Source-linked AI summary
High-resolution measurements of face-to-face contact patterns in a primary school
J. Stehlé, N. Voirin, A. Barrat, C. Cattuto, L. Isella, J. -F. Pinton, M. Quaggiotto, W. Van den Broeck, C. Régis, B. Lina, P. Vanhems
TL;DR
The paper addresses limited quantitative evidence on children’s mixing in schools and its relevance to respiratory-infection transmission. It measures time-resolved face-to-face proximity with RFID devices in a French primary school, finding structured contact patterns useful for disease modeling and evaluating control measures.
Problem
Limited quantitative information exists on children’s mixing patterns in school environments, despite the relevance of school contacts to respiratory-infection transmission opportunities.
Method
The study measured time-resolved face-to-face proximity among children and teachers using RFID-based proximity sensing in a French school.
Results
The study found structured contact patterns relevant to modeling disease propagation and evaluating school control measures.
Takeaways & Limitations
The observed contact properties can inform mathematical models and help prioritize preventive measures, case isolation, and class or school closures during epidemics.
Takeaways & Limitations
The measurements may be used to study proximity but not the occurrence of physical contacts.
Abstract
from arXiv · showhide
Little quantitative information is available on the mixing patterns of children in school environments. Describing and understanding contacts between children at school would help quantify the transmission opportunities of respiratory infections and identify situations within schools where the risk of transmission is higher. We report on measurements carried out in a French school (6-12 years children), where we collected data on the time-resolved face-to-face proximity of children and teachers using a proximity-sensing infrastructure based on radio frequency identification devices. Data on face-to-face interactions were collected on October 1st and 2nd, 2009. We recorded 77,602 contact events between 242 individuals. Each child has on average 323 contacts per day with 47 other children, leading to an average daily interaction time of 176 minutes. Most contacts are brief, but long contacts are also observed. Contacts occur mostly within each class, and each child spends on average three times more time in contact with classmates than with children of other classes. We describe the temporal evolution of the contact network and the trajectories followed by the children in the school, which constrain the contact patterns. We determine an exposure matrix aimed at informing mathematical models. This matrix exhibits a class and age structure which is very different from the homogeneous mixing hypothesis. The observed properties of the contact patterns between school children are relevant for modeling the propagation of diseases and for evaluating control measures. We discuss public health implications related to the management of schools in case of epidemics and pandemics. Our results can help define a prioritization of control measures based on preventive measures, case isolation, classes and school closures, that could reduce the disruption to education during epidemics.
Methods and Findings
The study addresses limited knowledge of how children mix in schools by measuring face-to-face proximity in a French primary school. These contact patterns were examined for their relevance to respiratory-infection transmission and school containment measures.
- Motivation: Children’s school mixing is poorly documented, although describing contacts could quantify respiratory-infection transmission opportunities.
- Methods: The study used RFID proximity sensing to collect unsupervised, time-resolved face-to-face data from children and teachers.
- Analysis and implications: The study characterized contact number and duration by age, class, time of day, and school spatial structure, highlighting situations relevant to transmission and containment.
- Study setting: The measurements took place over two days in a French primary school serving children aged 6–12 years across 10 classes.
- Study setting: The school included assigned class rooms and teachers, shared playground and canteen spaces, and staggered breaks and lunches because common areas lacked capacity.
Study design/Ethics statement and privacy
The study used informed consent, institutional notification and approval, encrypted communications, and badge identifiers linked only to class information. Wearable RFID badges measured close, face-to-face proximity continuously in monitored school spaces.
- Ethics: Relevant French ethics and privacy bodies were notified, and the study received approval from the relevant academic authorities.
- Ethics: Parents, teachers, and the school director provided verbal informed consent after the study’s aims and details were explained.
- Privacy: RFID communications were encrypted, and badge identifiers were associated only with participants’ classes rather than personal information.
- Measurement infrastructure: Unobtrusive chest-worn badges exchanged radio packets when individuals faced each other at close range, approximately 1–1.5 m.
- Contact definition: A contact was recorded in each 20-second interval when two badges exchanged at least one radio packet and continued while exchanges persisted.
Collected data
The study measured face-to-face proximity among children and teachers over two school days and analyzed contact structure across individuals, classes, time, and school spaces.
- Data collection: 232 children and 10 teachers across 10 classes were monitored over two days, with contacts recorded during school hours.The measurement periods ran from 8:45am to 5:20pm on the first day and 8:30am to 5:05pm on the second.
- Contact measures: Contacts were represented through occurrence, event frequency, and cumulative duration measures for individual pairs.The occurrence indicator records whether any contact occurred, frequency counts events, and cumulative duration sums their durations.
- Contact measures: Class-level analyses computed total and per-child contact counts and durations, producing symmetric and class-size-adjusted matrices.These quantities were calculated for whole study periods and separately, accounting for differing class sizes.
- Temporal analysis: Temporal analyses tracked each child’s distinct contacts and cumulative contact time, separated into same-class and different-class interactions.The resulting degree and strength measures describe how contact patterns build at class and school levels.
- Network and spatial analysis: Twenty-minute sliding-window networks and daily weighted networks captured slowly varying interactions and compared contact patterns across days.RFID reader signal rates also supported approximate badge localization and reconstruction of children’s trajectories within school premises.
Number of contacts
The study recorded many contacts per individual, with contact counts varying across days and classes but showing limited heterogeneity overall.
- Overall contact counts: 77,602 contact events involving 242 individuals were recorded across the two study days.There were 37,414 contacts on day 1 and 40,188 on day 2.
- Individual contact counts: 317 contacts per individual were recorded on day 1 and 338 on day 2.The squared coefficients of variation were CV^2=0.22 and CV^2=0.27, respectively.
- Group differences: Teachers had fewer total contacts than children, largely because classes contained more children than teachers.Figure 2 indicates that the difference mainly reflects the number of individuals available for contact.
- Variation across settings: Contact numbers showed limited heterogeneity across classes and study days.Differences were probably related to class schedules and school activities, including absences and group versus individual work.
- Individual contact counts: 50 distinct individuals were contacted on average on day 1 and 46.5 on day 2.The corresponding squared coefficients of variation were CV^2=0.14 and CV^2=0.18.
Duration of contacts
Contacts were usually brief, but their durations spanned a broad range, and cumulative pairwise contact times were heterogeneous and structured by class and age.
- Contact-event durations: 33 seconds was the average contact duration, with 88% of contacts lasting less than 1 minute.More than 0.2% of contacts exceeded 5 minutes, and no characteristic contact time scale was defined.
- Cumulative pairwise durations: 64% of individual pairs interacted for less than 2 minutes per day, while 9% spent more than 10 minutes together and 0.38% more than 1 hour.Pairwise cumulative interaction times therefore had a broad distribution.
- Cumulative pairwise durations: 207 seconds on day 1 and 236 seconds on day 2 were the average daily pairwise face-to-face proximity times.The squared coefficients of variation were 5.44 and 4.65, indicating broad distributions.
- Individual cumulative durations: 10,340 seconds on day 1 and 11,000 seconds on day 2 were the average total daily contact times per individual.Individual-level totals were more homogeneous, with CV^2=0.25 on day 1 and CV^2=0.33 on day 2.
- Class and age structure: Most contacts occurred within the same class, with additional concentration among children in the same grade.Contacts also declined with grade separation, and younger versus upper grades were separated in the mixing structure.
- Class and age structure: Each child spent three times more daily face-to-face time with classmates than with children from other classes.Teachers had sparse contacts with one another because they spent most of their time in class with children.
Temporal evolution of the contact patterns
Contact networks built rapidly within classes and expanded across classes mainly during breaks, while individual-level patterns were broadly reproducible across the two days.
- Temporal mixing: The average network degree peaked at breaks and around the beginning or end of lunch.These peaks corresponded to periods when children from different classes mixed.
- Network growth: After a few hours, each child had contacted all members of their own class, while new cross-class contacts appeared mainly during breaks.This produced plateaus in cumulative average degree between periods of cross-class mixing.
- Temporal mixing: Three times more daily contact time was spent with classmates than with children from other classes.Same-class contact time increased regularly, whereas different-class contact time changed substantially mainly during lunch.
- Day-to-day repetition: 0.53, 0.54, and 0.53 were the day-to-day Pearson correlations for contact number, contact time, and distinct persons contacted.These values indicate overall strong correlation between individual behavior across days.
- Day-to-day repetition: 26 repeated and 20 new contacts were observed per child on average on the second day relative to the first.Repeated contacts included 19 within-class and 7 cross-class contacts, while new contacts included 1.4 within-class and 18.4 cross-class contacts.
- Day-to-day repetition: 0.67 was the average cosine similarity between individual neighborhoods across the two days.Similarity was 0.74 within each child’s class and 0.2 for contacts with children from different classes.
Trajectories in space
Children’s scheduled movement through classrooms, public spaces, the courtyard, and cafeteria shaped when and how they mixed across classes and grades.
- Trajectories in space: Figure 10 traces children moving from classrooms to the courtyard and back, with some lunch routes passing through the cafeteria.
- Trajectories in space: Scheduled trajectories strongly contributed to mixing patterns between classes and grades.
- Trajectories in space: The study provided detailed measures of close face-to-face proximity interactions among primary-school children.
- Measurement approach: RFID badges recorded time-resolved, close-range face-to-face proximity over two school days without relying on questionnaires.
- Measurement approach: More than 95% participation and unobtrusive wearable devices enabled unsupervised contact detection.
Comparison with previous studies
The study’s contact counts and durations differ from previous studies because measurement ranges and contact definitions differ, but become compatible when only longer contacts are retained.
- Methodological differences: Unsupervised proximity detection does not rely on participant surveys or memory and is expected to produce larger total contact numbers and durations than survey methods.
- Methodological differences: The study detected less numerous and shorter contacts than a study using a broader 3-meter proximity range.
- Observed contact levels: Most contacts were short, but contact durations were highly heterogeneous and included a non-negligible fraction of long-lasting contacts.
- Threshold comparisons: At a cumulative-duration threshold of at least one minute, each child interacted with 21 different children for 163 minutes.
- Methodological differences: The results were quantitatively different from previous studies but became compatible after filtering to retain only the longest contacts.
Public health implications in the field of infectious diseases
School contacts show strong age and class structure, with activity peaks during breaks and lunch. These patterns suggest targeted interventions could reduce disruption compared with closing entire schools.
- Contact structure: Children mixed preferentially within their age group, an effect attributed largely to studying together.
- Contact structure: Contact matrices showed a hierarchical block-diagonal structure, with strong within-class entries and blocks for grades and junior or senior groups.
- Temporal structure: Breaks and lunch produced periods of higher contact activity and the largest increases in cumulative contact weights.
- Intervention implications: Each child spent three times more time with classmates than with children from other classes, suggesting selected-class closure as a possible alternative to whole-school closure.
- Intervention implications: Exposure matrices and aggregated contact networks could identify classes linked to the class of the first detected case for targeted closure.
- Intervention implications: Shifting class schedules could substantially reduce contacts between classes, particularly for preventing transmission from asymptomatic cases.
Informing mathematical models
The measured contact structure provides information for infectious-disease models, while its scope is limited by the school setting, measurement period, and proximity-based sensing.
- Exposure matrices: Exposure-matrix cells report the average number and duration of contacts between individuals in two categories, supporting refinement of young age groups in contact matrices.
- Model structure: Most contacts occurred within classes, with relatively few contacts across classes, limiting disease spread between classes.
- Model structure: The observed patterns indicate that homogeneous contact-duration or between-class mixing assumptions may mislead models, whereas homogeneous mixing within classes may be a good approximation.
- Study limitations: The infrastructure measured children in school buildings or playgrounds but not during sports activities, which can involve close proximity and physical contact.
- Measurement scope: The data provide badge proximity rather than physical-contact measurements, supporting use for respiratory-spread pathogens but not agents transmitted by skin contact.
- Study limitations: Data collection lasted only two days, limiting conclusions about longer timescales and motivating longer deployments for confirmation.
- Study limitations: Results depend on the school’s schedule and spatial structure, so generalization to other schools requires caution.
Reference Setting Contact definition Results
The study defines face-to-face contacts using RFID-detected proximity and finds frequent, heterogeneous interactions among schoolchildren. Children averaged 323 daily contacts with 47 other children, totaling 176 minutes of contact time.
- Contact definition: A contact is a continuous sequence of proximity between two individuals detected by the RFID devices.
- Results: On average, each child had 323 contacts per day with 47 distinct other children.
- Results: Contacts lasted 33 seconds on average, but their duration distributions were broad.
- Results: Each child accumulated 176 minutes of contact time per day on average.
- Reference Setting: The RFID system detected proximity at approximately 1 to 1.5 meters between individuals wearing the devices.