The optimal strategy to mitigate the population health impact of influenza epidemics is unclear. Safe and effective vaccines are the cornerstone for prevention of both seasonal and pandemic influenza; however, coverage remains suboptimal in the US despite long-standing recommendations for targeted vaccination of high risk groups and seniors, and the recent expansion to universal vaccination.[1], [2], [3], [4]

Children are crucial transmitters of influenza virus within communities.[5], [6] School-aged children have high influenza attack rates, are infected early in the course of epidemics, shed viruses for longer periods than adults and have a higher intensity of social contact than other age groups.[5], [7], [8], [9] Therefore, targeted vaccination of school-aged children, especially when there is a good match between vaccine and circulating strains, could reduce the impact of epidemics by preventing both influenza disease in vaccinated children and also transmission of the virus to other age groups.

Live attenuated influenza vaccines (LAIVs) are attractive for school-based interventions because compared to inactivated vaccines, they have higher efficacy in young children,[10], [11], [12] result in reduced virus shedding among vaccinees after influenza infection,[13] and their “needle-less” mode of administration has high acceptability.[11], [14], [15] School-based vaccination campaigns with LAIV have been reported to reduce influenza illness among vaccinees,[16], [17], [18] reduce influenza-like illnesses among their close contacts, and modestly reduce school absenteeism.[16], [17] Although mathematical models and experimental evidence suggest benefits of such targeted vaccination strategies among unvaccinated persons in selected populations (i.e. indirect or herd protection), [9], [19], [20], [21], [22], [23] consistent evidence of herd protection is limited.[14], [16], [24]

In 2005, Tennessee's Knox County Health Department launched a school-based influenza vaccination campaign with the aim of immunizing public school children from kindergarten through 12th grade. From October-December 2005, LAIV was offered through weekday in-school and weekend clinics. Children with signed parental consent forms and without known contraindications to LAIV were vaccinated. MedImmune donated LAIV and other campaign costs were assumed by the Knox County Health Department.[18], [25], [26], [27] Prior to the 2005–2006 influenza season, 24,198 (45%) of 53,420 Knox public school students were immunized with LAIV. Approximately 58% of 5,099 children aged 5–9 years received a second dose. In the following 2006–2007 campaign year, MedImmune donated LAIV and additional resources to strengthen vaccination efforts and to extend the campaign to private schools. Prior to the second (2006–2007) season, 47% of 54,786 public school children and 61% of 5,998 private school children were immunized. Approximately 53% of children aged 5–9 years received a second dose. The estimated overall coverage of eligible Knox school children aged ≥5 years was 41% and 48% for the 2005–2006 and 2006–2007 influenza seasons, respectively.[18], [25], [26], [27]

The second year of the Knox vaccination campaign (2006–2007) was evaluated in two studies that compared influenza activity in Knox with that of another urban Tennessee County. These studies suggested that the campaign resulted in a reduction in the proportion of influenza associated ED visits in school-aged children and influenza hospitalizations in adults aged 50 to 64 years.[25], [26] A third study reported that compared with children from surrounding counties, Knox County children had a reduced proportion of positive rapid influenza tests during the campaign seasons (2005–2007).[18] Nevertheless, none of these studies could quantify the campaign effects on actual ED and hospitalization rates among school-aged children, and the effects on non-targeted age groups remains unclear. In the present study we assessed the population impact of the two-year countywide Knox influenza vaccination campaign.

Our findings indicate that the large Knox school-based influenza vaccination campaign was associated with a significant reduction in morbidity attributable to influenza. These declines coincided temporally with the vaccination campaign and were observed exclusively in the 5–17 year old target population of the campaign, with consistent results observed in secondary analyses. Nevertheless, our study had limited power to detect significant declines among non-targeted age groups, and none were detected.

The incidence of MAARI ED visits attributable to influenza declined in the target population, indicating a direct protection of the vaccination campaign. No significant effects were observed on MAARI hospitalizations but these events were rare, and despite the large sample size of our study, we could not rule out sizable campaign effects on this outcome. Based on the analyses of data from pre-campaign years for the target population, we previously estimated that our study would be powered to detect a 20% decline in ED visits rates, but a much larger 91% decline in hospitalization rates.[28]

Although previous studies suggested that indirect protection of unvaccinated groups against influenza-like-illness and reductions in absenteeism could be achieved through vaccination of school-aged children,[14], [15], [16], [17], [41], [42] we were unable to show significant disease reductions among age groups that were not directly targeted during the Knox vaccination campaign. It is possible that the influenza vaccination coverage among school children (approximately 45%) was not high enough to achieve significant protective levels in other groups. A previous study reported significant indirect protection against MAARI among adults aged ≥35 years with even lower vaccination coverage of children aged 1.5 through 18 years (e.g. approximately 25%, 35% and 50% coverage during three consecutive influenza seasons).[14] However, in that study significant protection was not demonstrated among children or younger adults who were vaccinated.[14] It is important to note that most contemporary school or community-based influenza studies have achieved vaccination coverage of 50% or lower.[14], [15], [16] The feasibility of achieving higher vaccination coverage through promotion or education campaigns deserves further scrutiny.

Another factor limiting the ability of this study to demonstrate indirect effects could be vaccination coverage among non-targeted age groups. Prior studies showed parent/self-reported vaccination rates in Knox County residents of 36% for children aged <5 years and 70% for adults aged 50 years or older.[25], [26] Detecting indirect protection in populations where influenza vaccination is now universally recommended will be challenging, and limited to reductions in the residual disease burden in those populations. A recent article assessed the effect of school-based influenza vaccination on healthcare encounters in the Province of Ontario, Canada (2000–2007). This province has had a universal influenza vaccination program since year 2000. Within this context, Kwong et al[43] compared the incidence of influenza associated healthcare encounters between public health units (PHUs) that had school-based influenza vaccination available in at least 50% of elementary and secondary schools, with PHUs without such availability. Vaccination coverage in children in secondary schools was 39% in PHUs with and 30% in those without vaccination programs available. There was a modest reduction in the frequency of physician office visits for influenza-associated pneumonia and influenza in children attending secondary schools in PHUs with vaccination programs compared to those without such programs (estimated vaccine effectiveness: 19%), but no significant reductions in emergency department visits or hospitalizations among school aged-children. There were no significant differences in visits for the more common influenza-associated acute respiratory illnesses in the targeted age group; and, similar to our study, no benefits were detected in other non-targeted groups.[43]

A recent cluster randomized clinical trial demonstrated that vaccination of 83% of children aged 3–15 years in Hutterite colonies significantly reduced the incidence of influenza illness in both immunized children and their unvaccinated close contacts. Although the study also reported non-significant declines in physician visits for respiratory diseases and influenza like-illnesses in colonies randomized to influenza vaccines, the trial was not adequately powered to assess these outcomes.[20] The very high vaccine coverage achieved, the particular demographics and low background vaccination coverage in the study communities likely contributed to the observed indirect protection. Additional studies in populations with low background vaccination rates may help quantify the minimum vaccination coverage required to attain measureable indirect benefits.[24]

The identification of a proper comparison population had implications for our assessment of effectiveness. Previous studies have demonstrated substantial differences in the seasonal incidence of influenza-related healthcare encounters between different communities.[44] Our assessment of the campaign effectiveness used Knox-surrounding counties as comparison because in spite of minor demographic differences between study populations (Table 1), the incidence of healthcare utilization attributable to influenza was similar in five consecutive pre-campaign influenza seasons.[28]

The two different analytical approaches used to estimate MAARI attributable to influenza found similar estimates.[34] Our analyses suggest that the overall effectiveness of the Knox campaign in preventing ED visits among children aged 5–17 years was 35% (95% CI: 16–54). This estimate encompassed both direct and indirect protection in this age group,[43] and suggests that the Knox vaccination campaign prevented approximately 300 influenza ED visits in this target population during each influenza season. Using the estimated effectiveness, we also calculated crude estimates of vaccine efficacy for Knox children aged 5–17 years. Taking into account that ∼45% of all Knox school-aged children were immunized, the observed excess rate ratio of 0.65 and assuming that influenza rates were the same for unvaccinated children from Knox and Knox-surrounding counties, the estimated vaccine efficacy provided by LAIV against ED visits would be 77.8% (RR: 0.65 = [1−(0.45*Vaccine Efficacy)]/1). These estimates assumed no vaccination of school-aged children in Knox-surrounding counties; accounting for influenza vaccination in control communities would increase the vaccine efficacy estimates.

We used all data available for ED visits and hospitalizations for the study populations. This information was systematically collected for administrative purposes and thus, unlikely to be affected by recall issues. Study data included all hospital-based care without regard to payment sources and thus it was virtually free of selection issues. Moreover, the use of laboratory-confirmed influenza information to define influenza seasons reduced misclassification of periods with influenza activity.

Despite those strengths, the interpretation of our findings requires the consideration of several caveats. First, our data lacked virological confirmation for all observed MAARIs and potential misclassification of MAARI attributable to influenza cannot be ruled out. Nevertheless, our estimates were restricted to laboratory-confirmed influenza seasons and we consider that any residual misclassification would be similar in the two study populations. Second, despite its large sample size, our study did not detect significant changes in the non-targeted adult population combined, and had limited power to identify significant effects of the intervention on ED visits for adult age subgroups. Moreover, the limited number of hospitalizations observed reduced our ability to detect true effects on this outcome. Third, information on other potential causes of MAARI during winter months (e.g. respiratory syncytial virus [RSV]) was not specifically accounted for in our analyses. However, previously studies suggest that estimates including or excluding RSV data did not significantly affect influenza estimates.[29], [36], [45] Also, RSV activity would be of greater concern in younger age groups than among children aged 5-17 years. Nevertheless, lack of adjustment for RSV illness may have limited our ability to detect indirect protection in children aged <5 years. Finally, we did not measure influenza vaccination coverage in other age groups. High vaccination rates in seniors and other age groups may have made it difficult to demonstrate indirect effects in these groups.

In conclusion, our findings indicate that a large school-based influenza vaccination campaign with LAIV that attained 45% vaccination coverage, significantly reduced ED visit rates attributable to influenza among Knox County children aged 5–17 years, the main target of the campaign. During campaign years there were approximately 5 fewer influenza-associated ED visits per 1000 Knox children aged 5–17 years annually. Higher vaccination coverage and/or larger studies will be needed to determine the effect of similar interventions in other age groups.