions such as quarantine, isolation and antivirals. For contact tracing, persons were asked to name on average five members of their peer group. The contacts of all new symptomatic or hospitalized cases were traced with a mean delay of 1 d. Contacts were asked to take precautionary measures. Interventions were active before arrival of infected persons in the city

No intervention
Attack rate decreased from baseline of 74% to 40% when combination of isolation, quarantine and antivirals is implemented. Addition of contact tracing to the combination of interventions further reduced attack rate to 34%, but increased considerably the proportion of population in quarantine Peak CM, 2017 (19) 1.54 (1) Initial infected population of 1000 persons during the early phase of an epidemic (2); no substantial depletion of susceptibles within first few generations of transmission Symptomatic contacts were isolated immediately, asymptomatic contacts were placed under quarantine (in a high performance scenario, delay in contact tracing was 0.5 ± 0.5 d, 90% of contacts were traced, 50% of traced contacts were infected) Asymptomatic contacts were placed under symptom monitoring instead of quarantine Combination of contact tracing with quarantine is more effective in reducing reproduction number compared with combination of contact tracing with symptom monitoring Upper bound of R 0 was 21 (1) Early stage of disease outbreak in a community with homogenous mixing (2) Proportion of pre-symptomatic transmission is 30%-50% Isolation of symptomatic persons, contact-tracing and quarantine of some persons who were infected before symptomatic persons isolated; Interventions were implemented without delay. Efficacy of isolation considered were 75%, 90%, and 100%; contact tracing and isolation were assumed to be fully effective.
Not available Control of influenza is challenging even at high level (90%) of quarantine and contact tracing, due to the considerable proportion of pre-symptomatic transmission.
an der Heiden M, 2009 (20) 1.34, 1.58, 2.04 (1) Model based on the population structure of Germany: 71,000,000 adult and 11,000,000 children (<15 y old), whole population is completely susceptible at the beginning of the epidemic (2); Children are 2.06 times more susceptible than adults, 86% of infected persons show development of symptoms (1) Intensive case-based measures (CCM1; consisting of isolation and treatment of cases, contact tracing, quarantine and post-exposure prophylaxis of some household and community contacts) (2); Less-intensive measures (CCM2; isolation and treatment of cases, quarantine and post-exposure prophylaxis of only household contacts); CCM1 and CCM2 were assumed to be 75% and 50% respectively in their effectiveness to reduce secondary cases

No intervention
(1) When the initial 500 cases were subjected to CCM1 and the subsequent 10,000 cases CCM2, the peak of the epidemic is delayed for up to 6 weeks (R 0 1.58, 5 imported cases per day, case detection rate 10%-30%). If only CCM1 was adopted, the delay was estimated to be 6-20 d (case detection rate 10%-30%) (2); Effectiveness of these combination of interventions is affected by the R 0 of the influenza strain and case detection rate, i.e., higher R 0 causes interventions to be ineffective at an earlier time point.

Quarantine of Exposed Persons
Terminology Terms relevant to quarantine are defined below (Appendix Table 7): Appendix Table 7. Definition of terms relevant to quarantine Term Definition Quarantine Imposed "separation or restriction of movement" of persons who are "exposed, who may or may not be infected but are not ill," and "may become infectious to others" (1).
Household quarantine Confinement (commonly at home) of non-ill household contacts of a person with proven or suspected influenza (1,2).
Home quarantine Home confinement of non-ill contacts of a person with proven or suspected influenza.

Self-quarantine
Voluntary confinement of non-ill contacts of a person with proven or suspected influenza.
Work quarantine 1) Measures taken by workers "who have been exposed and who work in a setting where the disease is especially liable to transmit (or where there are people at higher risk from infection), e.g. people working in elderly homes and nurses in high risk units" (1). 2) Measures taken by healthcare workers who "chose to stay away from their families when off-duty so as not to carry the infection home" (1).

Maritime quarantine
Monitoring of all passengers and crew for a defined period before disembarking from a ship is permitted in a jurisdiction (21).
Onboard quarantine Monitoring of all passengers and crew for a defined period before disembarking from a flight is permitted (22). Also known as 'airport quarantine' (22).

Search Strategy
A literature search was conducted by using PubMed, MEDLINE, EMBASE, and CENTRAL to identify literature that were available from 1946 through July 23, 2018. Similar to isolation, no limitation on language and study design were applied for the literature search.
Literatures in languages other than English were excluded during full-text screening. Studies reporting the effectiveness of quarantine on control of influenza in nonhealthcare settings were included. Systematic reviews and metaanalyses, as well as studies involving clinical settings were excluded. Two reviewers (M.W.F. and H.G.) independently screened the titles, abstracts and fulltexts to identify articles for inclusion (Appendix Table 8).
Quarantine measures studied include home quarantine, household quarantine, border quarantine as well as maritime quarantine. Quarantine was studied as a single intervention or as a combination with other interventions, commonly with isolation and antiviral prophylaxis. These included studies focused mostly on reduction of attack rate, transmissibility, and delay in epidemic peak as outcomes-of-interest.
Appendix Figure 3. Flowchart of literature search and study selection for quarantine.
Five studies suggested reduction in attack rate with implementation of household quarantine measures (7,10,12,14,29 (7). Combination of quarantine with other interventions such as home isolation, provision of antiviral prophylaxis, school closure and workplace distancing were suggested to further reduce the cumulative incidence of infections (7,10,14).
Household quarantine has also been suggested to be highly effective in reducing peak and total number of cases in a pandemic, provided that compliance is high (27). Longini et al.
reported similar findings, that is the effectiveness of household quarantine in reducing number of cases is conditioned by high compliance at 70% and relatively low R0, in addition to early implementation (23). Border quarantine on the other hand has been suggested to cause minimal impact on reduction of number of cases (26).
Both analyses of historical data of the 1918-19 pandemic studied the effectiveness of interventions on mortality rates (5,21). When a combination of isolation and quarantine was implemented, excess death rates due to pneumonia and influenza decreased in New York City and Denver (5). Maritime quarantine in the pacific islands have also delayed or prevented arrival of the epidemic, indirectly reducing mortality rates in the jurisdictions (21). would prevent 99% of entry of infectious travelers into small island nations (24).

Increased Risk for Household Contacts
Although it showed a reduction of the infection rate in the intervention cluster, the intervention study of Miyaki et al. also reported that more persons became ill in the intervention group when there was an ill family member (29). The likelihood of a household contact (concurrently quarantined with an isolated individual) becoming a secondary case has been estimated to increase with each day of quarantine (30).

Page 16 of 57
Appendix consisting of isolation and treatment of cases, contact tracing, quarantine and postexposure prophylaxis of some household and community contacts) (2); Less-intensive measures (CCM2; isolation and treatment of cases, quarantine and post-exposure prophylaxis of only household contacts); CCM1 and CCM2 were assumed to be 75% and 50% respectively in their effectiveness to reduce secondary cases No intervention (1) When the initial 500 cases were subjected to CCM1 and the subsequent 10,000 cases CCM2, the peak of the epidemic is delayed for up to 6 weeks (R 0 1.58, 5 imported cases per day, case detection rate 10%-30%). If only CCM1 was adopted, the delay was estimated to be 6-20 d (case detection rate 10%-30%) (2); Effectiveness of these combination of interventions is affected by the R 0 of the influenza strain and case detection rate, i.e., higher R 0 causes interventions to be ineffective at an earlier time point. Saundershastings P, 2017 (12) 1.5-2. 5 (1) Model based on the population structure of Ottawa-Gatineau census metropolitan area in 2011 Combination of quarantine with other interventions including vaccination, antiviral treatment and prophylaxis, school closure, reduction in community contact, personal protective measures, and isolation; best estimate for compliance for quarantine is 15% (1) Home quarantine (70% compliance) for 6 d, which prevents 56% of all transmission from those infected within their household.
(2) Home quarantine (50% compliance), which prevents 40% of transmission from household contacts (3)  (1) Household quarantine (home confinement at all times with compliance 25%, 50%, 75%, and 100%). (2) Combination of household quarantine with school closure and avoiding social activities; Delay between interventions and outbreak threshold was less than one day No intervention At 50% compliance, household quarantine reduced 12.5% and 20.8% of total number of cases and peak cases respectively, as well as delayed epidemic peak. A combination of all 3 interventions did not add much benefit in reducing the total number of cases, however reduced the peak cases by 56%, and delayed the epdemic peak School is closed after a substantial incidence of ILI-related illnesses is reported among children and/or staffs in that school. Pre-emptive Closure/ Dismissal School is closed before a substantial transmission among children and staff is reported.

Search Strategy
The latest systematic review to review the effects of school closures on influenza  Table 12).
Appendix Table 12. Search strategy for school closures Search terms Search date Reviewers #1: "school closure" OR "class dismissal" OR "school holiday" OR "community mitigation" OR "social distancing" #2: "influenza" OR "flu" #3: #1 AND #2  Sixteen studies demonstrated that reactive school closure could be a useful control measure during influenza epidemics or pandemics, with impacts that included reducing the incidence and reducing the peak size (Appendix Table 14). Several studies reported a reduction in number of confirmed or ILI cases (36,37,39,41,45,47,48). One study also showed a reduction in total infected cases by 32.7% (total reduced number of cases from 127.1 to 85.5) (44). Another observational study suggested a reduction in the peak of the epidemic curve by 24% during the 4day closure and also a reduction of the total number of infected students by 8% (40). However, 2 observational studies in China did not identify a significant difference for total attack rate between the control (school closure not implemented) and intervention group (school closed) (34,35). Two studies in the United States showed that absenteeism was lower after school reopening compared with before school closure (42,43).  (38). An observational study from Japan reported that school closure was more effective than class closure (dismissal of that particular class with substantial increase in influenza incidence) (48). In another study from Japan, a 2-day school closure in the outbreak situation (after a 10% of absentee occurrence in a school) was associated with the interruption of an outbreak within a week (46). One detailed study of transmission in a school in Pennsylvania identified no effect of the reactive closure that was implemented when 27% of students already had symptoms (33).
Effectiveness of preemptive school closure was studied in 13 articles (Appendix Table   15). A study showed that preemptive school closure had an advantage to delay the epidemic peak for more than a week, affect the modeled mean peak, and reduce overall attack rate from 9.7% to 8.6% (49). Bootsma  One study estimated a 29%-37% reduction in influenza transmission by the 18-day period of mandatory school closures and other social distancing measures including closure of restaurants and theaters, and cancellation events (52). A study in Mexico City estimated that effective reproduction ratio declined from 1.6 before closure to less than 1 during closure (55).
Wu et al. estimated that the reproduction number was reduced from 1.7 to 1.5 during the pre-emptive closures and to 1.1 during the rest of the summer holiday (60). One study in Mexico showed a 80% reduction of contact rate during closure period and a subsequent planned holiday (58). However, closing kindergartens and primary schools for 2 weeks in Hong Kong did not show any significant effect on community transmission, although the incidence remained low after the peak during preemptive closure (54).
Twenty-eight studies monitored the change of influenza incidence across planned school holidays, for example the scheduled winter holiday each year, to estimate the impact of school closure on influenza transmission (Appendix Table 16). Of these studies, 8 showed that planned holidays could reduce influenza transmission (58,61,63,69,70,72,81,85). One study demonstrated that school holidays reduced the reproductive number R0 of influenza A(H1N1)pdm09 by 14%-27% in different regions of India compared with a nonholiday period (61). One study also reported an association of school holiday with a reduction of 63% to 100% in transmission in Canada (70). Another study reported a reduction of R0 from 1.25 to 0.79 during the 8 daysnational holidays in China, but reported that the 8-week summer school holiday had a limited effect on incidence of ILI (85). Two studies in the United Kingdom and Mexico showed that school closures could reduce contact rate by around 48%-80% (58,63). Two studies in Belgium and the Netherlands suggested that holidays delayed the epidemic peak by >1 week and reduced the peak incidence by 4%-27% (77,82). A study from the United States showed that absenteeism in Adrian reduced by ≈6% (79), whereas Rodriguez et al. reported no difference between closed schools and those did not close (80).
Observational studies also reported a reduction in incidence of influenza associated with planned school holidays (45,47,62,(64)(65)(66)(67)(68)71,72,(74)(75)(76)78,81,83,84). Studies showed that summer or winter holidays were associated with the reduction of ILI incidences by showing significant changes of ILI incidence rate ratios of school children to adults during the breaks (65,67,75 Figure 1); at junior high school, school closure significantly reduced the number of H1N1 case but not in class closure ( Figure 2) _ _ _ ILI: fever plus cough and/or sore throat *ILI rate ratio is compared at school district with 51%-100% school being closed vs. district with 1%-50% of school being closed. ^Author mentioned the recommended period of school closure is >4 d ^^Closure duration is significantly related with the number of cases within the 7-d of school opening Appendix The reproduction number was reduced from 1.7 to 1.5 during the pre-emptive closures and to 1.1 during the rest of the summer holiday ARI: Presence of at least 2 of the following symptoms: fever, cough, sore throat, or runny nose ILI: fever plus cough and/or sore throat #School closure combined with other interventions ^Pre-emptive closure followed by planned holidays *Assuming schools were closed for 4 weeks and the attack rate in children was 3-fold higher than in adult Page 27 of 57   Workplace measures include teleworking, flexible leave policies, working from home, weekend extension, staggered work shifts, and social distancing at workplaces. All randomized controlled trial, epidemiologic study or simulation study in nonhealthcare workplaces were included in this review. Reviews, commentaries, editorial articles, studies on workplace closure, and studies on generic social distancing irrelevant to workplace were excluded from our review. The following outcomes were extracted from the studies: cumulative attack rate, peak attract rate, occurrence of peak, and others. Two reviewers (H.G. and J.X.) worked independently (Appendix Table 18). Two reviewers (H.G. and E.S.) independently screened titles, abstracts and full texts to identify eligible articles (Appendix Table 19).
Appendix Table 19. Search strategy for workplace closures Search terms Search date Reviewers #1: "workplace" OR "work site" OR "business" OR "organization" OR "office" #2: "closure" OR "close" #3: "influenza" OR "flu" #4: #1 AND #2 AND #3 There were 6 epidemiologic studies among the 18 included studies (29,(135)(136)(137)(138)(139). A cross-sectional study interviewed randomly selected US adults from the Knowledge Networks online research panel, and showed that persons who cannot work from home (for 7-10 days) were more likely to have ILI symptoms compared with those who could (135). Another cohort study suggested that respondents who could work from home had a 30% lower rate of attending work with severe ILI symptoms compared with employees who cannot, suggesting work from home may be able to reduce employee-to-employee transmission (137). A cohort study in Singapore estimated that enhanced surveillance and segregation of work units into smaller working subgroups had significantly lower serologically confirmed infections compared with subgroups using the standard pandemic plan (17% vs 44%) (136). were simulation studies reviewed by Ahmed et al. (134),and suggested that workplace measure alone reduced the cumulative attack rate by 23%, as well as delaying and reducing the peak influenza attack rate (10,11,(140)(141)(142)(143)(144)(145)(146)(147)(148).  Among these 10 studies, 8 of them studied combination of workplace closure with school closure, 1 targeted different single and multiple intervention strategies, and 1 evaluated the effectiveness of workplace closure alone (Appendix Table 21). All 10 studies were simulation studies and the main outcomes include the reduction of attack rate, peak number, and delay of epidemic peak.   (151). In addition, a study in Italy suggested that combining strategies including vaccination, prophylaxis and closure of schools, workplaces and public places could reduce the incidence from 50% to ≈15% (153).
However, a heuristic model using R0 of 1.7 and 2.0 suggested a small reduction in cumulative attack rate but a more substantial reduction in peak attack rates (<40%) when 100% SC and 10% WC was implemented. It also suggested that the effectiveness could increase if 50% of workplaces were closed, at the same time resulting in a higher economic cost (7) Avoiding crowding refers to the measures to reduce influenza transmission in crowded areas (e.g., large meetings, conferences, and religious pilgrimages, national and international events).

Search Strategy
Literature available from 1946 through October 17, 2018 were identified from PubMed, Medline, EMBASE, and CENTRAL. Two reviewers (S.G. and E.S.) screened each title, abstract and article that fully met the criteria (Appendix Table 22). Both epidemiologic and simulation studies relevant to the effectiveness of avoiding crowding (e.g., cancellation or postponement of events and limitation of attendance) in public area are included. Studies that only reported outbreak events in a crowded area or perceptions on mass gathering without specific data related to the effectiveness of avoiding crowding; and reviews, letters, news, or summary articles were excluded.

Findings
We identified 3 studies for the systematic review after reviewing 815 titles and 121 abstracts identified from the 4 databases and other sources. Appendix Figure 9 shows the study selection process. Among these 3 articles, 2 were based on the 1918 influenza pandemic, and 1 focused on an influenza outbreak during the World Youth Day gathering in 2008 (details shown in Appendix Table 23). and lower total mortality rates (Spearman ρ = 0.37, p = 0.008) (5). There was a significant association between increased duration of interventions and a reduction in the total mortality rate (Spearman ρ = −0.39, p = 0.005) (5). Another study by Hatchett et al. also focused on the early bans on public gathering and closure of public places in reducing the excess death rate (57). In addition, during the 1-week long World Youth Day event in 2008, the group of youths who were accommodated in a single large place (17.2%) had a significantly higher attack rate compared with youths who lived in small classrooms (9.2%) (p<0.01) (157). Appendix

World Youth Day 2008 pilgrims
Pilgrims was sub-divided into smaller groups and accommodated in classrooms for 1 week.

Pilgrims was accommodated as a large group in a gymnasium
The attack rate was significantly (p<0.01) higher among pilgrims accommodated in the gymnasium (17.2%) than those staying in the classrooms (9.2%)