We obtained weekly age-specific ILI incidence rates during 2004–2010 from a systematic national surveillance system in Chile (13). ILI surveillance relies on 42 sentinel outpatient sites located throughout the country; these sites are representative of the general population and systematically report weekly age-specific physician visits for ILI (13) (online Technical Appendix, http://wwwnc.cdc.gov/EID/article/20/7/13-0967-Techapp1.pdf). We characterized the effect of the 2-week winter break period on influenza transmission during 2004–2010 by comparing trends in weekly ILI incidence rates among schoolchildren (5–14 and 15–19 years of age) and adults (20–64 and >65 years of age). To estimate changes in the age distribution of ILI patients, on the basis of methods used in previous work (8,14), we compared the weekly ratios of ILI incidence rates for schoolchildren and adults during the 2-week period before, during, and after the winter break by using a 1-sided Z test. We also considered a 6-week window (8,14) before and after the winter break as a sensitivity analysis. A decline in the schoolchildren-to-adult incidence rate ratio indicates a shift in the age distribution of patients toward adults, suggestive of decreased influenza transmission among schoolchildren (8).

In Chile, wintertime influenza activity peaks during May–September, which is typical of temperate regions in the Southern Hemisphere (15). The 2-week winter school break typically coincides with the influenza season and is synchronous throughout the country; ≈95% of educational institutions follow the break periods set by the Ministry of Education.

Figure 1 illustrates trends in ILI incidence rates among schoolchildren 5–19 years of age and adults >20 years of age throughout the year and the associated schoolchildren-to-adults incidence rate ratio. In Chile, ILI incidence displays bimodal patterns of activity; activity increases before and after the winter break, and transmission is reduced during the break. The schoolchildren-to-adults ratios decreased substantially (40%–68%) during the 2-week winter break period relative to the 2-week period immediately preceding the winter break (Table 1). Also, the reduction in ratios coinciding with the first week of the winter break occurred every year of our study, including during the pandemic (2009) and postpandemic (2010) seasons.

The reduction in the schoolchildren-to-adults incidence rate ratios was maintained for an average of 2 weeks after the end of the winter break. The decline in ratios was primarily caused by a decrease in ILI rates among schoolchildren; the average (+ standard error of the estimate) reduction in ILI incidence among schoolchildren (5–19 years of age) in the 2 weeks during the winter break compared with the 2 weeks before was 67.2% + 2.1% (p<0.001). This reduction occurred systematically in each winter of the study period. In contrast, the average reduction in adult ILI incidence (>20 years of age) was more modest but remained significant at 37.4% + 0.9% (p<0.001).

Furthermore, the incidence rate ratios for school-age children to middle-age adults significantly increased after the winter break, signaling a return toward a higher proportion of ILI cases among children, although the ratio did not return to prebreak levels (Table 1). In contrast, the ratio comparing rates for children with rates for adults did not change. Our results did not change when we used a 6-week period before and after the winter break period instead of a 2-week period (Table 2) or when we excluded the 2009 pandemic year from our analysis (Technical Appendix Figure 1).

We have shown that a two-thirds decline in ILI incidence among schoolchildren coincided with the onset of the school winter break in Chile; this pattern was consistent across the 7 years of the study. In line with a prior study in Argentina (6), the average reduction in schoolchildren-to-adults incidence rate ratio was sustained for up to 2 weeks after school sessions resumed. This time scale is consistent with the natural history of influenza virus infection, which has a serial interval (interval between cases) of 2–3 days, so that it takes a few successive chains of transmission to reach full-scale transmission.

Similar to findings from prior studies (5,6), our findings are based on analysis of ILI incidence, which is a broad indicator of respiratory disease activity in a community and is not entirely specific for influenza. Our results could be affected by changes in health-seeking behavior during the winter break. However, our ILI data are well correlated with influenza virus activity data (15) (Figure 2), and large increases in incidence among schoolchildren during winter 2009 coincide with the influenza A(H1N1)pdm09 virus pandemic period, suggesting that fluctuations in ILI incidence in Chile are primarily attributable to influenza. Our data also support the conclusion that school closure during pandemic situations is effective. Although the winter break took place near the peak of the 2009 influenza A(H1N1) pandemic in Chile, it was correlated with changes in the age distribution of patients hospitalized for influenza A(H1N1)pdm09 virus infection (7).

Overall, our study findings add to the body of information provided by empirical studies, supporting the implementation of school closure to achieve temporary reductions in ILI incidence rates, especially among school-age children, including in the Southern Hemisphere temperate setting (2–7). Our finding that ILI incidence was more modestly reduced among adults during winter breaks is consistent with past work on the age-specific transmission dynamics of influenza (2–7). School closure may be particularly useful in pandemic situations to gain time until pharmaceutical measures (vaccines, antiviral medications) become available and to mitigate the burden on health care institutions by reducing the surge of influenza patients. There is still, however, little information available from tropical and Southern Hemisphere settings, which are characterized by complex influenza seasonality patterns and/or low connectivity with the rest of the world and particular demographic and health conditions. Systematic multicountry and multiyear comparison of the effects of school closures could shed light on the effectiveness of school-based intervention policies under different epidemiologic, behavioral, and demographic situations.

Epidemiological surveillance of influenza-like illness (ILI) in Chile. Average weekly incidence rates for ILI among schoolchildren 5–15 years of age and adults >20 years of age, Chile, 2004–2010, excluding data from the 2009 influenza A(H1N1) pandemic year.