This retrospective cohort study shows the role of schools in the spread of an epidemic of mumps among unvaccinated children in a village with low vaccination coverage. When four schools in the same village were compared, the final mumps attack rates were much higher among the unvaccinated children with a connection to a orthodox protestant school than among the other unvaccinated children. Given that the vaccination coverage in the non-orthodox schools was above 90%, the low attack rates among the unvaccinated children with a connection to these schools can probably be attributed to a herd effect .
When the two orthodox protestant schools were compared, the unvaccinated children with a connection to orthodox protestant school A were affected earlier during the epidemic than the unvaccinated children with a connection to the other orthodox protestant school within the same village (i.e., school B). Given that we controlled for possibly confounding factors, this finding shows schools to play a role in the spread of infectious disease among orthodox protestant groups. Schools involve social clustering and, once the mumps has been introduced into a school, it can thus spread more easily among children at the same school than among other children. The school attended by unvaccinated children and their siblings -- together with household size -- thus determined whether the children got the mumps or not and when. And this suggests that school closure can influence the spread of an epidemic within an orthodox protestant population.
The question, of course, is whether or not school closure influences the final outcome of the epidemic. In a simulation study of pandemic influenza, the closing of schools and keeping children at home reduced the final attack rate by 90% -- without the further use of vaccines or antivirals. For this result, however, the children had to be quarantined for the extent of the epidemic, which is given the impact on education not desirable and obviously not achievable in real life .
When considering the effects of school closure, compliance with social distancing during school closure is of critical importance. Recent experiences with school closure for influenza prevention showed the majority of the children to visit at least one social event during the school closure period [22, 23]. Nevertheless, over all contact rates during a school closure period are likely to be considerably lower than during regular school periods. German school children reported four times less contacts on Sundays than on school days, for example .
According to an international diary study about 20% of the contact for people living in the Netherlands is leisure time contact e.g., during sports or other activities . The orthodox protestant way-of-life differs greatly from this, however. For religious reasons, members of this population refrain from sports, cinema, and television . Leisure time activities are nevertheless organized by the churches for such orthodox protestant children, which means that the variable religious denomination can serve as a proxy variable for leisure time activities. In the present study, religious denomination was nevertheless not found to significantly influence the spread of mumps. However, orthodox protestant children will --like other children-- visit family and friends. In the extra leisure time generated by school closure social distancing remains therefore of critical importance.
The perceived seriousness of a disease is an important determinant of compliance with social distancing . According to another study that we conducted, orthodox protestant parents perceive polio to be a particularly serious health threat and thus something that warrants not only social distancing but even consideration of vaccination (manuscript in preparation). Schools may also play a role in the spread of polio. At the beginning of the 1992-1993 polio epidemic, laboratory signs of polio infection were far more prevalent at the orthodox protestant schools of the siblings of the index case than at other schools . Therefore we recommend that school closure be considered during a next polio outbreak. We further recommend additional research and simulation studies in particular to gain more insight into the effects of school closure on the final attack rates of epidemics of vaccine preventable diseases in orthodox protestant populations, while also taking the durations of school closure and levels of vaccination coverage into account.
Some possible limitations on the present study
The overall response rate in our study was 54%. The response rates at the orthodox protestant schools A and B were slightly higher than the response rate of 48% in another study among a orthodox protestant population  and considerably higher than the response rate of 37% at school D where mumps did not appear to be an issue. Given that school D was the smallest school in the village and -- as a public school -- had a nationally representative vaccination coverage of > 95%, we do not think that the low response rate of this school affected our results. A non-response analysis was nevertheless not feasible as vaccination is a sensitive subject and the respondents in our study returned their questionnaires anonymously; we could not, thus, check the actual vaccination status of our respondents in the national register.
Several reports indicate that parental recall of vaccination may be inaccurate but that the inaccuracy concerns mostly the number of injections and vaccination dates [29–31]. Given that vaccination is a particularly sensitive topic among orthodox Protestants, we expected our respondents to recall at least whether or not their children are vaccinated against MMR. Nevertheless, recall inaccuracy is a possible limitation on the present study.
When the vaccination coverage of 35% among our respondents is compared to the registered vaccination coverage for the village (44%), unvaccinated respondents appear to be overrepresented. This overrepresentation can be explained, however, by the participation of students from orthodox protestant schools and their siblings who live in other villages.
The outcome variable in the present study was the clinical diagnosis of mumps. As mumps is generally construed to be a mild disease, only a minority of patients consult their GPs with regard to symptoms. Our case definition was therefore based upon clinical assessment by the parents while it is known that 30% of cases of mumps infection go without symptoms . The real amount of mumps may therefore be underestimated in the present study, but such underestimation should apply to all schools and therefore not affect our comparison of the schools.
In closing it should be noted that as part of our recruitment strategy, households with only children under four years of age or only high school students were not included in the study. In orthodox protestant families, mothers are supposed to stay at home to care for their children, which means that transmission via day care centers that are rarely frequented by orthodox Protestants is not very likely. Transmission among high school students may, however, be more important in the spread of the epidemic, particularly during the early stages . While the presence of high school students in a family with elementary school children did not influence the hazard of mumps, we cannot exclude the possibility that high school students played a role in the initial introduction of mumps into the village.