Despite variable recommendations for the containment of pH1N1 in Victoria (Australia), our findings suggest that compliance with both behavioural and pharmaceutical recommendations was high, particularly in case households. These closures occurred during a well-defined and relatively constricted time frame, at the very beginning of the pandemic strain's emergence in Australia, where Victoria was the first state to report person-to-person transmission. As Australia was one of the first countries to experience pH1N1 outbreaks during the Southern Hemisphere winter, local public health officials were uncertain of the likely severity of disease and acted according to the 'worst case scenario' recommendations of the AHMPPI 2008 during the initial Contain phase. Considerable media attention was focussed on school-based spread of infection and the associated public health response. Our findings may therefore be indicative of a 'best case' estimate of the public's compliance during a moderate to severe influenza pandemic.
Issues arising in the conduct of our survey highlighted the considerable logistic challenges involved in implementing this complex policy on a large scale. In seeking to quantify implementation of school closure measures in Melbourne during the 2009 pH1N1 response, it was first apparent that government records of the intervention did not accord with the level of stated school involvement. Reasons for this discrepancy were unclear, but based on discussions with principals, did not represent school refusal to comply with directives. An alternative explanation might relate to the practical challenges involved in centralized administration of a localized reactive public health intervention, applied across many sites. The highly variable quarantine duration recommended to families provides further support for this hypothesis.
Inevitable delays to response arising from the multiple steps to initiation of closure including: case diagnosis, public health reporting, contact identification and information dissemination were reflected in frequent reports of quarantine periods less than seven days. A quarantine duration of three days or less may not reliably exclude development of infection, given some variation in the length of the presymptomatic infectious period , particularly in children . Moreover, as the period of isolation was to extend for a total of 7 days following last contact with an infected individual, it must be assumed that those contained for a shorter period had already spent several days post-exposure mixing freely in the community, during the time at which they were most likely to be infectious.
Strengths and limitations of the study
This is the first study to evaluate implementation of school closure on such a large scale, with our 33 schools representing an intervention conducted across the whole of metropolitan Melbourne. The low response rate from invited participants in this study is consistent with that observed in similar surveys [17–19], but does introduce potential for ascertainment bias. In particular, we received a disproportionately low level of responses from less advantaged schools, limiting our ability to represent the whole population experience and possibly inflating estimates of compliance. Also of note, the proportion of households that contained a confirmed case (20%) was considerably higher than that in a recently published West Australian (WA) study of school closures (5%) . This may suggest that not only more affluent, but also more concerned and/or compliant parents were more likely to take part in our study. Study materials were not available in languages other than English, which may also have excluded vulnerable subgroups in the population sample. Unfortunately, as invitations to participate were distributed through schools due to privacy constraints, we are not able to characterize non-respondent households in more detail. Further, conduct of the survey several months after closures took place may have reduced motivation to participate, and introduced the possibility of recall bias.
Findings in relation to other studies of quarantine compliance
Why was compliance with quarantine recommendations so high in our sample? The study of school closures in WA, implemented later than in Victoria and with greater awareness of the generally mild nature of pH1N1 disease, found greater frequency of excursions outside the home (75%) than did our survey . Unlike our sample, the WA study included 'peers' as well as those children identified as actual 'contacts'. The latter were more likely to stay at home than their unexposed friends, exceeded only by cases, of whom there were relatively few . Frequent socialization was reported among students sent home during pH1N1 driven closures in the United States (US) , in keeping with earlier observations during a large seasonal influenza B epidemic, in which individual risk perception was assessed and reported to be low . Australian surveys have found a lower anticipated compliance with voluntary quarantine measures for seasonal influenza infection, compared with a pandemic virus .
Parental care in the home was associated with higher compliance with social restrictions. During pH1N1 associated elementary school closures in Pennsylvania, only one in five parents took time off work to care for children despite dual income earners in two thirds of households. In that study, 69% of affected children made excursions to locations outside the home during the closure period . A recent contact diary study reported a 50% reduction in child socialization during school holiday periods in the United Kingdom (UK) compared with term time, suggested to be predictive of behavior during a public health intervention . However, the relevance of this finding to an emergency school closure setting should be interpreted with caution, as making 'ad hoc' arrangements for child care at short notice may lead to very different patterns of child socialization, compared with periods of scheduled leave.
Oseltamivir was well accepted by respondents in this study, with almost all taking at least half of the course, and very few reporting side effects. In a 'real-time' survey from the UK, just under half of secondary school students and three quarters of primary school students completed a prescribed course of oseltamivir . Non-compliance was ascribed to gastro-intestinal side-effects in half, and may have been more reliably reported than in our study due to an absence of recall bias, although questionnaires were only completed by around 40% of the sample population . Similarly high rates of adverse events were seen among children receiving oseltamivir in a comprehensive school in the South-West of England, but with better compliance and a higher study participation rate (> 90%) .