Risk assessment of communicable respiratory diseases transmission based on social contact networks: a primary school contact data survey conducted with portable high-precision devices

Abstract

Background

During the 2020 COVID-19 pandemic, class suspension and school closures, as non-pharmacological interventions, effectively curbed on-campus communicable diseases transmission by minimizing contact. Targeted temporary measures taken for high-risk groups and activities, such as suspending a certain group activity and isolating students with symptoms at home, can significantly reduce transmission without the need for a complete suspension. However, there is currently a lack of in-depth analysis of teacher and student contact behavior. The aim is to assess the risk of communicable respiratory diseases transmission among different populations and activities in primary school.

Methods

We utilized Ultra-Wideband (UWB) wearable devices – a wireless positioning technology enabling centimeter-level proximity detection to record 143,328 close contacts among 292 teachers and students in a primary school throughout the day. By converting data into a network matrix, we constructed a dynamic contact network for the corresponding analysis and calculated network indicators. We also analyzed the contact patterns and distribution of contagion risk among different groups and activity types of contacts.

Results

Network analysis reveals that among the top 10% of individuals in terms of closeness centrality and eigenvector centrality, the proportion of younger students was the highest. Additionally,the duration of single contact was longer among younger students. High-risk contacts predominantly occurred during classes, after-school service, large recess activity and lunch break. The risk of contact was highest among teachers and students of the same grade, with an average contact duration of 57.46 and 56.17 min, while inter-grade durations were only 4.15 and 1.17 min. Moreover, same-grade teachers and students had higher proportion of medium–high risk and above interactions. Inter-grade contacts primarily occurred among students (38.99% medium–high-risk interactions), with 90.45% concentrated during morning self-study and after-school services.

Conclusions

The identification of high-risk contact activities and populations on campus provides important quantitative parameters for targeted intervention particular during the early stage of outbreak.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-24327-2.

Introduction

Communicable respiratory diseases like influenza, chickenpox and mycoplasma pneumonia tend to occur in school-age children [1–3]. In China, the infection rate of seasonal influenza (31%) and A (H3N2) subtype (17%) in children significantly exceeds those in adults (21% and 10%) [4]. Over 90% of reported influenza outbreaks annually occur in schools [5], with primary schools being the most affected [6]. Despite measures like class suspensions and free influenza vaccinations, the incidence of influenza among school-age children is still on the rise [7]. In addition, possibly influenced by the"immunity gap"from the zero-COVID policy adjustment, respiratory diseases caused by adenovirus, influenza virus, respiratory syncytial virus (RSV) and mycoplasma pneumoniae in Chinese children have surged [8], producing surging demand for medical services. Most governments temporarily closed schools at various times during the SARS-Cov-2 pandemic, impacting > 90% of students worldwide [9]. Targeted and temporary measures can be implemented for high-risk population and activities reducing the need for full school closure or suspension, so in-depth studies on school children were called for.

It is widely known that communicable diseases like influenza, COVID-19 spread through droplets and contact, with rapid transmission associated with social activities. Droplet organisms, such as influenza, typically transmit through large respiratory secretions within a distance of less than 2 m [10]. Studying students'close contact on campus is crucial for constructing communicable diseases transmission models and formulating effective epidemic prevention policies based on social contact adjustments [11]. However, previous studies relying on diaries, questionnaires, or positioning devices with low accuracy have explored social contact patterns and network structures among students but often overlooked variations in individual contact transmission risk due to duration and distance [12]. A recent study using high-precision wearable devices collected close contact data for 24 Chinese students, revealing the relationship between close contact behavior and airborne transmission of COVID-19 in classrooms [13]. Therefore, our study analyzes contact patterns, various social networks structural characteristics, and assesses contact risks for different groups and activity types based on reliable contact data (< 2 m) from students and teachers throughout the day on campus collected via UWB wearable devices.

Methods

Data collection

For this research, we chose Tb 1000 g UWB Tag, a wearable device from TSINGOAL, for recording close contact behavior and used it in a sample field trial in school (Appendix Figure A1). The sensor has a communication range of more than 5 m, an alarm distance range of 1 m ~ 3 m, a ranging precision up to 10 cm, and can work for an average of 12 h at the default 1 Hz refresh rate setting. We set the refresh rate of the sensor to 0.25 Hz, the detection range to 2 m, and collected the close contact information of school personnel, including tag ID, duration, distance and time.

The survey was conducted on May 30, 2023, at Shawan Primary School in Longgang District, Shenzhen, China. In 2021, Longgang District had a year-end permanent population of 4,015,300, accounting for 22.71% of Shenzhen's total population and establishing it as the city's second-largest administrative division. This School has 24 classes with 1,164 students from grades 1 to 6. Cluster sampling was employed in this study. The sample size calculation indicated that a minimum of 5 clusters were required. Given the younger age and lower compliance of students in Grades 1 and 2, which made wearing the devices impractical, the final selection, adhering to the cluster sampling principle, involved choosing two classes from each of Grades 3, 4, and 5, resulting in a total of six classes. The study population thus comprised 286 students(151 males and 136 females) and six teachers (Appendix Table A1). Students inclusion criteria: (1) enrolled elementary schoolchildren willing to participate; (2) without mobility impairments or disabilities; (3) on-campus during the survey period. Students exclusion criteria: (1) off-campus during the survey; (2) mobility impairments or disabilities; (3) parental consent not obtained. Teacher inclusion criteria: (1) teachers conducting instructional duties on-site with willingness to participate; (2) on-campus during the survey period. Teacher exclusion criteria: (1) absent during the survey; (2) unwilling to participate. The school day runs from 8:20 AM to 5:30 PM, with a 30-min large recess activity around 9:00 AM, a lunch break from 11:50 PM to 2:20 PM, and three recesses of 10–15 min around 10:10 AM, 11:00 AM and 3:15 PM. Before 8:20 AM and after 3:55 PM are morning self-study and after-school service respectively. During large recess activity, all students go down to the playground and do physical exercise (broadcast gymnastics, running, etc.) in an orderly queue. Lunches are served in a common canteen, and a shared playground is located outside the main building, but some classes have meals in class. 74 students went home for lunch on that day. In the morning, we put devices on students at the door of the classroom in the order they arrive at school. At the end of school at noon, return to school in the afternoon, and at the end of school in the afternoon, students are required to collect and re-wear their devices at a designated location, with a distance of 2.5 m from other students to ensure that the contact before the student's recovery is disconnected and recorded. We recorded the time when students left school and returned to school at noon to facilitate data screening.

Contact patterns

Through the base station, contact data in the recovered tags was uploaded to the terminal. Firstly, the data was exported to EXCEL for storage. R 4.3.0 is used to de-duplicate, clean and sort data. The contact database without names was generated by matching the corresponding grade, class and gender of the contacts according to the ID number. Secondly, the cumulative contact times, cumulative contact duration and average single contact time of different contact pairs and individuals were sorted out. To compare the contact differences between different genders, grades and teachers and students, the Shapiro–Wilk test and Bartlett's test in R 4.3.0 were employed to assess the normal distribution and homogeneity of variance, respectively. The Kruskal–Wallis test and Mann–Whitney test were conducted to determine if there were statistically significant differences between groups. In addition, the results are visually presented in the boxplot. Finally, for comparing the difference between the number of contacts and contact duration within and between grades, the contact pair data was transformed into matrix, and the contact data among different grades in the matrix were accumulated, presenting by heatmap.

Social network analysis

In the static weighted network, nodes represent individuals, and the contacts between individuals are regarded as edges, and the weight of connected edges represent the cumulative contact duration of the contact pairs [14]. Based on the school schedule and contact time, we categorized all interactions into six activities: morning self-study, class, recess, large recess activities, lunch break, and after-school services. Gephi (version 0.10.1) software was used for generating and visually exploring networks, as well as calculating the characteristics indicators of nodes and networks. We employed weighted degree, closeness centrality, betweenness centrality and eigenvector centrality to delineate the fundamental network positioning of nodes. Six indices—average degree, density, diameter, average distance, average clustering coefficient, and modularity—were employed to characterize the structural attributes of both the overall network and six distinct activities networks. These metrics provide insights into the degree and distance of connections among nodes, the localized connectivity among neighboring nodes, and the community division within the network (Appendix Table A2).

Risk levels

Based on our research needs and referring to a study [15] that measured close contact at summer camps, we defined four different categories of risk of infection for contacts based on proximity and duration in order to assess the risk of transmission of communicable respiratory diseases for different groups and different activity types of contacts (Table 1).

Table 1.

Classification of risk levels of contacts

The first category is associated with a high risk of contagion and includes all contacts with duration above 15 min and distance less than 0.1 m. The second category, medium–high risk, includes all contacts with duration above 10 min and distance below 1 m that are not included in the high-risk category. The third category, medium–low risk, includes contacts with duration above 5 min and distance below 1.5 m not included in the previous categories. The fourth category contains all remaining contacts, therefore associated to a low risk level.

Results

Contact patterns

Our study recorded 143,328 contact events among elementary school participants. Key findings include:the number of contacts per individual ranged from 213 to 1,878;The mean daily contact frequency was 981.7 (SD = 256.14), following a normal distribution;the number of unique contacts per participant ranged from 40 to 91, with a median of 58 (IQR: 58–64), showing a right-skewed distribution;single-contact duration ranged from 1 to 12,544 s, with a median of 39 s (IQR: 16–106), consistent with a power-law distribution;most interactions were brief:61.73% lasted ≤ 1 min;93.44% lasted ≤ 10 min; Only 0.31% exceeded 1 h. (Fig. 1 A–C).

Fig. 1.

Distributions of contact patterns. A Number of contacts. B Number of contacted individuals. C Number of contact duration (s)

The analysis indicates differences in contact patterns observed among distinct groups categorized by gender, grade, and activity status of participants (Fig. 2 A–B). For example, males exhibit a significantly higher median number of contacts, Reaching 1043, showcasing a difference of 114 contacts compared to females (P = 0.016). In contrast to students in other grades, those in the third grade manifested fewer contacts but prolonged duration (A: P < 0.001; B: P < 0.001). Moreover, students exhibited significantly higher contact frequencies than teachers, with a median increase of 215 contacts (median: 985 vs. 770 contacts; P = 0.032). Similarly, their median single-contact duration was 215 s longer (student median: 193 s vs. teacher median: 71 s; P < 0.001).

Fig. 2.

Contact patterns within and between different groups. A Number of contacts by gender, grade and activity status of the participants (student and teacher) based on the contact data. The annotations adjacent to the boxes indicate the median and the number of individuals in each subgroup. The asterisks above the boxes denote the level of significance of the difference between two groups (Kruskal–Wallis test or Mann–Whitney test): P < 0.05, P < 0.01, P < 0.001, P < 0.0001. B As in A, but counted by average single contact duration. C Contact matrix by class. The matrix entry for row A and column B gives the number of contacts (n_AB) measured between individuals of classes A and B based on collected data. The color intensity represents the number of contacts. D Same as C, but each cell representing accumulated duration (w_AB, in minutes) of contacts

By breaking down contacts by the class of contacting and contacted individuals, mixing patterns were elucidated by means of contact matrices by class (Fig. 2 C–D). The contact matrices revealed a dark-blue diagonal, indicating the majority of contacts occurred within classmates, accounting for 97.40% of the total number of contacts and 97.95% of the total contact duration for students. Furthermore, two-by-two lightly shaded blocks flanking the diagonal (e.g., Class 3 A and 3 C, Class 5 A and 5 C) illustrated a higher frequency and longer duration of contacts among children within the same grade compared to those across different grades, while this pattern was less discernible in some grades. Finally, due to the constrained teacher population, both contact frequency and distinct contacts remain relatively low. However, standardized contact matrix heatmaps reveal that teacher interactions and within-grade contacts exhibit contact frequency and duration ranked second only to within-class contacts (Appendix Figure A2).

Contact networks

We constructed a social network formed by contacts from 286 students and 6 teachers with a 20-s duration threshold (Fig. 3). It reveals how individuals engaged with peers, forming a large number of clusters. The contact data of the six classes investigated can form a good network modularization (Q = 0.809, n = 6), forming a highly recognizable subset group. The node with the highest degree (91) in the network also had the highest closeness centrality (0.591) and eigenvector centrality (1), indicating that this node has the highest Links, is closest to other nodes, and possesses the greatest influence within the network. In the third-grade network, the proportions of nodes ranking in the top 10% for weighted degree, closeness centrality, eigenvector centrality, and betweenness centrality were 20.00%, 50.00%, 62.01%, and 20.00%, respectively (Fig. 4). These proportions in fourth-grade network were 43.33%, 16.67%, 0%, 43.33%; in fifth-grade networks were 36.67%, 33.33%, 37.93%, 36.67%.

Fig. 3.

Static network diagram(a 20-s threshold for accumulated contact duration). Each numbered circle represents an individual, and its size indicates the degree. Links between circles represent contacts between individuals, with linewidth reflecting the accumulated duration of contact based on collected data. Various colors denote different classes, with red circles specifically representing six teachers

Fig. 4.

Distribution of network node importance metrics across grade levels. The black Line indicates the 10th percentile value of node importance metrics across the network. A Weighted degree. B Closeness centrality. C Betweenness centrality. D Eigenvector centrality

According to the school timetable, the overall network was subdivided into six sub-networks based on distinct types of activities and the diverse structural characteristics of these networks are exhibited (Table 2). Each sub-network had relatively high modularity (0.764 ~ 0.830) with six communities. The overall network exhibited a large scale, sparse connections between nodes but high clustering, high information dissemination efficiency, and a certain modular structure. The average degree, density and average clustering coefficient of the class network, recess network, and lunch break network were comparable. Among them, the class network exhibited the highest average clustering coefficient (0.857) and modularity (0.830), while the lunch break network had the shortest diameter (5) and average distance (2.540). On the contrary, the large recess activity network exhibits a relatively dispersed structure and sparse connections between nodes, with the largest diameter (7) and average distance (3.280), as well as the smallest average degree (29), density (0.102), and average clustering coefficient (0.684).

Table 2.

A summary of network metrics for different networks

Networks	Nodes	Edges	Average Degree	Weighted Average Degree	Diameter	Density	Average distance	Average clustering coefficient	Modularity (n)
Overall	292	8666	59	184,459.116	3	0.204	2.084	0.792	0.809(6)
Morning self-study	291	4959	34	11,490.179	6	0.118	2.906	0.725	0.764(6)
Class	292	6394	44	110,679.308	6	0.150	2.889	0.857	0.830(6)
Recess	292	6493	44	66,005.233	5	0.153	2.797	0.847	0.823(6)
Large recess activity	290	4266	29	25,427.690	7	0.102	3.280	0.684	0.821(6)
Lunch break	292	6339	43	49,979.712	5	0.149	2.540	0.773	0.782(6)
After-school service	263	4835	37	17,732.061	5	0.140	2.544	0.734	0.779(6)

The structural characteristics of the contact networks varied significantly depending on the proximity threshold and the assigned risk level (Tables 3 and 4). As the contact distance threshold became stricter (moving from < 2 m to < 1 m to < 0.5 m), the networks exhibited a pronounced decrease in connectivity and interaction intensity. The number of edges (Edges) dropped substantially (8666 → 7162 → 5259), and the average weighted degree (Weighted Average Degree), representing the cumulative contact duration per node, decreased dramatically (184,459 s → 119,323 s → 56,437 s). Concurrently, the diameter increased (3 → 4 → 5) and the average distance rose (2.084 → 2.289 → 2.540), indicating reduced overall connectivity and longer path lengths between individuals at closer ranges. Modularity slightly increased (0.809 → 0.812 → 0.816), suggesting marginally stronger community structure (likely class-based) when considering only very close contacts. The clustering coefficient peaked for < 1 m contacts (0.807) before falling for < 0.5 m (0.683) (Table 3).

Table 3.

Network characteristic indicators for different contact distances

Networks	Nodes	Edges	Average Degree	Weighted Average Degree	Diameter	Density	Average distance	Average clustering coefficient	Modularity (n)
< 2 m	292	8666	59	184,459.116	3	0.204	2.084	0.792	0.809(6)
< 1 m	289	7162	50	119,322.644	4	0.172	2.289	0.807	0.812(6)
< 0.5 m	289	5259	36	56,436.657	5	0.126	2.540	0.683	0.816(6)

Table 4.

Network characteristic indicators of different risk contacts

Networks	Nodes	Edges	Average Degree	Weighted Average Degree	Diameter	Density	Average distance	Average clustering coefficient	Modularity (n)
High risk	166	112	1.470	5766.530	7	0.009	2.226	0.052	0.952(48)
Medium high risk	292	1186	8.123	65,026.664	6	0.028	2.448	0.467	0.832(6)
Medium low risk	292	1435	9.829	22,738.404	8	0.034	3.207	0.395	0.826(6)
Low risk	292	6334	43.384	27,659.068	6	0.149	2.893	0.849	0.826(6)

Analysis by risk level revealed stark differences. The High-Risk network (< 0.1 m & > 15 min) was extremely sparse and fragmented, with very few edges (112), low average degree (1.470), the highest diameter (7), very low density (0.009), minimal clustering (0.052), and very high modularity (0.952) split across 48 small communities. In contrast, the Medium–High-Risk network (< 1 m & > 10 min, excluding High Risk) retained significant connectivity (1186 edges) and exhibited the highest interaction intensity per connection (Weighted Average Degree = 65,027 s), while maintaining relatively high modularity (0.832) similar to the overall network. The Low-Risk network (< 2 m & < 5 min) formed the most connected sub-network (6334 edges, highest Average Degree 43.38 and Density 0.149), but with much lower interaction intensity per connection (Weighted Average Degree = 27,659 s) compared to Medium–High-Risk contacts. These analyses demonstrate that while close-proximity interactions (< 0.5 m) are less frequent, they form distinct, somewhat isolated clusters. Crucially, Medium–High-Risk contacts, though constituting only 4.39% of total contacts (Sect. 3.3), form a network backbone characterized by sustained, high-intensity interactions concentrated within communities (classes), potentially playing a disproportionate role in transmission dynamics (Table 4).

Risk levels

After classifying contacts risk levels according to Table 1, 90.12% of all recorded contacts were deemed low-risk, comprising only 29.73% of the total duration. Medium–high-risk exposure contributed to 47.10% of the total exposure duration but represented just 4.39% of the total number of contact. Additionally, a minority of contacts presented a high risk of infection.

The ratio of time at risk to activity time illustrates the risk per unit time for each activity and the proportion of exposure time at different risk levels within the total exposure time of each activity (Fig. 5 A). It allows comparison across the different activities, independent of their duration. Notably, the distribution of contact risk levels is similar for class and large recess activities, as well as for after-school service and lunch break, potentially influenced by the duration and scope of the activities. Activities such as class, after-school service, large recess activity and lunch break had a higher occurrence of high-risk contacts per unit of time. In addition, considering both high-risk and medium–high-risk contacts, class and large recess activities emerge with higher overall risk levels.

Fig. 5.

Risk levels of type of activities, interaction and the percentage of contact duration in each activity in different contact type groups. A Distribution of the risk levels by type of activity, sorted according to a decreasing percentage of high-risk contacts. Four color bars for each activity, representing the four risk levels. The height of the bars represents the sum of the duration of all contacts during each activity divided by the total duration of the activity.The annotated percentages show the fraction of contact time within each risk level, for each activity. B Distribution of risk levels by group and type of interaction. The color bars, annotated with percentages, which refer to the right-hand scale, report the percentage of time of contact within each risk level. The grey background bars, which refer to the left-hand scale, report accumulated contact duration for each contact pair of the six categories. C The height of the colored bar represents the proportion of contact duration of various activities in the contact type group, with a percentage annotation above the bar. The upper left legend indicates various types of activities and corresponding colors

The average cumulative contact duration of each contact pair in the three types of contact in the same grade (S–S 56.17, S-T 13.12, T-T 57.46) is higher than that between different grades (S–S 4.15, S-T 0.39, T-T 1.17) (Fig. 5 B). High risk contacts exclusively occurred among students or between students and teachers within the grade, constituting 3.14% of the former's total duration and 0.33% of the latter's. Teachers and students in the same grade had higher risk of contact due to their longer average contact duration and their total proportions of medium–high risk and high risk were 52.40% and 50.62%, respectively. Due to the short cumulative contact time and all low-risk contact, the transmission risk of cross grade teacher to teacher contact and cross grade teacher to student contact is lower than that of cross grade student contact (38.99% medium–high-risk contact).

The contact data were categorized according to different contact types and activities, showing the distribution of the contact duration of different activities among six contact type groups. Primary interactions among same-grade students mostly occur during class and lunch break, constituting 73.30% of their total contact time (Fig. 5 C). Cross-grade student interactions predominantly take place during morning self-study and during after-school service due to different grades of students participated in school club activities, contributing to 90.45% of the total cross-grade contact time. Teacher interactions across grades occurred exclusively during lunch and lunch breaks.

Discussion

With advancing wireless communication technology, wearable devices are increasingly utilized to objectively record in an unsupervised manner, time-resolved data on the contact duration and frequency, employing technologies like RFID (Radio Frequency Identification) [12], GPS [16] and WiFi [17], etc. These wearable devices can collect social contact data, avoiding recall bias, but the selected transmission signal frequency of 0.05 Hz miss short-term contact. To address this limitation, our study employed UWB wearable devices to accurately record close contact (< 2 m) data of students and teachers at a signal transmission frequency of 0.25 Hz, and constructed a social contact network for the school population. As our knowledge, our study is the first one to capture the high-precision social contact data in whole-day campus in China. Our study also analyzed and compared the contact patterns and network characteristics of different groups and activity scenarios, including average degree, density, average distance, and clustering coefficient. Currently, researches on student social contact mostly focus on Europe and North America [12]. The Study has shown that school-age children exhibit relatively long total contact times [18]. Higher school-age students have stronger inter-grade interaction, while lower school-age children have higher modularity, indicating that intervention measures may be more effective for lower school-age children [19]. In our analysis results, the contact network with high modularity (0.809) manifests that each class are relatively independent, linked to the relatively fixed daily school schedule. Fourth and fifth-grade students had a significantly higher frequency of contact than third-grade students, indicating that higher school-age students had higher frequency of contact. However, third-grade students exhibited longer contact times, which potentially associated with a higher disease transmission risk. Meanwhile, some potential super spreaders also appeared among third-grade students. In addition, the results suggest that the main mode of contact among student groups is intra-class, followed by contact between classes of the same grade, and finally contacts between grades. For students, contact within the same grade primarily happened in the class and lunch breaks, while interactions cross grades were concentrated in morning self-study and after-school services. Risk classification assessment revealed that high-risk exposure mainly occurred among students in the same grade, particularly during class, large recess activity, after-school service and lunch break.

The findings indicated that the third-grade students exhibited a higher risk of infection compared to their counterparts in other grades. On the one hand, this Heightened risk can be attributed to longer duration of single contact among third-grade students. On the other hand, among the top 10% of individuals in terms of closeness centrality and eigenvector centrality, the proportion of third-grade students was higher than that of other grades. A large number of nodes with high degree indicates the presence of more potential super-spreaders. The high eigenvector centrality of the node indicates that the student's contacts are in contact with many students. In China, the incidence of common communicable respiratory diseases in men is higher than that in women [20], so the number of male contacts is more likely to contact infected people, which provides favorable conditions for the transmission of communicable diseases. The result is consistent with the high incidence of communicable respiratory diseases among lower-grade students [21]. Another study on the association between social network characteristics and tuberculosis infection also shows that people who are more central in their social networks and have more male contacts had a higher risk of tuberculosis infection [22]. In contrast to students within the bottom 10% of degree, those in the top 10% exhibited increased interaction with peers from diverse classes within the same grade (Appendix Figure A3), particularly during the lunch break (Appendix Figure A4). It is recommended that we should increase the vaccination rates of influenza, chickenpox, and other vaccine-preventable diseases among lower grade students. Additionally, class teachers could organize low-mobility activities (e.g., animation viewing, individual handicrafts) to promote stationary behaviors during lunch breaks. These activities reduce spontaneous interactions by: 1) Confining students to fixed seating arrangements; 2) Minimizing cross-class movement; 3) Replacing high-contact play with focused individual/paired tasks.

Some studies emphasize the significance of contact distance and duration in assessing the risk of COVID-19 transmission [23, 24]. Accordingly, our study introduces risk level indicators to evaluate the communicable respiratory diseases transmission risk corresponding to various groups and activity types of contact. Our findings reveals teachers and students in the same grade have higher risk of contact, with respective average contact times of 57.46 and 56.17 min. Among them, the total proportions of medium–high risk and high risk are 52.40% and 50.62%, respectively, mainly occurring in the highest risk activity – attending classes. In class, students are basically in a fixed position in the classroom keeping contact with surrounding students for a long time. During large recess activity, students performed physical exercises in class-based formations, maintaining orderly queues. Therefore, contact occurred in class or among same-grade students had higher risk. Considering the contact patterns, characteristics of class network and risk assessment, it is suggested that class suspension should be prioritized in appropriate scenarios during the pandemic. Suspending classes for different grades or even closing the school may not significantly contribute to interrupting the transmission of communicable diseases [25]. Besides, measures such as vaccination, wearing masks, and improving classroom ventilation can be taken to reduce the risk of infection [26, 27]. Without restricting social distance, activities range and time are the main influencing factors for contact risk assessment [15], so the magnitude of contact risk can be changed by adjusting them. However, the time and location of class are usually fixed, so if the classrooms and staffs allow, it is feasible to reduce the risk of transmission by decreasing the number of students in the class [28]. Moreover, study also shows contact between teachers in the same grade carried a higher risk of infection, which is consistent with the results of the study on Covid-19 transmission in schools in Britain and Germany [29, 30]. This suggests that we should pay attention to teachers'vaccination status, masks wearing and social distance to avoid the rapid spread of the disease among teachers, thereby increasing the transmission rate of contact between teachers and students.

Finally, in different types of cross-grade contact, contact among students of different grades had medium–high risk contact and longer contact duration, indicating that the transmission of diseases across grades is primarily through inter-grade student contact. While 90.45% of the cross-grade student contact duration concentrated in early self-study and after-school services, which is likely due to some students of different grades participating in club activities at that time. Therefore, when there is a case in the class, we can prevent the further spread of the disease to other grades by prohibiting all students in the class from participating in club activities.

However, the study has limitations. First, data collection was confined to a single day at one school, which may not capture weekly/seasonal behavioral variations. Second, the inability to track influenza symptoms and infection cases among students and teachers during the flu season through a cohort study has hindered the validation of our findings. However, the influenza vaccination rates for all classes are higher than 85%, forming a certain level of immune barrier. Given the consistent daily schedules among schools in the region, we can observe groups of students with lower vaccination rates when the next flu season arrives. Furthermore, based on our understanding, the daily schedules of public elementary schools in China are generally highly consistent, making our results highly relevant for reference in research at other schools.

Conclusion

In summary, research on factors influencing the transmission of communicable diseases, such as social contact networks in schools and childcare facilities, should be given renewed attention. Our findings indicate that there is a high risk of transmission within the class during the early stage of outbreak. Implementing measures like home isolation for infected students and class suspensions can reduce further transmission within the class. Controlling risky on-campus activities could restrain internal school transmission. Low-grade students should be concerned on vaccination coverage.

Authors’ contributions

X.Z., X.L., Q.L., L.Y. and Z.Z. designed research; Z.Z., R.L., S.D., L.L., Y.W., R.C. and X.M. performed research; R.L., Q.L., X.M., L.Y., K.Z. and K.L. analyzed data; Q.L., R.L. and X.L. wrote the paper; X.Z., Q.L. and Z.Z. reviewed throughout the manuscript.

Funding

This work was supported by the Key Project of Shenzhen Science and Technology Innovation Commission (JCYJ20210324115411030; JCYJ20200109150715644; JSGG20200225153023511); Shenzhen Key Medical Discipline Construction Fund (SZXK064); Shenzhen San-Ming Project of Medicine (SZSM202011008); Shenzhen Municipal Technological Project (JCYJ20200109150715644).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Before the study, parents and teachers were informed on the details and aims of the study. Informed consent was obtained from parents, teachers and from the director of the school. This experiment was approved by the Ethics Committee of Shenzhen Center for Disease Control and Prevention (SZCDCLL-[2022]010A) and conducted in accordance with the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.