Who meets whom, and how would closing schools slow an epidemic?
Contact diaries showed that people mostly mix with others of similar age, and plugging holiday-period contacts into an epidemic model suggested that closing schools would cut the reproduction number of a flu-like infection by about a sixth.
Source
Mining social mixing patterns for infectious disease models based on a two-day population survey in Belgium
Study at a glance
- Design
- Cross-sectional — Population contact-diary survey (one random weekday plus one random weekend day per person) recruited by random digit dialling, analysed with association rules, classification trees and weighted GEE, then used to parameterise a next-generation-matrix epidemic model.
- N
- N=750 · 750 Belgian residents (adults, teenagers and children, with parents completing diaries for young children), each recording contacts on two days; analyses are at the level of recorded contacts and participant-days.
- Population
- A telephone-recruited sample of Belgian residents of all ages, stratified by age and region (Flanders, Wallonia, Brussels), surveyed in spring 2006.
- Outcome
- Number and type of daily social contacts (conversation vs skin-to-skin touch, location, duration, frequency) and the age-by-age contact matrix; modelled change in the basic reproduction number under school-holiday and weekend mixing.
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The researchers recruited 750 people living in Belgium by random telephone dialling and asked each to keep a paper diary of everyone they talked to or touched on one weekday and one weekend day. People with very many work contacts estimated those separately, and the team imputed them rather than dropping them. They mined the diaries for patterns (which contacts involve touching, where and how long), modelled what predicts how many contacts a person has, built an age-by-age contact matrix, and used it in a model of a new flu-like infection to compare normal, school-holiday and weekend mixing.
What they found
Most close contacts happened at home, and touching was most likely for long, frequent, home or leisure contacts; overall 61% of contacts involved skin-to-skin touch. Contact numbers rose with household size and class size, adults in work or study had about twice as many contacts as unemployed or retired adults, and adults living with children averaged 10.4 close contacts versus 8.2 for those without. Mixing was strongly assortative by age with a parent–child component. With the reproduction number set to 2 under normal term-time mixing, holiday-period contact patterns lowered it to 1.69 (a 15.5% reduction), while weekend-only patterns lowered it to 1.33.
The limits
What it doesn't show
The school-closure result is a model projection built on self-reported diaries, not an observed drop in infections, and it assumes transmission is proportional to touching contacts with a fixed five-day infectious period. Recruitment used landline phones only, which the authors note under-represents people who only have mobile phones. Very high work-contact counts had to be imputed, which changed the average number of contacts substantially, and holiday periods differ from true school closures in ways (such as grandparent childcare) that may not transfer to an emergency closure.
Key terms
- Social contact matrix
- A table giving the average number of contacts that a person in one age group has with people in each other age group per day; it tells a model who can infect whom.
- Basic reproduction number (R0)
- The average number of new infections caused by one infectious person in a fully susceptible population; an epidemic grows when it is above 1.
- Assortative mixing
- The tendency of people to have most of their contacts with others similar to themselves, here people of a similar age.
- WAIFW matrix
- 'Who Acquires Infection From Whom': the matrix of transmission rates between age groups that dynamic epidemic models need, which contact data help estimate.
- Imputation
- Filling in values that were not recorded using a statistical rule based on the data that were recorded, here used for the large numbers of work contacts.
Flashcards
0 of 11 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
In the Belgian contact survey, what drove the modelled reduction in R0 during the school holiday?
Common questions
Does this prove that closing schools stops epidemics?
No. It shows that, in a model driven by real contact data, holiday-style mixing lowers R0 from 2 to about 1.69. That slows spread but does not push R0 below 1, and it is a projection rather than observed infection data.
Why not just use weekends to represent school closure?
Weekend patterns also remove work contacts for adults, so they exaggerate the effect (R0 fell to 1.33). The authors argue holiday periods, when schools shut but many adults still work, are the better stand-in.
Why did the team ask people with many work contacts to estimate them separately?
Listing dozens of brief customer or pupil contacts in a diary is burdensome and error-prone. Asking for an estimate and imputing ages kept these contacts in the data instead of silently dropping them, which raised the average contact count noticeably.
More on Infectious disease