Analyses of the 1957 (Asian) influenza pandemic in the United Kingdom and the impact of school closures

Abstract

Many countries plan to close schools during a future influenza pandemic, although the potential impact is poorly understood. We apply a model of the transmission dynamics of pandemic influenza to consultation, serological and clinical data from the United Kingdom from the 1957 (Asian) influenza pandemic, to estimate the basic reproduction number (R0), the proportion of infected individuals who experience clinical symptoms and the impact of school/nursery closures. The R0 for Asian influenza was about 1.8 and 60-65% of infected individuals were estimated to have experienced clinical symptoms. During a future pandemic, closure of schools/nurseries could reduce the epidemic size only by a very small amount (<10%) if R0 is high (e.g. 2.5 or 3.5), and modest reductions, e.g. 22% might be possible if it is low (1.8) and schools are closed early, depending on assumptions about contact patterns. Further data on contact patterns and their dependence on school closures are needed.

Fig. 1

Comparison between the best-fitting model-predictions of the weekly numbers of cases per 100 000 in different age groups which were reported to the general practice in Wales [15] for matrices W1, W2, W3 and W4 and the observed data, assuming that the average durations of the latent and infectious periods were each 2 days.

Fig. 2

Comparison between (a) the best-fitting model predictions of the expected proportion of individuals who were positive to antibodies of the Asian strain of influenza by the end of the pandemic against data collected in Sheffield in November 1957 [17] and (b) the best-fitting model predictions of the proportion of individuals who experienced clinical disease during the Asian influenza pandemic wave, and the corresponding observed proportion of individuals who experienced disease during the pandemic in a practice in South East London [16]. These predictions are based on the assumption that the latent and infectious periods were each 2 days.

Fig. 3

Model predictions of the impact of school closures on the overall weekly influenza disease incidence rate, assuming that individuals contact each other according to WAIFW matrix W3 and W4 (left- and right-hand panels respectively), for different assumptions about the threshold weekly influenza disease incidence rate at which school/nursery closures are introduced, the effect of school/nursery closures on contact between individuals of pre-school and school age and the basic reproduction number. Predictions based on matrix W3 are similar to those based on matrix W2; those based on W1 are similar to those based on matrix W4. The epidemic curves predicted using assumptions that schools/nurseries are closed at a threshold incidence of 50 and 200 per 100 000 per week were identical. The latent and infectious periods are here assumed to be 2 days.

Fig. 4

Model predictions of the impact of school/nursery closures on the average reduction in the epidemic size, for different assumptions about the threshold weekly influenza disease rate at which school/nursery closures are introduced, the effect of school/nursery closures on contact between individuals of pre-school and school age, and the basic reproduction number. The lower and upper limits of the vertical bars reflect the minimum and maximum reductions predicted under each WAIFW matrix. The latent and infectious periods are here assumed to be 2 days.

Fig. 5

Model predictions of the impact of school/nursery closures on the weekly influenza disease incidence rate in specific age groups, assuming that individuals contact each other according to WAIFW matrices W3 and W4 (left- and right-hand panels respectively). The basic reproduction number is here assumed to be 1·8 and school/nursery closures are assumed to reduce contact between nursery/school-aged children by 75%. Predictions based on matrix W3 are similar to those based on matrix W2; those based on W1 are similar to those based on matrix W4. The latent and infectious periods are here assumed to be 2 days.

Fig. 6

Model predictions of the impact of school closures, which are introduced once the disease incidence exceeds 50/100 000 per week, on the reduction in the proportion of individuals experiencing influenza, assuming that the basic reproduction number is 1·8 and that school closures reduce contact between individuals of nursery/school-age by 75%, for different assumptions about contact between individuals in the overall population (matrices W1, W2, W3 and W4). The latent and infectious periods are here assumed to be 2 days.

Fig. 7

Model predictions of the time period during which schools would be closed (left-hand panel) and the duration of the epidemic (right-hand panel), defined as the time period during which the disease incidence rate was above 10/100 000 per week, for different assumptions about the threshold weekly influenza disease rate at which school closures are introduced, the effect of school closures on contact between individuals of pre-school and school age and the basic reproduction number. The lower and upper limit of each vertical bar reflects the minimum and maximum durations predicted under each WAIFW matrix. The latent and infectious periods are here assumed to be 2 days.