Epidemiological Surveillance To Establish Thresholds For Inuenza Among Children In Satellite Cities Of A metropolitan area of Tokyo, Japan

Background: Few reports have longitudinally investigated seasonal inuenza epidemiological surveillance data of pediatric populations in the metropolitan areas of Japan. We aimed to provide descriptive characteristics of circulating inuenza and to investigate the usefulness of setting thresholds for inuenza in children (0–15 years old) in two satellite cities of a metropolitan area of Tokyo, Japan, for ve consecutive seasons of the inuenza epidemic. Methods: The survey was conducted annually during the inuenza season, from 2014 to 2018 (ending March 2019), at preschools (kindergartens and nursery schools), elementary schools, and junior high schools located in Toda and Warabi cities, Saitama prefecture. We investigated the epidemiological characteristics and established thresholds using the World Health Organization method. Results: Of the 108,362 children (21,024 to 22,088 throughout ve seasons) who received the questionnaire, 76,753 (70.8%; 14,652 to 15,808) responded. After exclusion of responses without basic information, 64,586 children were included in the analysis, of which 13,754 (21.3%) had tested positive for inuenza. Inuenza type A was generally dominant, whereas type B was responsible for a substantial share of all inuenza cases (>40% in seasons 2015 and 2017, when type A circulated with low incidence). The weeks when the inuenza epidemic peaked had no clear seasonal pattern among the surveyed years, i.e., the peaks appeared at week 51 (mid-December) or later, whereas the World Health Organization methods reported that the median period when a peak was observed was at 3 weeks (mid-January), regardless of school age group. Conclusions: The present information obtained from the epidemiological survey regarding seasonal inuenza in children would be useful for general practitioners, health policymakers, and planners who establish prevention and control methods against inuenza.


Background
The World Health Organization (WHO) estimated that annual epidemics cause 3 to 5 million in uenza cases with severe illness worldwide [1]. The epidemiology of in uenza changes markedly each year and from location to location [2]. In general, approximately 80% of in uenza cases are caused by in uenza type A, whereas in uenza type B accounts for approximately 20% of the total global cases [3]. When considering the threat of in uenza, school children are the primary vulnerable population, because they have the highest rates of in uenza transmission and infection among infected populations [4]. In the Asia-Paci c region, in uenza type B appeared to cause more illness in children between ages of 1 and 10 years than in other age groups [5]. Although surveillance data of in uenza have been reported in various forms for populations across Japan [6][7][8], studies that investigated seasonal in uenza among school children from the capital city of Japan, Tokyo, the most populous metropolitan area of the country, are scanty.
A variety of thresholds for the in uenza epidemic has been proposed [9][10][11][12]. These include the WHO global standards for the collection, reporting, and analysis of seasonal in uenza epidemiological surveillance data [9]. The WHO further recommends obtaining average epidemic curves and seasonal and alert thresholds as established tools to help control annual in uenza epidemics [9]. Thresholds using the WHO methods are simple to implement and can be adapted easily for any in uenza surveillance system with adequate historical data [13]. In some countries, the WHO method can be utilized to inform key decision makers in the areas of public health regarding in uenza outbreak management [14][15][16]. Based on our recent surveillance, data of children (from preschool to junior high school age) during ve consecutive in uenza seasons in two satellite cities of the metropolitan area of Tokyo, Japan, we aimed to provide descriptive characteristics on circulating in uenza and to investigate the usefulness of establishing thresholds for the in uenza epidemic.

Study area
The study area comprised two cities, Toda and Warabi, which are located to the north of Tokyo, Japan.

Study procedure
Throughout ve consecutive seasons, from 2014 to 2018 (ending March 2019), an annual survey was conducted of parents of children who were attending preschool (kindergarten or nursery school), elementary school, and junior high school in the Toda and Warabi regions. The questionnaire obtained parent responses about the following information regarding their children: school, sex, sibling, underlying disease, vaccination, and number of children with in uenza by type. In clinical practice in Japan, the type of in uenza (types A or B) is typically diagnosed by the children's local doctor or an emergency outpatient healthcare provider who administers an in uenza antigen rapid test kit covered by the health insurance.
The survey was conducted every June, and the responses pertained to the preceding season.

Statistical analysis
We demonstrated the number of children, percentage of in uenza cases by type, week of in uenza epidemic peak by in uenza type for each season; and seasonal, average, and alert thresholds of in uenza. Data were also distributed by school group (preschool, 0-6 years old; elementary school, 7-12 years old; junior high school, 13-15 years old; in Japan, no repeat school year system has been applied for these school children). In the present analysis, we de ned each in uenza season as beginning in October of the start year and ending in March of the succeeding year; e.g., the 2014 season began in October 2014 and ended in March 2015. The epidemic peak was de ned for each in uenza type as the week with the highest number of detected in uenza cases.
According to the WHO protocol [9], we calculated the average and upper limit of 90% con dence interval curves and the seasonal, high, and alert thresholds based on the number of children with in uenza each week throughout the ve seasons. The average curves denote the peak weekly mean, and the 90% upper curve was for the upper limits of the 90% con dence intervals (CI) of the peak weekly mean [9,13]. For these curves, the WHO protocol suggests utilizing the normal distribution to assign thresholds based on the mean and standard deviation of the aligned data for weekly counts [9]. The seasonal threshold was de ned as the annual median amplitude of the number of children with in uenza per week throughout the study period; therefore, half of the study weeks are necessarily above the seasonal threshold and these correspond to the seasonality in the in uenza epidemic (e.g., from week 40 of 2014 to week 13 of 2015). The high threshold was de ned as number of children with in uenza higher than the average peak for each of the ve seasons, i.e., the peak number of children with in uenza of the average epidemic curves. Theoretically, we can expect that seasonal peaks can be over the high threshold in two or three of the ve seasons. Finally, we de ned the alert threshold as extraordinarily severe seasons, such as a pandemic [9,13,15] derived from the upper limits of the 90% CI of the high threshold de ned as above.
Data for the total number of children studied and for each school group from week 40 of 2014 to week 13 of 2019 were plotted against the calculated seasonal, high, and alert thresholds. We analyzed data using Stata version 16.0 (Stata Corp., College Station, TX, USA).

Results
A total of 76,753 responses were collected using the survey conducted via post mail for 108,362 parents of children attending preschool, elementary school, or junior high school during the 2014 to 2018 seasons (the collection rate for the questionnaire was 70.8%). We excluded respondents who did not answer basic information (n = 4,445) and had in uenza vaccination before September 30 or incidence of in uenza after April 1 for each season (n = 7,722) [18]. The present analysis, therefore, consisted of 64,586 respondents ( Fig. 1).

The prevalence of children with in uenza and their distribution by in uenza type
The total number of children who were reported to have been infected with in uenza was 13,754 (21.3% of the analyzed respondents; Table 1). With respect to the de ned in uenza season, the numbers in   The week of the in uenza epidemic peak by in uenza type As presented in Table 1

Curves and thresholds by the WHO methods
The thresholds from the WHO method are summarized in Table 3. Peak median number of children with in uenza was similar to the corresponding peak mean number. The average epidemic curve determined by the weekly mean and the upper 90% CI threshold determined by the weekly upper 90% CI curve are presented in Figs. 2 and 3, respectively. The median week of the peak was observed at week 3 (mid-January; Table 3 and Figs. 2 and 3). The plotted curve of number of children with in uenza crossed the seasonal threshold multiple times regardless of the ve seasons. The start of the in uenza season was between week 43 and week 1 (late October and early January), and the end of the in uenza season was between week 8 and week 13 (late February and late March). The peak seasonal in uenza activities in 2015, 2016, and 2017 did not reach high thresholds (Fig. 4). The peak seasonal in uenza activity varied when children with in uenza was strati ed according to their school groups, as illustrated in Fig. 5. In none of the ve seasons did the plotted curve of the in uenza incidence cross the alert threshold (Fig. 4).
The results were almost con rmatory when classi ed by school group, except for the junior high school group during the 2014 season, in which number of children with in uenza was close to the alert threshold ( Fig. 5).

Discussion
We presented data on the circulation of in uenza in children who were attending preschool, elementary school, or junior high school in Toda or Warabi, Japan, during ve consecutive in uenza seasons from 2014 to 2018. We also successfully established seasonal, high, and alert thresholds based on surveillance data from ve consecutive seasons of in uenza using the WHO method.

Number of children with in uenza and their distribution by in uenza type
In Japan, the governmental agency (Ministry of Health, Labour, and Welfare) in collaboration with the National Institute of Infectious Diseases (NIID) provides a weekly in uenza outbreak report [19]. This report is based on a school surveillance in which the absence of children and temporary closure of schools are recorded; therefore, the date of onset of in uenza cannot be assessed. Nonetheless, the total number of temporary school closures peaked in 2017, which supports our ndings that the greatest number of in uenza cases occurred in 2017. In junior high school children, our survey was different from the national report [19], i.e., the greatest prevalence of in uenza cases were reported in 2016 (23.4%; Table 2). In our survey, in uenza type B was responsible for a substantial share of all in uenza cases, approximately > 40% in seasons where in uenza type A circulated with low incidence in 2015 and 2017 ( Table 1). The results of our survey are essentially similar to the outbreak trend summary of in uenza for each season, as reported by NIID of Japan [19], whereas the prevalence of in uenza type B among children in junior high school was approximately > 50% in 2015 (Table 2). Outbreaks differed by region, even within a single country, and surveillance in a local area of the present study was demanded.
Week of the in uenza epidemic peak In the national report [19], each peak of temporary suspension of facilities occurred in weeks 4, 7, 4, 5, and 4 in the 2014, 2015, 2016, 2017, and 2018 seasons, respectively. The week of the in uenza epidemic peak in our survey occurred consistently earlier than that in the national report, although the gross tendency was similar. Our survey was different from the national report [19], which was based on the dates of absence from school in children with in uenza, and it is likely that the week with the highest number of detected in uenza cases proceeded the week of temporary school closures. Regional characteristics such as metropolitan area versus nationwide data [19] might affect the peak shift; however, relevant report has been lacking and we cannot say much about this difference.
The peaks of in uenza type B occurred later than those of type A (Tables 1 and 2). The epidemic order is in accord with that observed in the northern hemisphere [20]. Understanding the geographical and temporal patterns of seasonal in uenza could help strengthen in uenza surveillance for the early detection of epidemics [21]. As Mosnier and his colleagues reported [22], timely data on the circulation of in uenza collected within in uenza surveillance systems is essential for optimizing in uenza prevention and control strategies [21,22].
Establishment and assessment of seasonal, high, and alert thresholds We used WHO method-based thresholds and epidemic curves to assess the in uenza seasons from 2014 to 2019; these data represent the onset and end of the in uenza season, as well as the severity of in uenza. Epidemic peaks occurred at week 51 or later, in particular at ≥ 2 weeks among preschool children (Tables 1 and 2). We did not nd any apparent seasonal pattern on the weekly uctuations or on the week of in uenza epidemic peaks. Meanwhile, the seasonal, high, and alert thresholds were explored without being affected by the diverse seasonal patterns. The NIID provides another warning outbreak system. When the number of in uenza cases that occur at regional sentinel sites is more than 30 per week per site, the warning is issued to a responsible public health center [23]. We provided three threshold levels (seasonal, high, and alert thresholds) for children and each school group in Toda and Warabi based on the surveillance data that were captured in the same region. Our community-based survey in children can be used for pandemic in uenza assessment by accounting for a potential increase, with high sensitivity. It should be noted that we should alter these threshold values according to region and country.
Previous studies have proposed that the analysis of surveillance data could provide information about seasonal thresholds and epidemic curves that might help healthcare personnel in clinical management [16]. Although there is room for debate on the de nition of the period of in uenza activity [24], future surveillance data would allow for an assessment of early warnings with a revision of the nominated threshold levels [9].

Limitations
Although our approach can be crucial in guiding research in health plans for the prevention and control of in uenza in a community, our study has several limitations. First, because the questionnaires were answered by parents of targeted children in preschool, elementary school, or junior high school, preschool-aged children who were not attending kindergarten or nursery school and children who were attending school out of town were excluded from the analysis. In the study area (Toda and Warabi), the total number of children who were aged 15 years or younger was 27,562, according to the 2015 census.
Although approximately half of the respondents appropriately answered the questions in this metropolitan area survey, we cannot guarantee that the current ndings accurately represent the epidemiology of children in the general population. External validity also requires further investigation. Second, we used a questionnaire in which we did not request detailed medical information, so the answers might not be accurate. However, we emphasize that the in uenza antigen rapid test is widely available in Japan, and patients with a fever who visit an outpatient clinic are generally automatically checked using the test kit to detect in uenza and its type. We are, thus, con dent about the diagnosis of in uenza, even though it is based on the questionnaire. Third, the survey was conducted in adherence with the anonymous principle, and we could not identify each respondent. Although the respondents overlapped among the ve seasons, the overlapping degree varied; this might have caused bias in comparisons among the seasons. Finally, our survey was completed in March 2019, and it was not affected by COVID-19 and related confounding circumstances. No one knows whether the current estimates regarding the in uenza epidemic will be applicable after the COVID-19 pandemic has subsided; this is the same issue for the epidemiology of most infectious diseases.

Conclusions
The present study gives valuable insight into data on the circulation of in uenza and the measurement of three threshold levels for children, which would support management approaches for annual in uenza epidemics in the community. The vaccination timing should be rede ned over time and adapted to each country as more local surveillance data become available [18]. Our ndings based on in uenza surveillance for children are useful for general practitioners, health policymakers, and disease control planners who are concerned with the prevention and control of in uenza. More local surveillance data is expected to be utilized to review and improve the infectious disease prevention in each region.
Declarations Figure 1 Selection of the study population