Effects of School Holidays on Seasonal Influenza in South Korea, 2014–2016

Abstract

School closures are considered as a potential nonpharmaceutical intervention to mitigate severe influenza epidemics and pandemics. In this study, we assessed the effects of scheduled school closure on influenza transmission using influenza surveillance data before, during, and after spring breaks in South Korea, 2014–2016. During the spring breaks, influenza transmission was reduced by 27%–39%, while the overall reduction in transmissibility was estimated to be 6%–23%, with greater effects observed among school-aged children.

Schools are recognized as an important location for influenza transmission as school-aged children have more social contacts than other age groups [1]. Changes in social contact patterns may shape the transmission of influenza in the community [2], and school closures is a social distancing measures that can be used to mitigate severe influenza epidemics and pandemics [3]. While schools are not often closed for public health reasons, regular school holidays sometimes coincide with influenza peak seasons and provided us with an opportunity to assess their impact on influenza transmission. A number of previous studies have examined the relationship between school closure and influenza dynamics [3]. Preemptive school closure is likely to reduce influenza transmission in the community, while reactive school closure and routine school holidays may have a more temporary effect. However, such studies have not been carried out in South Korea.

In November 2014, the Korean Ministry of Education announced a plan for diversification of the school curriculum, which included abolishing the unified period of spring breaks [4]. Here, we examined influenza-like illness (ILI) and influenza virus surveillance data in South Korea to estimate the transmissibility of seasonal influenza viruses and the effect of spring school holidays on seasonal influenza transmission.

METHODS

Sources of Data

South Korea, located at the east of mainland Asia, has a population of 51.4 million and a temperate climate with 4 seasons. Influenza epidemics in Korea occur in the winter each year, between October and May. In South Korea, the spring school breaks usually start in February and last for 2–3 weeks, covering the Lunar New Year holiday. In 2013–2014, 2014–2015, and 2015–2016, the influenza peak seasons coincided with spring breaks, permitting us to include them in the analyses, while the 4 influenza peak seasons from 2016–2017 to 2018–2019 occurred early, before February, and had ended before the spring break [5]. We examined the effect of the spring school breaks on transmission of influenza virus in the community in 2013–2014 to 2015–2016 influenza seasons using influenza surveillance data from the Korean national influenza surveillance system, comprising 200 sentinel outpatient sites located throughout the country [5]. We obtained the weekly ILI consultation rate from the sentinel sites and the weekly influenza virus detection rate, determined by reverse transcription polymerase chain reaction (RT-PCR), in clinical samples. We multiplied these together to obtain a proxy measure of influenza activity in the community, which has been shown have a good linear correlation with the incidence rate of influenza virus infections [6].

Statistical Analysis

To assess the effect of spring school breaks on influenza transmission, we first estimated influenza transmissibility over time using the effective reproduction number (R_t) [7]. R_t represents the mean number of secondary infections per primary case at time t, and an R_t value below 1 indicates that the epidemic is under control. We first applied flexible cubic splines to weekly surveillance data to obtain interpolated daily data, as previously described [8]. The serial interval distribution, defined as the duration between symptom onsets of the primary and secondary cases, was assumed to follow a Weibull distribution with mean of 3.2 days and standard deviation of 1.3 days [9]. We estimated daily R_t from epidemiological week 51 to week 21 for each influenza season in 2013–2014 to 2015–2016, and compared the mean R_t during the spring breaks versus the preceding and following 2-week period. We further carried out similar analysis for different age groups (0–6, 7–18, 19–49, 50–64, and > 65 years).

We also estimated the overall impact of each school holiday by fitting a regression model for R_t accounting for depletion of susceptibles in the population. Based on epidemic theory, $R_t=R_0{S}_t{e}^{\lambda B_t}$, where R₀ is the basic reproduction number (measure of intrinsic transmissibility), S_t is the community susceptible proportion at that time t, and B_t is an indicator variable, defined as $B_t=1$ during spring break and 0 otherwise [8]. λ is the coefficient representing the effect of spring breaks on R_t (λ = 0 indicates no effect from spring breaks and λ < 0 indicates reduced transmissibility during spring breaks). By using an approximation $S_t\approx S_0{e}^{\alpha h_t}$ (see Supplementary Material), $R_t=R_0{S}_0{e}^{\alpha h_t}{e}^{\lambda B_t}$, where S₀ is the initial proportion of susceptibles in the population, α is the coefficient of $h_t=\underset{x=1}{\sum ^{t-1}}{I}_x$, a variable indicating the depletion of susceptibles at time t, and I_t is the number of infectious cases at time t. This defines the log-linear multivariable regression model $\ln (R_t)=constant+\alpha h_t+\lambda B_t$, where α and λ were estimated (see Supplementary Material). The overall reduction in R_t is estimated by $(1-e^{\widehat{\lambda }})$.

RESULTS

Figure 1 shows influenza activity and estimated R_t for the 3 influenza seasons. Mean R_t ranged from 1.13 to 1.40 during the 2 weeks before the spring breaks, decreased to 0.82–0.90 during the break, and rebounded slightly to 0.96–1.10 in the 2 weeks following the breaks (Table 1 and Figure 1). This suggested an immediate 27%–36% reduction in influenza transmission during the spring breaks compared to the preceding 2 weeks. Similar effects on influenza activity were observed across different age groups, including preschool (0–6 years, 28%–36%), school-aged children (7–18 years, 34%–42%), and young adults (19–49 years, 19%–36%) during the study period (Supplementary Figure and Table 1). In the regression analysis for R_t, the 3 spring breaks were associated with 17.8% (95% confidence interval [CI], 14.2%–21.3%), 22.8% (95% CI, 17.0%–28.1%), and 6.4% (95% CI, .2%–14.3%) reductions in influenza transmission (Table 1).

Figure 1.

Influenza activity and estimated effective reproductive number (R_t) in South Korea, 2014–2016. Black solid lines indicate influenza activity measured by multiplying influenza-like illness consultation rate with influenza virus detection rate from the Korean national surveillance database. Shaded gray bars represent the period of spring school breaks and Lunar New Year (light gray bars). Gray solid lines indicate daily estimated R_t and gray dashed lines indicate pointwise 95% confidence intervals (CIs) of R_t; gray horizontal lines indicate transmission threshold (R_t = 1).

Table 1.

Estimated Daily Reproductive Number R_t During, Before, and After Spring School Breaks With Respective Reductions in R_t, South Korea, 2014–2016

	Mean Estimated Rt (95% CI)			Reduction in Rt, Spring Break vs 2 wk Before, % (95% CI)	Overall Reduction in Rt Associated with Spring Break, % (95% CI)a
Influenza Season	During 2 wk Before Spring Break Started	During Spring Break	During 2 wk After Spring Break Ended
2013–2014	1.13 (1.11–1.14)	0.82 (0.81–0.83)	0.96 (0.95–0.97)	27.3 (25.9–28.7)	17.8 (14.2–21.3)
2014–2015	1.27 (1.25–1.28)	0.86 (0.84–0.88)	1.10 (1.10–1.11)	31.9 (30.0–33.8)	22.8 (17.0–28.1)
2015–2016	1.40 (1.39–1.41)	0.90 (0.88–0.91)	1.02 (1.00–1.03)	36.0 (34.6–37.5)	6.4 (0.2–14.3)

^aEstimated from the regression model for R_t (see Supplementary Material).

DISCUSSION

We found that scheduled spring school breaks around influenza peaks were associated with a moderate (6%–23%) reduction in influenza transmission in South Korea from 2013–2014 to 2015–2016. Such a reduction could have a large effect on transmission, given a typical reproduction number of 1.2–1.4 for seasonal influenza [10]. Our results are comparable to previous studies on reactive school closures, where reductions of 16%–25% in influenza transmission were reported [8, 11]. The overall reductions in influenza transmission during the spring breaks were highest among the school-aged children, consistent with the findings of previous studies [11, 12]. Our estimates are likely to provide a lower bound for the effect of preemptive school closures around the epidemic peak, during which the general population would be more likely to adopt enhanced preventive measures and other social distancing measures [12]. In each of the 3 influenza season we studied, R_t decreased further at the start of the spring break, and in the 2014–2015 and 2015–2016 influenza seasons R_t rebounded near the end of the spring break and increased by 27% and 13%, respectively, after schools resumed (Table 1). This is a clear indication that transmission was suppressed during the school holidays [12]. However, we also observed a clear rebound in transmissibility after the Lunar New Year holiday in 2014–2015, possibly due to the increased social interactions through gatherings outside the school settings during long spring break. In November 2014, the Korean Ministry of Education recommended revision of the school curriculum and study timetable [4], and so the timing of the spring school breaks became more variable in the 2015–2016 school year. Incidentally, we found a comparatively weaker overall effect of the spring school breaks in 2015–2016 (Table 1), where the transmissibility was sustained at a higher level during the spring breaks compared to the 2 previous influenza seasons (Table 1). This suggested that the timing, duration, or synchronicity of school breaks might have an impact on transmissibility of influenza epidemics [13].

In our study, the changes in transmissibility during spring breaks were estimated based on a proxy measure for seasonal influenza virus activity, which is likely affected by health care-seeking behavior and changes in laboratory testing practice. However, the consultation rate was still high at the beginning of the spring breaks in 2014–2015 and 2015–2016 as the prime travel seasons in Korea are during summer and winter vacations (July–August and December–January). Furthermore, laboratory testing is conducted by the National Research Institution of Public Health and Environment, which has consistent laboratory testing practice over the year [5]. The influenza vaccine coverages among individuals younger than 18 years were similar in 2012 to 2015 (2012, 49.7%; 2013, not available; 2014, 47.8%; and 2015, 49.3%) [14]. The annual influenza vaccination campaign started early, in around October each year, and its effect was reflected in the estimated R_t. The social contact patterns before, during, and after the spring breaks and Lunar New Year could help to further examine influenza transmission between and across different age groups; however, such data were not available. A previous study demonstrated that school-aged children, who play an important role in introducing influenza virus infections into households, usually reduced social contacts among themselves but increased contacts with adults during school holidays [2]. We did not observe lags in the timing of the epidemic peaks among adults (Supplementary Figure). This is consistent with the notion that children may not necessarily lead the influenza epidemic waves [15].

Our findings support the conclusion that school holidays around the peak of the influenza season are likely to limit further increase in influenza activity among school-aged children, as well as other age groups, and may relieve burden on the health care system. However, it should be noted that emergency school closures can cause adverse consequences in the community such as additional cost for the alternative childcare arrangements, particularly when the duration of closure is prolonged.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. The authors thank Julie Au for technical assistance and staff from the Korean Education Development Institute for helpful discussions about the Korean school holidays.

Disclaimer. The funding bodies had no role in study design, data collection and analysis, preparation of the manuscript, or the decision to publish.

Financial support. This work was supported by the National Research Foundation of Korea Ministry of Education Basic Science Research Program (grant number NRF-2018R1A6A3A03012236); Harvard Center for Communicable Disease Dynamics from the National Institute of General Medical Sciences (grant number U54 GM088558); and a commissioned grant from the Health and Medical Research Fund and the Research Grants Council of the Hong Kong Special Administrative Region, China (project number T11-705/14N).

Potential conflicts of interest. B. J. C. has received honoraria from Roche and Sanofi Pasteur. All other authors report no potential conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.