The Relationship Between School Holidays and Transmission of Influenza in England and Wales

Abstract

School closure is often considered as an influenza control measure, but its effects on transmission are poorly understood. We used 2 approaches to estimate how school holidays affect the contact parameter (the per capita rate of contact sufficient for infection transmission) for influenza using primary care data from England and Wales (1967-2000). Firstly, we fitted an age-structured susceptible-infectious-recovered model to each year's data to estimate the proportional change in the contact parameter during school holidays as compared with termtime. Secondly, we calculated the percentage difference in the contact parameter between holidays and termtime from weekly values of the contact parameter, estimated directly from simple mass-action models. Estimates were combined using random-effects meta-analysis, where appropriate. From fitting to the data, the difference in the contact parameter among children aged 5-14 years during holidays as compared with termtime ranged from a 36% reduction to a 17% increase; estimates were too heterogeneous for meta-analysis. Based on the simple mass-action model, the contact parameter was 17% (95% confidence interval: 10, 25) lower during holidays than during termtime. Results were robust to the assumed proportions of infections that were reported and individuals who were susceptible when the influenza season started. We conclude that school closure may reduce transmission during influenza outbreaks.

Keywords: disease transmission; human; schools; infectious; influenza.

Figure 1.

Proportions of the population (all ages) (A) and children aged 5–14 years (B) susceptible to influenza infection at the start of each influenza year from 1967/1968 to 1999/2000 in England and Wales. Results were based on season-specific serological data where available; where data were not available, 30% of individuals were assumed to be susceptible at the start of each season. Light gray columns show values from published data sources; dark gray columns show assumed values. Error bars show 95% confidence intervals for data sets which provided denominators; where there are no error bars, denominators were not available.

Figure 2.

Estimated percentage difference in the contact parameter (among children under age 15 years) for influenza during holidays compared with termtime based on fitting of the age-structured model to influenza-like illness general practitioner consultation data, England and Wales, 1967–2000. X's, single dominant subtype; circles, more than 1 subtype circulating. Error bars show 95% confidence intervals.

Figure 3.

Estimated percentage change in the value of the contact parameter for influenza during school holidays based on application of the simple mass-action model to influenza-like illness general practitioner consultation data for children aged 5–14 years (reporting fraction was assumed to be 50%), England and Wales, 1967–2000. X's, single dominant subtype; circles, more than 1 subtype circulating; squares, unknown number of subtypes circulating. Dotted lines indicate years in which there were ≤2 estimates of the contact parameter during termtime and/or holidays. Error bars show 95% confidence intervals.