Our results suggest that entry screening did not lead to substantial delays in local H1N1 transmission (Table ). This empirical study is consistent with theoretical results from previous modeling studies [3
] and findings from previous pandemics [7
]. While longer delays in local transmission to the summer in countries in the Northern hemisphere could have substantially aided pandemic mitigation, due to seasonal factors [3
] and school vacations [10
] leading to lower peak attack rates [12
], the observed delays in the present pandemic suggest entry screening provided around 1-2 weeks of additional time for preparation and planning.
While our study focused on the impact of entry screening, some nations also implemented other containment and mitigation measures, such as isolation of suspected or confirmed cases, quarantine of their contacts with or without antiviral chemoprophylaxis, school closures or other social distancing measures, and public health campaigns to improve hygiene. Most nations enhanced their influenza surveillance. If countries that expended greater effort into entry screening also had more effective containment and mitigation measures in the general population, these might have led us to overestimate the effect of entry screening. Conversely, if countries that expanded greater effort into entry screening also tended to have better influenza surveillance and were able to identify local transmission earlier, we may have underestimated the effect of entry screening. Other differences between countries in laboratory capacity and availability of public health resources may also have confounded our evaluation, and all of these factors are limitations of our study.
Previous mathematical modeling studies have questioned the value of entry screening, since it could only delay rather than prevent local epidemics [3
]. However, most models assumed that source countries would conduct exit screening and infectious cases would not travel [3
]. In such a scenario it is not surprising that entry screening would add little benefit, since most journeys are shorter than the average 1.5-2 day incubation period for influenza A virus infections [5
]. Screening is unlikely to identify 100% of ill travelers, while some might use antipyretics to reduce a fever prior to passing through thermal scanners, or fail to report symptoms on declaration forms. Many individuals with subclinical or asymptomatic illness would not be identified, and could initiate outbreaks after arrival [14
]. In Hong Kong, only one third of confirmed imported H1N1 cases were identified through screening on entry to Hong Kong, the majority of imported cases were identified through the local health care system after arrival (T. Tsang, personal communication). A similar experience has been reported in Singapore [15
]. Nevertheless, entry screening could act as a deterrent to traveling when ill or lead to other indirect benefits such as improving public awareness of the pandemic.
For entry screening to be successfully employed, substantial resources are required to identify the small fraction of travelers who may have H1N1 infection [16
]. Further resources may be needed to isolate identified cases, and trace and quarantine close contacts. An important caveat is that a delay in inevitable local transmission of a pandemic virus may not be desirable if it would defer local transmission into a season associated with higher transmissibility such as the winter in temperate regions [12
], or if it led to importation and local transmission of antiviral resistant strains [17
In addition to the caveats on potential confounding by resource availability, competing interventions, and other differences discussed above, there are a number of further limitations to our study. First, identification, confirmation and notification of H1N1 cases is unlikely to have been perfect given the mild and self-limiting nature of most infections, and dates of importation and local transmission that we report may lag behind the true events of interest. Nations that devoted greater resources to entry screening may have identified imported cases earlier. Secondly, we have not considered the size of local epidemics, or how the degree of connectivity with source regions (for example the number of travelers per day) might relate to time delays between imported and local cases. Thirdly, by focusing on nations with at least 100 confirmed cases by July 6, 2009 we may have excluded nations where entry screening was more effective in delaying local transmission, or excluded some nations with fewer resources available for surveillance and confirmation of local cases. Fourthly, while we searched for the dates of reporting of the first imported case and first local case, these dates may not have corresponded exactly to the dates of identification and confirmation of those cases, since in some cases delays may have occurred for various reasons including political considerations. Finally, we collected data from online sources including official government websites, and we have included the hyperlinks in Additional file 1
, but information available on the internet could be inaccurate.