Re‐opening schools in Croatia did not have a negative impact on children under 14, but it could not be ruled out in older children

School closures have been used to help tackle COVID‐19. Meta‐analyses and reviews suggest that children are unlikely to be the main drivers of the pandemic and that opening schools and kindergartens is unlikely to increase COVID‐19 mortality rates in older people.¹

Croatia closed its schools on 16 March 2020, and they reopened in week 37, 7 September 2020, after the summer holidays. Public Health guidelines for the reopening included mandatory hand hygiene and face masks for children grades 5 (10/11 year olds) and above. Schools were also told to adopt other social distancing measures, including staggered arrival and departure times and student bubbles. They were also offered three different models for primary and secondary school students: just classroom learning, just distance learning or a mixture of the two. More than 90% of schools chose just classroom learning until the peak of the second wave in mid‐December 2020, when schools individually decided based on local epidemiologic situations for one of the other two options: mixed or distance.

This study investigated the effect that opening schools in week 37 had on incidence trends from weeks 32–42, by analysing COVID‐19 testing data using joinpoint regression. The dependent variable (crude age‐specific rate) was calculated by the program from age and time specific counts and population estimates at the end of 2019. The independent variable was time (week), which was treated like continuous variable. Logarithmic transformation was applied, with the statistical significance for the age‐specific average percentage change per week (APC) set at 0.05. Joinpoint Regression Program 4.8.0.1.was used for the analysis.

During the study period, 20,106 children tested positive, increasing from 343 to 5,031 per week over the study period. The age‐specific rates ranged from 0/1000 for students aged 7–14 years in week 32 to 1.86/1000 for those aged 15–19 years in week 42. The increases in all age groups were 0–6 years (26.0 times), 15–19 (9.3), 20–64 (16.0) and 65 plus (16.4). The age‐specific seven‐day cumulative incidence rates and trends by age group are presented in Figure 1.

FIGURE 1.

Age‐specific seven‐day cumulative incidence rates per 1000 residents by age group

The joinpoint regression analysis revealed a statistically significant increase in the 7–14 and 66 plus age groups, with no change in trends (APC 27.9 and 21.4, respectively; p < 0.001). In week 40 a significant change in trend occurred in the 15–19 (p = 0.011) and 20–65 (p = 0.029) age groups. There were no significant APC changes, or trend changes, in the 0–6‐year group (APC = 9.1, p = 0.110).

Research suggested that opening schools should have had a significant impact on the reproduction number and changes in epidemiologic dynamics after two weeks.² However, our study showed no effect on incidence rates in children aged 0–6 years, primary school children and older residents. The only significant changes were in people aged 15–19 and 20–65 in week 40.

A limitation of our study was that it was hard to separate the impact of opening schools from seasonal changes and non‐pharmaceutical measures. For example, minor measures were in place up to 26 November (week 48), including maximum numbers for gatherings and restricted hours for bar and restaurants. Mask was mandatory for everyone over the age of two, with exceptions in younger children due to potential compliance difficulties. Stricter measures were introduced on 27 November, including significant restrictions on gatherings, limited contact in public places, restrictions on travel and mandatory masks.²

Another limitation was that we only included cases confirmed with polymerase chain reaction. In order to exclude potential impact of testing capacity on incidence rates we also analysed testing rates between 32 and 42 weeks, APC per week was statistically significant and comparable in all groups, and therefore, any impact would have been uniform.

There were no incidence trend changes associated with children under 14 years of age when schools reopened. Although a negative impact could not be excluded in older children and working populations, it had not spilled over to people aged 66 plus.

Primary schools opening in Croatia were associated with lower total years of life lost, due to reduced educational achievements during closure.³ School closures have had a negative impact on children's activity levels, the mental health of parents and children⁴ and parents’ working hours, including healthcare workers. They should be used as a last resort, consider study findings and be implemented with greater effects and less harm.⁵

CONFLICT OF INTEREST

The authors have no conflicts of interest to declare.