This paper is in the following e-collection/theme issue:

Reviews on Public Health Technology and Innovation (69) Surveillance Systems (462) Digital Health Reviews (1289) Influenza and Influenza-Like-Illness (ILI) Surveillance (217)

Published on in Vol 9 (2023)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/41329, first published .
Monitoring School Absenteeism for Influenza-Like Illness Surveillance: Systematic Review and Meta-analysis

Monitoring School Absenteeism for Influenza-Like Illness Surveillance: Systematic Review and Meta-analysis

Monitoring School Absenteeism for Influenza-Like Illness Surveillance: Systematic Review and Meta-analysis

Authors of this article:

Tim K Tsang1 Author Orcid Image ;   Xiaotong Huang1 Author Orcid Image ;   Yiyang Guo1 Author Orcid Image ;   Eric H Y Lau1, 2 Author Orcid Image ;   Benjamin J Cowling1, 2 Author Orcid Image ;   Dennis K M Ip1 Author Orcid Image
Article Authors Cited by (5) Tweetations (6) Metrics

    Review

    1WHO Collaborating Centre for Infectious Disease Epidemiology and Control, School of Public Health, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong, China (Hong Kong)

    2Laboratory of Data Discovery for Health Limited, Hong Kong Science and Technology Park, Hong Kong, China (Hong Kong)

    Corresponding Author:

    Tim K Tsang, PhD

    WHO Collaborating Centre for Infectious Disease Epidemiology and Control, School of Public Health

    Li Ka Shing Faculty of Medicine

    The University of Hong Kong

    7 Sassoon Road

    Pokfulam

    Hong Kong

    China (Hong Kong)

    Phone: 852 39179715

    Email: timtsang@connect.hku.hk


    Background: Influenza causes considerable disease burden each year, particularly in children. Monitoring school absenteeism has long been proposed as a surveillance tool of influenza activity in the community, but the practice of school absenteeism could be varying, and the potential of such usage remains unclear.

    Objective: The aim of this paper is to determine the potential of monitoring school absenteeism as a surveillance tool of influenza.

    Methods: We conducted a systematic review of the published literature on the relationship between school absenteeism and influenza activity in the community. We categorized the types of school absenteeism and influenza activity in the community to determine the correlation between these data streams. We also extracted this correlation with different lags in community surveillance to determine the potential of using school absenteeism as a leading indicator of influenza activity.

    Results: Among the 35 identified studies, 22 (63%), 12 (34%), and 8 (23%) studies monitored all-cause, illness-specific, and influenza-like illness (ILI)–specific absents, respectively, and 16 (46%) used quantitative approaches and provided 33 estimates on the temporal correlation between school absenteeism and influenza activity in the community. The pooled estimate of correlation between school absenteeism and community surveillance without lag, with 1-week lag, and with 2-week lag were 0.44 (95% CI 0.34, 0.53), 0.29 (95% CI 0.15, 0.42), and 0.21 (95% CI 0.11, 0.31), respectively. The correlation between influenza activity in the community and ILI-specific absenteeism was higher than that between influenza activity in community all-cause absenteeism. Among the 19 studies that used qualitative approaches, 15 (79%) concluded that school absenteeism was in concordance with, coincided with, or was associated with community surveillance. Of the 35 identified studies, only 6 (17%) attempted to predict influenza activity in the community from school absenteeism surveillance.

    Conclusions: There was a moderate correlation between school absenteeism and influenza activity in the community. The smaller correlation between school absenteeism and community surveillance with lag, compared to without lag, suggested that careful application was required to use school absenteeism as a leading indicator of influenza epidemics. ILI-specific absenteeism could monitor influenza activity more closely, but the required resource or school participation willingness may require careful consideration to weight against the associated costs. Further development is required to use and optimize the use of school absenteeism to predict influenza activity. In particular, the potential of using more advanced statistical models and validation of the predictions should be explored.

    JMIR Public Health Surveill 2023;9:e41329

    doi:10.2196/41329

    Keywords

    influenza; surveillance; school absenteeism; monitoring; school attendance; influenza-like illness; correlation; trend; pattern; predict; prediction; influenza activity; infection; surveillance tolls 


    Influenza virus causes substantial morbidity and mortality in humans each year on average [1,2]. Influenza activity and hence infections decreased due to public health and social measures of COVID-19 [3]. However, rebound of influenza virus activity is expected [4,5], given the relaxation of public health and social measures [6,7] and low population immunity of influenza [8]. Therefore, monitoring influenza activity is important in the post–COVID-19 era. Common modes of surveillance for influenza activity include the following: (1) sentinel surveillance, in which the consultation rates or the number of influenza-like illness (ILI) in outpatient and private medical practitioner clinics is recorded and (2) laboratory surveillance, in which respiratory specimens are collected and the proportion of positive tests of the influenza virus are recorded.

    Children are believed to be a major driver of influenza virus transmission since they have more frequent person-to-person close contacts with low preexisting immunity [9-11]. Therefore, school is a high-risk setting for the transmission of influenza and other respiratory viruses. Preventing transmission in schools could block transmissions to family members and further reduce community epidemics [12]. School surveillance offers an opportunity for the early detection of these viruses [13]. In some regions, student absenteeism surveillance could be in real time, and hence it could have minimal reporting delay, with relative low cost [14,15].

    Some studies explore the potential use of school absenteeism to monitoring the influenza activity in the community. In some regions, school absenteeism has been integrated as a part of routine disease surveillance, particularly for respiratory virus infections [16-21], although there are also usages for other diseases [22,23]. However, the component of school absenteeism surveillance could be different, such as the use of all-cause and illness-specific absenteeism. Moreover, their performances are rarely assessed. Therefore, validating and optimizing the performance of the use of school absenteeism on the surveillance of influenza activity would be critical.

    In this paper, we conducted a systematic review to collect information on studies that described both school absenteeism surveillance and community surveillance. We summarized the types of school absenteeism surveillance and community surveillance methods to assess the relationship between these 2 surveillances, and the potential use of school absenteeism surveillance to predict influenza activity in the community.


    Definition of School Absenteeism Surveillance and Community Surveillance

    School absenteeism surveillance was defined as the time series recording the absent rate per day or week in a school. Absenteeism could be all-cause, specific to ILI, or specific to respiratory illness but not limited to ILI.

    Community surveillance was defined as the time series monitoring the influenza activity in a region. There were the following two main types of surveillance: (1) ILI rate or count per day or week and (2) the proportion of laboratory specimens testing positive for influenza virus.

    Search Strategy and Selection Criteria

    This systematic review was conducted following the PRISMA (Preferred Reporting Items for Systematic Review and Meta-analysis) statement [24]. A standardized search was done in PubMed, Embase, and Web of Science, using the search term “((school AND (absent OR absence OR absenteeism) AND (flu OR influenza))”. The search was done on May 12, 2022, with no language restrictions. Additional relevant articles from the reference sections were also reviewed.

    Two authors (XH and YG) independently screened the titles and extracted data from the included studies. Disagreements were resolved by consensus with a third author (TKT). Studies identified from different databases were deduplicated.

    Eligible articles were those reporting daily or weekly student absenteeism data and community ILI surveillance in the same region. There was no restriction on the methods used for finding the relationship between school absenteeism and influenza surveillance in the community. Studies without both school absenteeism and community surveillance data were excluded. Articles that met the following were also excluded: (1) the study summarized the findings published elsewhere; (2) the study used data generated from simulation or prediction; (3) the study analyzed the combination of data from studies published elsewhere; or (4) the full text was not available.

    Data were extracted from included studies using a standard form, with the following information in following 3 major components: school absenteeism, influenza surveillance in the community, and comparison between school absenteeism and influenza surveillance in community. Each mentioned variable below was a column in the standard form. For school absenteeism, the number of schools, the school size, the grade of students, methods used for school absenteeism surveillance, the types of absenteeism (all-cause, ILI-related, or illness-specific), illness ascertainment methods for illness-related absenteeism, and the treatment of holiday in the analysis were extracted. For influenza surveillance in the community, the extracted data included the types of surveillance (influenza-like illness with or without laboratory test) and the information of laboratory surveillance (influenza strain and the use of test-positive number or proportion). For comparison between school absenteeism and influenza surveillance in the community, information included methods, period and lag time (day/week) of comparison, and estimated correlation coefficients.

    Data Analysis

    For studies reporting regression coefficients, we used the approach by Rodgers and Nicewander [25] to transform them to Pearson correlation coefficient. We conducted random effects meta-analyses using the inverse variance method and restricted maximum likelihood estimator for heterogeneity to obtain the pooled correlation between school absenteeism and community surveillance [26-29]. Cochran Q test and the I2 statistic were used to identify and quantify heterogeneity among included studies [30]. An I2 value of more than 75% indicated high heterogeneity [31]. We conducted subgroup meta-analyses by the types of school absenteeism surveillance (all-cause vs ILI-related vs illness-specific), by the type of community surveillance (ILI vs lab-confirmed), with or without considering the delay effect of school absenteeism on community influenza. Meta-analysis on subgroups was only performed when there were at least 5 estimates. However, it should be noted that as I2 could be biased in small studies, we only quoted the I2 estimate when the number of studies was at least 10 [32]. We also conducted metaregressions to explore the impact of the following factors: types of school absenteeism, types of community surveillance, primary school or lower grade in school absenteeism surveillance, use of count data in school absenteeism surveillance, use of weekly data, and study type (prospective vs retrospective) in the analysis.

    For studies using qualitative methods, in which only plots or tables of time series of school absenteeism and influenza activity in the community were provided without using any statistical comparison such as estimating correlations, a qualitative description of the relationship between these 2 surveillance systems was extracted. The used terms for the relationship in those articles was summarized. For the studies attempting to predict community surveillance by using school absenteeism, the following information about prediction was extracted: the types of data and its time period used for training for the models for prediction, data type and prediction methods, the type of data for validating the developed models and its time period, and the evaluation method.


    Overview

    In the systematic review, we identified 3579 studies in our search, and 1144 duplicated articles were excluded (Figure 1). After screening the titles and the abstracts of the remaining articles, we identified 171 studies that may contain relevant information for full-text screening. Among these, 35 articles (Table S1 in Multimedia Appendix 1) met the inclusion criteria and were included in this review [15-19,33-62], of which 9 [19,33,37,41,46,48,54,61,62] and 6 [17,18,40,42,51,53] were prospective and retrospective studies, respectively. In terms of school absenteeism surveillance, 22 [15,16,34-39,41,47-49,​51,53-57,59-62], 12 [15,17-19,35-37,44,45,50,54,61], and 8 [33,40,42,43,46,52,58,61] of the studies provided counts on all-cause, ILI-related, and illness-specific absenteeism, respectively. In terms of community surveillance, 25 [16-19,33-37,39-45,47,49-51,55,58-60,62] and 26 [15-17,19,33,34,36-43,46-49,52-54,56,57,59-61] studies provided the number of ILI cases and laboratory-confirmed cases, respectively. Moreover, 24 [15,17,19,33,35-38,40-45,​47-49,51,54,56-58,60,62] and 11 [16,18,34,39,46,50,52,​53,55,59,61] studies used weekly and daily scale of time series for analysis, respectively. In total, 7060 (we excluded 5 studies that did not provide this information) schools were included in our study. Of these, 2 articles [16,56] did not provide any comparison between school absenteeism and community surveillance; 16 articles [17-19,33,37,40-42,46,48,​51,53-55,61,62] with 7703 school years performed quantitative analysis by using statistical method to compute the correlation between school absenteeism and community surveillance; and 19 articles [15,16,34-36,38,39,43-45,47,49,50,52,56-60] with 2123 school years performed qualitative analysis comparing their trends visually by figures or plots. In addition, 6 [17,18,42,46,54,61], 5 [17,18,42,46,61], and 1 [37] articles considered time lag and reported comparisons of school absenteeism and community surveillance with 1-week, 2-week, and 3-week lags, respectively. Moreover, 25 studies excluded holidays in the comparison, while 10 studies did not provide any information. Among the 25 studies that excluded holidays in the comparison, 15 of them [16-18,33,37,45,48-53,55,61,62] used weekly absolute count in school absenteeism surveillance, which may not have accounted for the reduction in denominator due to holidays within the week. Overall, among 16 and 19 studies using quantitative and qualitative analysis, 15 and 17 studies supported the association between school absenteeism and community surveillance, respectively (Table 1).

    ‎
    Figure 1. Process of systematic review. ILI: influenza-like illness.
    View this figure
    Table 1. Summary of included studies for pooled analysis (n=35).
    CharacteristicsStudies, nReference
    Methods to assess correlation

    		Quantitative analysis16N/Aa

    		Regression4[33,41,51,62]

    		Correlation11[17-19,37,40,42,46,48,53,54,61]

    		Other measures1[55]

    		Qualitative analysis by comparing of trends in plots of time series19[15,16,34-36,38,39,43-45,47,49,50,52,56-60]
    Time lag between school absent and surveillance

    		0 week14[17-19,33,40-42,46,48,51,53,54,61,62]

    		1 week6[17,18,42,46,54,61]

    		Others1[37]
    Association between school absent and surveillance

    		By quantitative analysis


    		Significance for 0-week lag6[17-19,33,40,46]


    		Significance for 1-week lag2[17,18]


    		Significance for 2-week lag2[17,18]


    		Significance for 3-week lag1[37]


    		Unknown significance8[41,42,51,53-55,61,62]


    		No significant association1[48]

    		By quantitative analysis


    		With association17[16,34-36,38,39,43-45,47,49,50,56-60]


    		No detected association2[15,52]

    aN/A: not applicable.

    Correlation Between School Absenteeism and Community Surveillance

    For quantitative analysis, all studies used regression coefficient or Pearson or Spearman correlation coefficient for evaluating the relationship between school absenteeism and community surveillance, except 1 [63], which used receiver operating characteristic to determine the accuracy of using school absenteeism to predict the occurrence of influenza outbreak, which was not included in our calculation of pooled correlation. Overall, 14 studies [17-19,33,40-42,46,48,51,53,54] provided 33 correlation estimates (Figure 2 [17-19,33,40-42,​46,48,51,53,54,61,62] and Table S2 in Multimedia Appendix 1), and 30 estimates supported a positive correlation between school absenteeism and community surveillance [18,42]. The correlation (without lag) between school absences and community surveillance was 0.44 (95% CI 0.34, 0.53), with high heterogeneity (Figure 3). In terms of community surveillance, the correlation between school absences and community surveillance using laboratory-confirmed cases (correlation: 0.45; 95% CI 0.33, 0.57) was higher than using the number of ILI (correlation: 0.41; 95% CI 0.23, 0.59), which was less specific. In terms of school absenteeism, the correlation between community and ILI-specific absenteeism (correlation: 0.51; 95% CI 0.30, 0.73) and illness-specific absenteeism (correlation: 0.43; 95% CI 0.30, 0.57) was higher than all-cause absenteeism (correlation: 0.36; 95% CI 0.11, 0.62). While there was a positive relationship between school absenteeism and the community surveillance with 1-week (correlation: 0.29; 95% CI 0.15, 0.42) and 2-week lag (correlation: 0.21; 95% CI 0.11, 0.31), the positive correlation was smaller compared with those without lag.

    In the metaregression (Table 2), we estimated that the correlation between school absenteeism and the community surveillance with 1-week and 2-week lag was 0.19 (95% CI 0.04, 0.34) and 0.26 (95% CI 0.10, 0.41) lower than without lag. We found that the correlation between community surveillance and ILI-related absenteeism was 0.25 (95% CI 0.04, 0.45) higher than that between community surveillance and all-cause absenteeism. The correlation between school absenteeism and laboratory-confirmed community surveillance was 0.17 (95% CI 0.04, 0.30) higher than that between school absenteeism and ILI surveillance. We found that the correlation between school absenteeism and community surveillance from prospective studies was 0.19 (95% CI 0.05, 0.33) higher than from retrospective studies.

    ‎
    Figure 2. The temporal correlation between school absenteeism and influenza activity in community from identified studies. ILI: influenza-like illness; PCR: Polymerase Chain Reaction.
    View this figure
    ‎
    Figure 3. The pooled estimate of temporal correlations between school absenteeism and influenza activity in the community by types of school absenteeism surveillance and types of surveillance of influenza activity in community. ILI: influenza-like illness.
    View this figure
    Table 2. Factors affecting the correlation between school absenteeism and community surveillance by metaregression.
    VariablesNo lag (95% CI)1-week lag (95% CI)2-week lag (95% CI)Overall (0- to 2-week lag; 95% CI)
    School absence type

    		All-cause absenteeismReferenceReferenceReferenceReference

    		ILIa-related absenteeism0.20 (–0.11, 0.50)0.39 (–0.11, 0.89)0.21 (–0.16, 0.57)0.25 (0.04, 0.45)

    		Illness-specific absenteeism0.10 (–0.16, 0.35)0.16 (–0.32, 0.64)0.14 (–0.20, 0.47)0.12 (–0.07, 0.30)
    Community surveillance type

    		ILIReferenceReferenceReferenceReference

    		Laboratory confirmed0.08 (–0.14, 0.30)0.30 (0.05, 0.54)0.23 (0.03, 0.43)0.17 (0.04, 0.30)
    Study type

    		Retrospective studyReferenceReferenceReferenceReference

    		Prospective study0.25 (0.04, 0.45)0.12 (–0.20, 0.44)0.07 (–0.19, 0.33)0.19 (0.05, 0.33)
    Lagb

    		No lagN/AcN/AN/AReference

    		1-week lagN/AN/AN/A–0.19 (–0.34, –0.04)

    		2-week lagN/AN/AN/A–0.26 (–0.41, –0.10)
    Only primary school and lower grade in school surveillance–0.10 (–0.53, 0.34)–0.21 (–0.59, 0.16)–0.15 (–0.41, 0.11)–0.14 (–0.36, 0.08)
    Using count in school absenteeism (Reference: using rate or proportion)0.03 (–0.23, 0.29)–0.16 (–0.50, 0.18)0.22 (–0.10, 0.54)0.01 (–0.16, 0.18)
    Using weekly scale (Reference: daily scale)0.11 (–0.15, 0.38)0.09 (–0.22, 0.41)–0.07 (–0.33, 0.19)0.06 (–0.10, 0.22)

    aILI: influenza-like illness.

    bNumber of weeks that school absenteeism data were leading community surveillance.

    cN/A: not applicable.

    Qualitative Analysis for the Relationship School Absenteeism and Community Surveillance

    In the qualitative analysis, there were 19 articles that reported the relationship between school absenteeism and community surveillance by presenting the time series of these surveillances (Table 3). Of these 19 studies, 2 [16,56] did not compare the relationship in plots of time series. Overall, a similar pattern was reported within school absenteeism and community surveillance in all studies (Table 3), except 2, which reported no detected association [15,52]. Moreover, 6 studies [35,36,43,47,49,50] compared the peak of school absent and peak of epidemics. Among these, 4 [35,36,43,50], 1 [49], and 1 [47] studies found that school absenteeism peaked before, at the same time, and after the peak of community surveillance, respectively. In addition, 15 studies concluded that school absenteeism was in concordance with, coincided with, or was associated with community surveillance, using terms including “trends are coincident” [34,47,59], “associated with” [16,34,36,56,57,60], “concordance with” [38,43,58], “similar trend” [39,44,50], or ‘’mirror” [45].

    Table 3. Summary of term count for qualitative results.
    Term combined and term usedStudyComparisonCount
    School absenteeism is in concordance with, coincided with, or is associated with community ILIa16

    		Increases in absences coincided with community-wide ILI outbreaksBesculides [34], 2005All-cause absenteeism vs community influenza isolates1

    		The second wave of school absenteeism coincided with the second round of community ILISchoub [47], 1994All-cause absenteeism vs laboratory isolations1

    		The drop in school absenteeism coincided with the epidemicMcCormick [59], 2010All-cause absenteeism vs community ILI surveillance or virus-confirmed influenza1

    		School absenteeism is associated with community ILI6


    		The extent of school absenteeism is associated with the extent of community ILI peakBesculides [34], 2005All-cause absenteeism vs community influenza isolates


    		School absenteeism is associated with community ILICheng [36], 2012ILI-related absenteeism vs laboratory isolations


    		Influenza activity was reflected by the school absenteeism ratesChin [57], 1974All-cause absenteeism vs laboratory-confirmed influenza


    		Influenza activity was associated with or reflected by school absenteeismRubin [60], 1975All-cause absenteeism vs community ILI surveillance


    		School absenteeism was less timely than laboratory dataChu [16], 2013All-cause absenteeism vs laboratory-confirmed influenza


    		Close temporal correlation between school absenteeism and the isolation of strains of influenza virusOlson [56], 1980All-cause absenteeism vs laboratory-confirmed influenza

    		School absenteeism is concordant with community ILI3


    		Jaeger [38], 2011All-cause absenteeism vs laboratory-confirmed influenza


    		Mook [43], 2007Illness-defined absenteeism vs community ILI surveillance


    		Lenaway [58], 1995ILI-related absenteeism vs community ILI surveillance

    		Similar trend3


    		Janusz [39], 2011All-cause absenteeism vs community ILI surveillance


    		Nasrullah [44], 2012ILI-related absenteeism vs community ILI surveillance


    		Suzue [50], 2012ILI-related absenteeism vs community ILI surveillance

    		School absenteeism mirrored or corresponded to community ILIRead [45], 2021Influenza-confirmed absenteeism vs community virus–confirmed influenza1
    Comparison by peak time6

    		Peak at the same timeSigmundsdottir [49], 2010All-cause absenteeism vs laboratory confirmed cases1

    		School absenteeism peak preceded or was ahead of the epidemic peak4


    		Bollaerts [35], 2010All-cause absenteeism vs community ILI surveillance


    		Cheng [36], 2012ILI-related absenteeism vs laboratory isolations


    		Mook [43], 2007Illness-defined absenteeism vs community ILI surveillance or laboratory isolations


    		Suzue [50], 2012ILI-related absenteeism vs community ILI surveillance

    		Virus isolation commence and peak before the school absenteeismSchoub [47], 1994All-cause absenteeism vs laboratory isolations1
    No detected association2

    		No related outbreaks were detected, and no peaks were foundTan [52], 2014ILI-related absenteeism vs laboratory-confirmed cases1

    		Within seasons, cases peaked in winter, whereas county-level absences varied throughout the yearQuandelacy [15], 2021ILI-related absenteeism or all-cause absenteeism vs community virus–confirmed influenza1

    aILI: influenza-like illness.

    Prediction of Community Surveillance From School Surveillance

    Six studies attempted to predict community ILI from school absenteeism (Table S3 in Multimedia Appendix 1), 3 of which [15,41,62] predicted the influenza rate (case number), and 3 [18,53,55] predicted the occurrence of outbreaks. All the studies agreed that school absenteeism surveillance was of good use for influenza outbreak detection [15,18,41,53,62], except 1 using all-cause absenteeism [55]. Among these, 4 studies [15,53,55,62] and 2 studies [18,41] used the all-cause absenteeism and ILI-related absenteeism for prediction, respectively. Moreover, 1 [18] and 5 [15,41,53,55,62] studies used mechanistic models (SEIR) and statistical models, respectively, 3 of which [41,53,62] considered the autocorrelation of the time series. In term of evaluation, those studies that predicted the occurrence of outbreaks used receiver operating characteristic curves [18,55] and false-positive rate (1 minus sensitivity) [53]. For those studies that predicted the number of cases, 1 used the model fit [41], 1 used mean absolute error [15], and 1 used Akaike Information Criterion value [62]. Two articles [15,53] conducted out-sample testing to evaluate the performance.


    Principal Findings

    In this study, we summarized the practice of school absenteeism surveillance and their potential use on monitoring and predicting influenza activity in the community. Studies could broadly be classified as quantitative and qualitative. In quantitative studies, most studies used temporal correlation to quantify the relationship between school absenteeism and influenza activity in the community. In qualitative studies, time series between these 2 data streams were compared visually. Overall, we found a moderate correlation between school absenteeism and influenza activity in the community based on quantitative studies. This suggested the potential to use school absenteeism data to monitor or predict influenza activity in the community.

    The measure of school absence could be classified to nonspecific (all-cause absent) and specific (illness- or ILI-specific absence). We found that ILI-specific absenteeism had a higher correlation with the community surveillance, compared with all-cause absenteeism, which was consistent with the findings from another review [64]. While using specific measure of school absenteeism could be slightly more accurate [54], the implementation of using more specific absenteeism surveillance in schools should be considered jointly with the associated costs. Those specific measures may require more resources to obtain, such as follow-up of the reasons of absence, which may also jeopardize the timeliness of the school absence data. Moreover, the willingness of school participation may be lower due to the higher requirement of resource; hence the sample size may decrease. Such factors should be considered when deciding the use of all-cause absence or ILI-specific absence to monitor influenza activity. The actual cost of implementation of using more specific absenteeism surveillance in schools was likely different by regions and countries; therefore, our studies could not provide recommendation on this.

    A number of studies proposed that one of the values of school absenteeism surveillance was its lead on the traditional surveillance [35,36,43,50]. However, we found that the correlation between school absenteeism and community surveillance without lag was the highest. In addition, the correlation between school absenteeism and 1-week lag of community surveillance was only marginally smaller than community surveillance without lag. Hence, the school absenteeism surveillance could at most lead influenza outbreaks by 1 week. Therefore, despite the moderate correlation between school absenteeism and community surveillance, the use of school absenteeism as a surveillance tool may require further exploration and development of methodology. In particular, we found that the correlations between school absenteeism and community surveillance from prospective studies were higher than from retrospective studies, suggesting that prospective data collection could improve the accuracy by carefully checking the data with timely correction.

    More than half of the studies only use qualitative approaches to explore the relationship between school absenteeism and influenza activity in the community, in which only figures of time series of these 2 surveillances were compared visually. Such approaches were relatively subjective, and further quantitative comparison should be performed. Furthermore, almost all quantitative comparisons were based on temporal correlation, which required further steps to determine the usefulness of school absenteeism for monitoring or predicting influenza activity in the community, such as the development of prediction or forecasting tools.

    Only 6 out of 35 studies attempted to develop methods to use the school absenteeism data to predicted influenza activity in the community. However, those applied approaches were suboptimal compared with other influenza-forecasting approaches [65,66]; particularly, the potential of more advanced statistical approaches was less explored, as well as the lack of validation. In terms of modeling fitting, only 2 out of 6 studies conducted out-sample evaluation on their forecasting approach to avoid overfitting. In terms of model evaluation, all studies evaluated the point forecast, but none of them evaluated the prediction intervals from those models. The evaluation of interval forecast by some proper score rules would also be important to use the prediction performance [66,67] and to support the accuracy of the predictions. In future studies, using school absenteeism data in more well-developed models should be explored, including mechanistic models (ie, SIR-type) [65,66], and statistical models, such as generalized additive model and random forecast regression, which could particularly handle nonlinear relationship [68].

    Limitations

    Our study may have some limitations. First, we did not summarize the change of rule on the school attending in the identified studies. For example, many regions may not allow students to attend school when they have fever, particularly during or after the 2009 pandemic influenza outbreaks. Therefore, our study may not be able to evaluate this impact on the relationship between school absenteeism and community surveillance. Second, we did not stratify the analysis by type of reported correlation coefficients (Pearson or Spearman). However, we expected that the direction of correlation should be the same. Third, most studies did not report information of burden or costs of implementing school surveillance. Therefore, it was impossible for us to determine the real-life impact or cost-effectiveness of different types of surveillance. Finally, we could not rule out other potential cofounders in the identified relationship between school absenteeism and influenza activity in the community. For example, information of school or class size was not available.

    Conclusion

    We found there was a moderate correlation between school absenteeism and influenza activity in the community. We found that the correlation between influenza activity and ILI-specific absenteeism was higher than all-cause absenteeism. However, implementing more specific surveillance in school may require careful consideration, since more resources may be required, and it may have a negative impact on the willingness of school participation. There was potential for using school absenteeism as a surveillance and prediction tool of influenza activity. A further development of methodology was required to use and optimize such usage. In particular, more statistical models should be explored, and the validation of prediction performance is missed in most studies.

    Acknowledgments

    We thank Can Wang for technical assistance. This study was supported by the Theme-based Research Scheme project no. T11-712/19N from the Hong Kong Government, the Health and Medical Research Fund, Food and Health Bureau, Government of the Hong Kong Special Administrative Region (grant no. 21200292 and 11101092).

    Conflicts of Interest

    BJC reports honoraria from Sanofi Pasteur, GSK, Moderna, and Roche. The authors report no other potential conflicts of interest.

    ‎
    Multimedia Appendix 1

    Supplementary tables.

    DOCX File , 36 KB
    1. Dawood FS, Iuliano AD, Reed C, Meltzer MI, Shay DK, Cheng P, et al. Estimated global mortality associated with the first 12 months of 2009 pandemic influenza A H1N1 virus circulation: a modelling study. Lancet Infect Dis 2012 Sep;12(9):687-695. [CrossRef] [Medline]
    2. Iuliano AD, Roguski KM, Chang HH, Muscatello DJ, Palekar R, Tempia S, Global Seasonal Influenza-associated Mortality Collaborator Network. Estimates of global seasonal influenza-associated respiratory mortality: a modelling study. Lancet 2018 Mar 31;391(10127):1285-1300 [FREE Full text] [CrossRef] [Medline]
    3. Cowling BJ, Ali ST, Ng TWY, Tsang TK, Li JCM, Fong MW, et al. Impact assessment of non-pharmaceutical interventions against coronavirus disease 2019 and influenza in Hong Kong: an observational study. Lancet Public Health 2020 May;5(5):e279-e288 [FREE Full text] [CrossRef] [Medline]
    4. Wu K, Wu X, Wang W, Hong L. Epidemiology of influenza under the coronavirus disease 2019 pandemic in Nanjing, China. J Med Virol 2022 May;94(5):1959-1966 [FREE Full text] [CrossRef] [Medline]
    5. Ali ST, Lau YC, Shan S, Ryu S, Du Z, Wang L, et al. Prediction of upcoming global infection burden of influenza seasons after relaxation of public health and social measures during the COVID-19 pandemic: a modelling study. Lancet Glob Health 2022 Nov;10(11):e1612-e1622 [FREE Full text] [CrossRef] [Medline]
    6. Sovann LY, Sar B, Kab V, Yann S, Kinzer M, Raftery P, et al. An influenza A (H3N2) virus outbreak in the Kingdom of Cambodia during the COVID-19 pandemic of 2020. Int J Infect Dis 2021 Feb;103:352-357 [FREE Full text] [CrossRef] [Medline]
    7. Zheng L, Qi J, Wu J, Zheng M. Changes in Influenza Activity and Circulating Subtypes During the COVID-19 Outbreak in China. Front Med (Lausanne) 2022;9:829799 [FREE Full text] [CrossRef] [Medline]
    8. Lee K, Jalal H, Raviotta JM, Krauland MG, Zimmerman RK, Burke DS, et al. Estimating the Impact of Low Influenza Activity in 2020 on Population Immunity and Future Influenza Seasons in the United States. Open Forum Infect Dis 2022 Jan;9(1):ofab607 [FREE Full text] [CrossRef] [Medline]
    9. Mossong J, Hens N, Jit M, Beutels P, Auranen K, Mikolajczyk R, et al. Social contacts and mixing patterns relevant to the spread of infectious diseases. PLoS Med 2008 Mar 25;5(3):e74 [FREE Full text] [CrossRef] [Medline]
    10. Ranjeva S, Subramanian R, Fang VJ, Leung GM, Ip DKM, Perera RAPM, et al. Age-specific differences in the dynamics of protective immunity to influenza. Nat Commun 2019 Apr 10;10(1):1660 [FREE Full text] [CrossRef] [Medline]
    11. Kwok KO, Cowling BJ, Wei VWI, Wu KM, Read JM, Lessler J, et al. Social contacts and the locations in which they occur as risk factors for influenza infection. Proc Biol Sci 2014 Aug 22;281(1789):20140709 [FREE Full text] [CrossRef] [Medline]
    12. Tsang TK, Lau LLH, Cauchemez S, Cowling BJ. Household Transmission of Influenza Virus. Trends Microbiol 2016 Feb;24(2):123-133 [FREE Full text] [CrossRef] [Medline]
    13. Yeng PK, Woldaregay AZ, Solvoll T, Hartvigsen G. Cluster Detection Mechanisms for Syndromic Surveillance Systems: Systematic Review and Framework Development. JMIR Public Health Surveill 2020 May 26;6(2):e11512 [FREE Full text] [CrossRef] [Medline]
    14. Worby CJ, Chaves SS, Wallinga J, Lipsitch M, Finelli L, Goldstein E. On the relative role of different age groups in influenza epidemics. Epidemics 2015 Dec;13:10-16 [FREE Full text] [CrossRef] [Medline]
    15. Quandelacy TM, Zimmer S, Lessler J, Vukotich C, Bieltz R, Grantz KH, et al. Predicting virologically confirmed influenza using school absences in Allegheny County, Pennsylvania, USA during the 2007-2015 influenza seasons. Influenza Other Respir Viruses 2021 Nov;15(6):757-766 [FREE Full text] [CrossRef] [Medline]
    16. Chu A, Savage R, Whelan M, Rosella LC, Crowcroft NS, Willison D, et al. Assessing the relative timeliness of Ontario's syndromic surveillance systems for early detection of the 2009 influenza H1N1 pandemic waves. Can J Public Health 2013 May 14;104(4):340-347 [FREE Full text] [CrossRef] [Medline]
    17. Dong X, Boulton ML, Carlson B, Montgomery JP, Wells EV. Syndromic surveillance for influenza in Tianjin, China: 2013-14. J Public Health (Oxf) 2017 Jun 01;39(2):274-281. [CrossRef] [Medline]
    18. Fan Y, Wang Y, Jiang H, Yang W, Yu M, Yan W, et al. Evaluation of outbreak detection performance using multi-stream syndromic surveillance for influenza-like illness in rural Hubei Province, China: a temporal simulation model based on healthcare-seeking behaviors. PLoS One 2014 Nov 19;9(11):e112255 [FREE Full text] [CrossRef] [Medline]
    19. Kom Mogto CA, De Serres G, Douville Fradet M, Lebel G, Toutant S, Gilca R, et al. School absenteeism as an adjunct surveillance indicator: experience during the second wave of the 2009 H1N1 pandemic in Quebec, Canada. PLoS One 2012;7(3):e34084 [FREE Full text] [CrossRef] [Medline]
    20. Koppeschaar CE, Colizza V, Guerrisi C, Turbelin C, Duggan J, Edmunds WJ, et al. Influenzanet: Citizens Among 10 Countries Collaborating to Monitor Influenza in Europe. JMIR Public Health Surveill 2017 Sep 19;3(3):e66 [FREE Full text] [CrossRef] [Medline]
    21. Donaldson AL, Harris JP, Vivancos R, Hungerford D, Hall I, O'Brien SJ. School Attendance Registers for the Syndromic Surveillance of Infectious Intestinal Disease in UK Children: Protocol for a Retrospective Analysis. JMIR Res Protoc 2022 Jan 20;11(1):e30078 [FREE Full text] [CrossRef] [Medline]
    22. Adnan M, Gao X, Bai X, Newbern E, Sherwood J, Jones N, et al. Potential Early Identification of a Large Campylobacter Outbreak Using Alternative Surveillance Data Sources: Autoregressive Modelling and Spatiotemporal Clustering. JMIR Public Health Surveill 2020 Sep 17;6(3):e18281 [FREE Full text] [CrossRef] [Medline]
    23. Lawpoolsri S, Khamsiriwatchara A, Liulark W, Taweeseneepitch K, Sangvichean A, Thongprarong W, et al. Real-time monitoring of school absenteeism to enhance disease surveillance: a pilot study of a mobile electronic reporting system. JMIR Mhealth Uhealth 2014 May 12;2(2):e22 [FREE Full text] [CrossRef] [Medline]
    24. Moher D, Liberati A, Tetzlaff J, Altman DG, PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. BMJ 2009 Jul 21;339:b2535 [FREE Full text] [CrossRef] [Medline]
    25. Lee Rodgers J, Nicewander WA. Thirteen Ways to Look at the Correlation Coefficient. The American Statistician 1988 Feb;42(1):59-66. [CrossRef]
    26. Hedges LV, Vevea JL. Fixed- and random-effects models in meta-analysis. Psychological Methods 1998 Dec;3(4):486-504. [CrossRef]
    27. Veroniki AA, Jackson D, Bender R, Kuss O, Langan D, Higgins JPT, et al. Methods to calculate uncertainty in the estimated overall effect size from a random-effects meta-analysis. Res Synth Methods 2019 Mar;10(1):23-43 [FREE Full text] [CrossRef] [Medline]
    28. Thompson SG, Sharp SJ. Explaining heterogeneity in meta-analysis: a comparison of methods. Stat Med 1999 Oct 30;18(20):2693-2708. [CrossRef] [Medline]
    29. Langan D, Higgins JPT, Jackson D, Bowden J, Veroniki AA, Kontopantelis E, et al. A comparison of heterogeneity variance estimators in simulated random-effects meta-analyses. Res Synth Methods 2019 Mar;10(1):83-98 [FREE Full text] [CrossRef] [Medline]
    30. Cochran WG. The Combination of Estimates from Different Experiments. Biometrics 1954 Mar;10(1):101. [CrossRef]
    31. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ 2003 Sep 06;327(7414):557-560 [FREE Full text] [CrossRef] [Medline]
    32. von Hippel PT. The heterogeneity statistic I(2) can be biased in small meta-analyses. BMC Med Res Methodol 2015 Apr 14;15(1):35 [FREE Full text] [CrossRef] [Medline]
    33. Aldridge RW, Hayward AC, Field N, Warren-Gash C, Smith C, Pebody R, Decipher my Data projectschools. Are School Absences Correlated with Influenza Surveillance Data in England? Results from Decipher My Data-A Research Project Conducted through Scientific Engagement with Schools. PLoS One 2016 Mar 2;11(3):e0146964 [FREE Full text] [CrossRef] [Medline]
    34. Besculides M, Heffernan R, Mostashari F, Weiss D. Evaluation of school absenteeism data for early outbreak detection, New York City. BMC Public Health 2005 Oct 07;5(1):105 [FREE Full text] [CrossRef] [Medline]
    35. Bollaerts K, Antoine J, Robesyn E, Van Proeyen L, Vomberg J, Feys E, et al. Timeliness of syndromic influenza surveillance through work and school absenteeism. Arch Public Health 2010 Sep 29;68(3):115-120. [CrossRef]
    36. Cheng CKY, Cowling BJ, Lau EHY, Ho LM, Leung GM, Ip DKM. Electronic school absenteeism monitoring and influenza surveillance, Hong Kong. Emerg Infect Dis 2012 May;18(5):885-887 [FREE Full text] [CrossRef] [Medline]
    37. Ip DK, Lau EH, So HC, Xiao J, Lam CK, Fang VJ, et al. A Smart Card-Based Electronic School Absenteeism System for Influenza-Like Illness Surveillance in Hong Kong: Design, Implementation, and Feasibility Assessment. JMIR Public Health Surveill 2017 Oct 06;3(4):e67 [FREE Full text] [CrossRef] [Medline]
    38. Jaeger V, Shick-Porter M, Moore D, Grant D, Wolfe V. GotFlu channel: an online syndromic surveillance tool supporting college health practice and public health work. J Am Coll Health 2011 Apr 13;59(5):415-418. [CrossRef] [Medline]
    39. Janusz KB, Cortes JE, Serdarevic F, Jones RC, Jones JD, Ritger KA, Chicago H1N1 Epidemiology Team. Influenza-like illness in a community surrounding a school-based outbreak of 2009 pandemic influenza A (H1N1) virus-Chicago, Illinois, 2009. Clin Infect Dis 2011 Jan 01;52(Suppl 1):S94-101 [FREE Full text] [CrossRef] [Medline]
    40. Kar EO, Elliot AJ, Bagnall H, Foord DGF, Pnaiser R, Osman H, et al. Absenteeism in schools during the 2009 influenza A(H1N1) pandemic: a useful tool for early detection of influenza activity in the community? Epidemiol. Infect 2011 Oct 21;140(7):1328-1336. [CrossRef]
    41. Lau EHY, Cheng CKY, Ip DKM, Cowling BJ. Situational awareness of influenza activity based on multiple streams of surveillance data using multivariate dynamic linear model. PLoS One 2012;7(5):e38346 [FREE Full text] [CrossRef] [Medline]
    42. Ma T, Englund H, Bjelkmar P, Wallensten A, Hulth A. Syndromic surveillance of influenza activity in Sweden: an evaluation of three tools. Epidemiol. Infect 2014 Dec 04;143(11):2390-2398. [CrossRef]
    43. Mook P, Joseph C, Gates P, Phin N. Pilot scheme for monitoring sickness absence in schools during the 2006/07 winter in England: can these data be used as a proxy for influenza activity? Euro Surveill 2007 Dec 01;12(12):E11-E12. [CrossRef] [Medline]
    44. Nasrullah M, Breiding MJ, Smith W, McCullum I, Soetebier K, Liang JL, et al. Response to 2009 pandemic influenza A H1N1 among public schools of Georgia, United States--fall 2009. Int J Infect Dis 2012 May;16(5):e382-e390 [FREE Full text] [CrossRef] [Medline]
    45. Read JM, Zimmer S, Vukotich C, Schweizer ML, Galloway D, Lingle C, et al. Influenza and other respiratory viral infections associated with absence from school among schoolchildren in Pittsburgh, Pennsylvania, USA: a cohort study. BMC Infect Dis 2021 Mar 22;21(1):291 [FREE Full text] [CrossRef] [Medline]
    46. Schmidt W, Pebody R, Mangtani P. School absence data for influenza surveillance: A pilot study in the United Kingdom. Eurosurveillance 2010;15(3):1-6. [CrossRef]
    47. Schoub BD, Johnson S, McAnerney J, Blackburn NK. Benefits and limitations of the Witwatersrand influenza and acute respiratory infections surveillance programme. S Afr Med J 1994 Oct;84(10):674-678. [Medline]
    48. Short VL, Marriott CK, Ostroff S, Waller K. Description and Evaluation of the 2009–2010 Pennsylvania Influenza Sentinel School Monitoring System. Am J Public Health 2011 Nov;101(11):2178-2183. [CrossRef]
    49. Sigmundsdottir G, Gudnason T, Ólafsson Ö, Baldvinsdottir G, Atladottir A, Löve A, et al. Surveillance of influenza in Iceland during the 2009 pandemic. Euro Surveill 2010 Dec 09;15(49):3 [FREE Full text] [CrossRef] [Medline]
    50. Suzue T, Hoshikawa Y, Nishihara S, Fujikawa A, Miyatake N, Sakano N, et al. The new school absentees reporting system for pandemic influenza A/H1N1 2009 infection in Japan. PLoS One 2012;7(2):e30639 [FREE Full text] [CrossRef] [Medline]
    51. Dai Y, Luo K, Wang X. Thermoelectric properties of graphene-like nanoribbon studied from the perspective of symmetry. Sci Rep 2020 Jun 04;10(1):9105 [FREE Full text] [CrossRef] [Medline]
    52. Tan L, Cheng L, Yan W, Zhang J, Xu B, Diwan V, et al. Using daily syndrome-specific absence data for early detection of school outbreaks: a pilot study in rural China. Public Health 2014 Sep;128(9):792-798. [CrossRef] [Medline]
    53. Ward MA, Stanley A, Deeth LE, Deardon R, Feng Z, Trotz-Williams LA. Methods for detecting seasonal influenza epidemics using a school absenteeism surveillance system. BMC Public Health 2019 Sep 05;19(1):1232 [FREE Full text] [CrossRef] [Medline]
    54. Williams NJ, Ghosh TS, Bisgard KM, Vogt RL. Comparison of 3 school-based influenza surveillance indicators: lessons learned from 2009 pandemic influenza A (H1N1)--Denver Metropolitan Region, Colorado. J Public Health Manag Pract 2013;19(2):119-125. [CrossRef] [Medline]
    55. Egger JR, Hoen AG, Brownstein JS, Buckeridge DL, Olson DR, Konty KJ. Usefulness of school absenteeism data for predicting influenza outbreaks, United States. Emerg Infect Dis 2012 Aug;18(8):1375-1377 [FREE Full text] [CrossRef] [Medline]
    56. Olson JG. School absenteeism during an outbreak of B/Hong Kong/5/72-like influenza virus in Taipei, Taiwan. Southeast Asian Journal of Tropical Medicine and Public Health 1980;11(4):429-434.
    57. Chin J, Magoffin RL, Lennette EH. The epidemiology of influenza in California, 1968-1973. West J Med 1974 Aug;121(2):94-99 [FREE Full text] [Medline]
    58. Lenaway DD, Ambler A. Evaluation of a school-based influenza surveillance system. Public Health Rep 1995;110(3):333-337 [FREE Full text] [Medline]
    59. McCormick JB, Yan C, Ballou J, Salinas Y, Reininger B, Gay J, et al. Response to H1N1 in a U.S.-Mexico border community. Biosecur Bioterror 2010 Sep;8(3):233-242 [FREE Full text] [CrossRef] [Medline]
    60. Rubin RJ, Gregg MB. Influenza surveillance in the United States 1972-1974. Am J Epidemiol 1975 Sep;102(3):225-232. [CrossRef] [Medline]
    61. Temte JL, Barlow S, Goss M, Temte E, Schemmel A, Bell C, et al. Cause-specific student absenteeism monitoring in K-12 schools for detection of increased influenza activity in the surrounding community-Dane County, Wisconsin, 2014-2020. PLoS One 2022;17(4):e0267111 [FREE Full text] [CrossRef] [Medline]
    62. Weiss Z. Detecting the Onset of Infectious Disease Outbreaks Using School Sign-out Logs. Epidemiology 2019 May;30(3):e18-e19. [CrossRef] [Medline]
    63. Egger J, Konty K, Wilson E, Karpati A, Matte T, Weiss D, et al. The effect of school dismissal on rates of influenza-like illness in New York City schools during the spring 2009 novel H1N1 outbreak. J Sch Health 2012 Mar;82(3):123-130. [CrossRef] [Medline]
    64. Donaldson AL, Hardstaff JL, Harris JP, Vivancos R, O'Brien SJ. School-based surveillance of acute infectious disease in children: a systematic review. BMC Infect Dis 2021 Aug 03;21(1):744 [FREE Full text] [CrossRef] [Medline]
    65. Reich NG, McGowan CJ, Yamana TK, Tushar A, Ray EL, Osthus D, et al. Accuracy of real-time multi-model ensemble forecasts for seasonal influenza in the U.S. PLoS Comput Biol 2019 Nov;15(11):e1007486 [FREE Full text] [CrossRef] [Medline]
    66. Ray EL, Reich NG. Prediction of infectious disease epidemics via weighted density ensembles. PLoS Comput Biol 2018 Feb;14(2):e1005910 [FREE Full text] [CrossRef] [Medline]
    67. Bracher J, Ray EL, Gneiting T, Reich NG. Evaluating epidemic forecasts in an interval format. PLoS Comput Biol 2021 Feb;17(2):e1008618 [FREE Full text] [CrossRef] [Medline]
    68. Paireau J, Andronico A, Hozé N, Layan M, Crépey P, Roumagnac A, et al. An ensemble model based on early predictors to forecast COVID-19 health care demand in France. Proc Natl Acad Sci U S A 2022 May 03;119(18):e2103302119 [FREE Full text] [CrossRef] [Medline]

    ‎
    ILI: influenza-like illness
    PRISMA: Preferred Reporting Items for Systematic Review and Meta-analysis

    Edited by A Mavragani; submitted 22.07.22; peer-reviewed by Y Katayama, F Mallor, B Smarr; comments to author 21.10.22; revised version received 04.11.22; accepted 25.11.22; published 11.01.23

    Copyright

    ©Tim K Tsang, Xiaotong Huang, Yiyang Guo, Eric H Y Lau, Benjamin J Cowling, Dennis K M Ip. Originally published in JMIR Public Health and Surveillance (https://publichealth.jmir.org), 11.01.2023.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on https://publichealth.jmir.org, as well as this copyright and license information must be included.