Infection and transmission risks in schools and contribution to the COVID-19 pandemic in Germany – a retrospective observational study using nation-wide and regional health and education agency notification data

INTRODUCTION

Schools have been a key target domain for nonpharmaceutical interventions (NPIs) in the SARS-CoV-2 pandemic. However, infection dynamics in schools, their impact on the wider population and the effect of NPIs remain ill-defined. Measures include isolation and quarantine, masking and testing, hygiene measures, ventilation and social distancing, including reduced attendance class models and school closures(1-3).

In Germany, the 16 federal states hold sovereignty over education policy and in large parts health system management. During the SARS-CoV-2 pandemic in Germany, schools were first closed on March 17, 2020, nationally and gradually reopened in May 2020. Subsequently, federal states instituted diverse NPIs around social distancing, cloth masking and reduced class sizes. Full school closures were implemented again nationally on December 16, 2020, until reopening under federal state-specific NPIs in February/March 2021. In February 2021 the German Association of Scientific Medical Societies published an evidence-based guideline on NPIs to reduce transmission of SARS-CoV-2 in schools(4). The variability of measures between federal states diminished but remained in some parts such as masking of primary school students or masking of students during class. Different testing strategies (e.g. antigen testing, pooled PCR testing, mandatory/voluntary) for students and teachers were implemented in spring 2021. For a third time school closures, now depending on local infection dynamics, were instituted in April 2021 along with mandatory testing in schools nationally. All federal states returned to face-to-face classes and reduced stringency of measures prior to the summer vacation 2021. In September 2021 the adult population in Germany surpassed a vaccination coverage of 60%(5). The National Advisory Committee for Immunisations (Ständige Impfkommission, STIKO) issued a vaccine recommendation for adolescents from 12-17 years of age and for children at risk in June 2021 and a general recommendation for all adolescents in August 2021(6).

Whereas school closures have proven effective in containing SARS-CoV-2 transmission(7-9), they are associated with adverse effects on the physical and mental health of students and exacerbate ethnic and socioeconomic disparities(10-14). Legislators implemented NPIs and testing in schools to reduce setting specific transmission(15) and avoid closures. Evidence of effects of measures was initially derived from modelling studies or indirectly from non-school settings(3). The sovereignty of federal states over education policy resulted in 16 different and parallel approaches. This allows for comparative evaluation of infection dynamics and the effects of NPIs in schools.

The risk of infection and transmission in schools relative to the population remain contentious. In a systematic review(16) we showed that during phases with low population incidence, there is no increased risk of infection in students or school staff, however with increasing infection dynamics in the population, the risk of infection in students and staff increases. In Germany and Europe, studies on school transmission risk and contribution to overall transmission in the population are based on geographically and temporally limited data (3, 17-20), with few exceptions(21).

In Germany, data on infection dynamics is collected at the county level by over 400 public health agencies and reported to the national Robert Koch Institute (RKI). Detailed data on (school) setting-specific infection dynamics, contacts, and their risk of infection are not reported to the RKI and can only be obtained individually from the 400+ public health and education agencies. Thus, studies investigating school setting transmission in Germany are limited to small geographic units and short time periods.

In addition, most current studies remain descriptive and do not account for age- and setting-specific infection dynamics and underdetection of cases(18-20, 22). Both, setting specific transmission risk as well as setting specific contribution to overall infections is crucial to inform pandemic policy on effects of NPIs for schools.

In this study we obtained federal state and county level data for school settings as well as detailed regional data obtained from individual public health and educational agencies to assess infection risk, secondary attack rates, the contribution of schools to the overall transmission as well as the effect of NPIs in schools during the first 18 months of the pandemic in Germany.

METHODOLOGY

Data analysis

We performed a descriptive analysis of infection risk for students and teachers as cumulative risks and crude risk ratios for specified periods of time with 95% Confidence Intervals (CI). We calculated secondary attack rates (SAR) as percentages with 95% CI of those being identified as notified infections in all known contacts. We adapted the pandemic phases as proposed by the RKI(26) for school policy intervals (Table 1). An overview of population incidences across phases is in the supplement (S4 Figure).

View this table:

• View inline
• View popup
• Download powerpoint

Table 1

Overview of the Pandemic Phases and Outline of School Policies

We investigated infection control measures and other factors influencing infection risk of teachers or students in a multiple linear regression model. We used an ordinary least squares (OLS) model to test for statistical associations of the official active cases per 100000 students and teachers as reported by the KMK for the county level on the two-week-lags of the attendance rates (by educational level), the stringency of mask mandates (Table 1 in S3 Text) and the mandate and stringency of testing in schools. We controlled for the two-week incidence per 100000 inhabitants of the corresponding county, the percentage of fully vaccinated persons in the corresponding federal state, the socioeconomic status and geography of the county (S3 Text, S11 Text). Furthermore, we tested whether the school specific NPIs had an over proportional effect on the school population compared to the overall population. We proceeded similarly to the first approach, yet did not include a control group as an explanatory variable but instead divided the active cases among the students or teachers, respectively, by the 14-day incidence of the total population of the district. This way we estimated the effect of the NPIs on cases in schools relative to cases in the population, thus estimating whether the NPIs in schools have a significantly different effect on the overall population as on the school population. Further information on the regression analyses are in the supplement (S5 Tables, S3 Text).

The infection dynamics in schools were estimated using a SEIRS (Susceptible-Exposed-Infectious-Recovered-Susceptible) model previously described(27). This distinguishes between healthy but susceptible individuals, those infected but not yet infectious (exposed), and symptomatic and asymptomatic patients. In addition, we included compartments for hospitalisations, patients entering intensive care units (ICUs) and persons with long-COVID, i.e. those who continue to have sequelae after recovery. In the final state, the patients are recovered or dead. Furthermore, we assume a reinfection process. We split the recovery compartment into a compartment for those recovered from COVID and a long-COVID compartment, since we assume that both have a different reinfection rate. Fig.1 illustrates the model as a flow chart.

• Download figure
• Open in new tab

Figure 1 Compartmental SEIR model

An individual in age group i is classified either as susceptible (S), exposed (E), asymptomatically infected (IA), symptomatically infected (IS), hospitalized (H), in intensive care (U), suffering under long-COVID (L), fully recovered (RF), recovered from long-COVID (R_L), and dead (D). P₁ − P₈ are health state transition rates. The Greek letters quantify transition probabilities between the statuses.

The remaining parameters represent time periods of transitions between the different statuses as well as probabilities for the transitions and are estimated from international data and literature research. The data for the model is composed of reports from KMK(23), RKI (25) and the DIVI Intensive Care Register(28). We accounted for age-specific underdetection taking pandemic period- and age-specific underdetection ratios from a large seroprevalence study in Germany(29).

RESULTS

Risk of infection for students and teachers in Germany according to nation-wide data

Data obtained from KMK was collected on the federal state level for W46-50/2020 and W08-40/2021 (at time of analysis), covering 304676 student and 32992 teacher cases. Cumulative risk of notified SARS-CoV-2 infection for students was 0.96% W46-50/2020, 1.22% in phase 4, 1.16% in phase 5 and for teachers 1.77%, 1.53% and 0.47% respectively. At the county level in W09-25, infection risk was 1.68% for students and 1.51% for teachers. Quarantine risk of students and teachers over federal states varied widely from 1-21% during all phases of the pandemic (Table 2, Table 1 in S7 Tables).

View this table:

• View inline
• View popup

Table 2

Risk of infection over pandemic phases with mimima and maxima risk per federal state/ county

On the federal state level, the crude infection risk ratio for students to the population showed an infection risk similar or lower to the general population in W46-50/2020. It was lower in phase 4 and highest in phase 5, where it exceeded that of the general population (Fig.2a). For teachers, the crude risk ratio is highest in W46-50 2020. It is lower in phase 4 and is similar or lower in phase 5 (Fig.2b). The crude infection risk ratio relative to the population is higher for teachers than students in W46-50/2020 and higher for students than teachers in phase 5.

• Download figure
• Open in new tab

Figure 2 Crude risk ratio of (a) Students to Population and (b) Teachers to Population

Crude risk ratio of infection of students (a) and teachers (b) compared to the infection risk of the population in the respective federal state and time interval. A crude risk ratio of 1 indicates equal risk of infection.

Risk of infection according to regional data

Data obtained from local authorities covered a total general population of 3970903, with 545409 (13.7%) students and 60401 (1.5%) teaching and other staff. We included 15433 index cases, 49131 contacts and 4637 secondary cases reported during the school year 2020/21 across 5 regions of Germany. Further information is in the supplement (Table S1).

Infection risk in phases 2-4 ranged from 2-7.6% for the general population, 1.3-5.8% for students and 2.4-3.2% for staff (Table 2). Infection risk was lower for students and staff than the general population during periods of school closure (Figures 1-7 in S6 Figures). Student infection risk increased with student age. The under 10-year-olds show a lower risk of infection than the 10-to-14 and 15-to-19-year-olds in most phases and regions (Fig. 3). Infection risk was higher for students in more advanced school forms with data available from 2 regions (S8 Text). Staff infection risk was similar across age groups. There was a trend towards higher infection risk in special needs and primary school staff and lower risk in secondary school staff (S8 Text). Further data is in the supplemental (Table 2 in S7 Tables, Tables 1-7 in S6 Figures, S8 Text).

• Download figure
• Open in new tab

Figure 3

Infection risks of students by age in four regions

Risk of infection after contact with a notified case (secondary attack rates)

Overall – irrespective of contact setting - risk of infection after contact with a notified case (SAR) for the whole observation period ranges from 4.6-12.8% across all included counties. For the different phases, ranges are 4.1-16.7% (phase 2), 4.9-9.1% (phase 3a), 5.3-19.4% (phase 3b) and 5.4-19% (phase 4). SAR increased with index as well as with contact age with data available for four regions (S9 Tables, S10 Text).

For all phases in all regions, 15-to-19-year-old student index cases had a higher SAR than under 10-year-olds and in several regions higher than 10-to-14-year-olds. SARs increased also with contact age and with more advanced school forms (S9 Tables, S10 Text).

Data from one region allows differentiation of SARs by contact settings. Whereas overall most contacts (64%) were recorded for schools as the contact area, most secondary cases (84%) were reported in household contacts (Fig. 4). Total SARs are 1.2% for contacts in schools and 23.2% for contacts in households, with 8.2% for all contact areas (S9 Tables). Regarding pandemic phases, household SARs in students are 1.5 to 2 times higher in phase 4 than phase 2 and phase 3a, whereas school specific SARs do not change significantly (Fig. 5).

• Download figure
• Open in new tab

Figure 4

Contacts and Secondary Cases in Phases 2-4 (W21 2020 to W23 2021) (R1)

• Download figure
• Open in new tab

Figure 5:

School and Household SAR in Students and Staff over Time

Absolute and proportional contacts across contact areas changed over time. Whereas in phase 3a the overall SAR was 4.9%, with schools at 1.2% SAR and 74.8% of contacts reported, and the household SAR at 19.1% with 20% of contacts, in phase 3b (school closure) the SAR changed to 16% overall as the share of household contacts was now 65.6%. Average reported contacts remained similar, with 10.6 (phase 3a) and 10.4 (phase 3b) in schools and 2.97 (phase 3a) and 2.77 (phase 3b) in households. In total, 2.2 times more contacts were reported in schools, whereas 8.8 times more secondary cases originated in households relative to schools (S9 Tables, Fig. 4 and 5).

Contact and secondary case data for staff was available from one region. Overall SAR for staff was 8%. Household SARs were 1.5 to 2 times higher in phase 4 than phase 3a and phase 2. School-specific SARs of staff members were higher for phase 4 at 4.3% compared to 1.4% in phase 3a and the average of 1.5% (Fig. 4).

Among school contacts of staff, 14% were over 18 years old. 57.7% (30 out of 52) of secondary cases of infected staff in schools occurred in this age group with a SAR of 4%, compared to 0.8% in the under 10-year-olds and 1% in the 10-to-18-year-olds.

Contribution of contacts in schools to overall transmission in the population according to nation-wide data from educational agencies

We used the compartmental model described in Fig.1 and data from the KMK. Not accounting for underestimation of infections, we found a high variability in the contribution of contacts from school infections to the overall transmission during the third wave from 2-12%. Accounting for age-specific underestimation of notified infections in comparison to actual infections based on seroprevalence estimates, this range was 2-13% (Fig.6). The contribution was lowest during vacations or school closures. The peaks occurred with open schools. The W24 peak coincides with a 7-day average population incidence of 6 cases per 100000 inhabitants and the W40 peaks occurs at 65 per 100000 7-day population incidence. The local peak at W12 occurs with reduced presence class models and a population incidence of 130 per 100000 inhabitants.

• Download figure
• Open in new tab

Figure 6

Estimated force of infection and contribution of contacts of infected persons in school to overall transmission in Germany week 9-39 2021, accounting (c&d) and not accounting for underdetection (a&b). Pink areas represent periods when the majority (at least 7) federal states have school vacations. Vacation periods in federal states in Germany differ by multiple weeks and will span periods prior and post to the areas marked. W16-21 include county-specific incidence-dependent school closures which affected less than 50% of counties in Germany

DISCUSSION

In this study of infection and transmission risks during the pandemic in Germany, we could show that the contribution of infection dynamics in schools to that of the general population was up to 13% under hygiene and testing measures. Furthermore, NPIs such as strict mask mandates (in all schools during class) and reduced presence models are associated with an effective reduction in infection and transmission risks in schools and their contribution to population infections.

Most infections and transmission in school took place in phase 3a when schools operated in normal class models with some cloth mask rules and general hygiene concepts. Teachers showed a higher risk of infection than the general population. Infection risk for students increased with age, as previously shown(17,19). During winter 2020/21 school closures (phase 3b), total reported infections and transmission in schools were low and infection risk for both students and staff was lower than the general population. Meanwhile, more time spent in the household lead to longer exposure and thus a higher SAR in household members, yet fewer secondary cases in total.

In phase 4 schools reopened with stricter NPIs including reduced presence class models and stricter mask mandates with testing and incidence-dependent school closures introduced later on. Meanwhile, the prevalence of the more transmissible B.1.1.7 (alpha) variant rose in Germany. Infection risks for students and teachers were similar to that of the population. School-setting SAR of both students and teachers remained low (<5% during phases 2-4), whereas household SAR more than doubled. The stricter NPIs likely counteracted the higher transmissibility of the alpha variant in schools.

This hypothesis is supported by our regression analysis. We estimate a reduction by 24-30% of infection risk for both students and teachers relative to the population and by more than 55/100.000 per 14 days in absolute case numbers in those counties implementing mandatory surgical mask wearing during class in all schools. Similar observations have been made in Arizona, USA, where the odds of school-setting outbreaks of SARS-CoV-2 were 3.5 times higher in schools without a mask policy(32).

Furthermore, all interventions reducing attendance rate of students (rotating and distance classes as well as school closures and vacations) reduced the risk of infection of students and teachers. Concurrently, we see a higher risk of infection for students and teachers in counties with higher deprivation indices. This was known for Germany, though not specifically for schools(33, 34).

These findings are reflected in our analysis on the contribution of schools to the general population infection dynamics. School vacation and closures reduced this contribution to as low as 2%. Previous modelling for the first and second wave in Europe showed that school closures resulted in >30% and <15% reductions of transmission(8, 35). The peak at 12% contribution in W24 occurs with open schools, loosening population NPIs and loosening but still strict hygiene and testing measures in schools. Meanwhile, the lower peak at 7% early in phase 4 occurred during reduced attendance class models and stricter NPIs.

During phase 5 we observe a continuation but loosening of mask and testing mandates with full attendance classes in a low population infection environment. The crude infection risk ratio of teachers relative to the population in phase 5 remained steady while it was higher for students than the population. This is likely due to vaccination coverage rising in the adult population and mandatory testing in schools. Whereas teachers as a vocational group were vaccinated with priority, vaccination for students age 12-17 was not available until W33. Our regression analysis showed the protective effect of higher vaccination rates in the population as well as a higher number of notified cases among students with mandatory testing, likely through reducing underdetection. Summer vacations reduced the contribution to population infection dynamics to a trough of 3%. Peaks occurred at 13% during open schools in normal presence classes.

We further found that SARs are higher both with higher age of the of index cases as well as contacts. SARs have previously been reported only for early phases of the pandemic from German public health agency data, showing SARs below 2%(17). The household SARs observed are in line with previous findings for the general population(36). The majority of secondary cases of both students and staff happened in households. In schools, age-specific SARs show that transmission took place mostly from staff to staff and from student to student more than student to staff. This is in line with previous findings(16).

Interestingly, in sensitivity analyses (S11 Text) of the regression analysis, we find a decrease in infection risk of primary school students when attendance in school increased, whereas attendance is positively correlated with infection risk for all other student and staff groups. Previous studies showed a similar effect for the end of vacations(37). The infection risk in school for primary school students is possibly lower due to stricter adherence to the rules in schools and by parents and less time spent in households with a higher transmission risk.

Limitations of our work are inherent in the notification process to both public health and educational agencies as well as the gathering of aggregate county-specific data (S12 Text). Both notification data itself as well as contact data is an underestimate of the actual infection dynamic. We attempted to account for underdetection with age-specific estimates taken from seroprevalence studies. This could make our estimate of contribution of school-contacts to overall transmission a potential underestimate. Residual confounding in the regression analysis is a possibility as we use aggregate county or school measures and did not have full access to individual-level confounding factors, e.g., on distribution of parental professions or industrial make-up of the counties included. We tried to include the most important confounding factors on infection dynamics in the population as well as deprivation and urbanity of counties in the analysis. SARs are limited by several factors. Contact tracing is inherently limited by both interviewer capacity and ability and interviewee memory and honesty, however with schools and households as contained domains the error margin is limited. Among contacts, the data did not indicate whether these were still susceptible to infection. However, it can be assumed that biases are similar in each region as they are situated within the same country and timeline of events. A reduction in the proportion of the susceptible population through vaccination or infection is assumed relevant from March 2021, where it is included as a parameter in regression analysis. All considered, the SARs presented here described further transmission after cases have been detected in the school with limitations.

Despite these limitations we conclude that school setting contribution to overall transmission in the population is relevant, but variable over different time periods in the pandemic and responsive to NPIs. In Germany, school-related NPIs, in particular masking and reduced attendance, have been successful in mitigating the spread of the virus among both students and teachers.

FUNDING STATEMENT

This project received financial support by the Standing Conference of Ministers of Education and Cultural Affairs (KMK) in Germany. TH, BL, PV, IR, and GK received funding from the European Union’s Horizon 2020 research and innovation program (101003480; Project CORESMA) and from the Initiative and Networking Fund of the Helmholtz Association. BL and SG received funding from the BMBF Netzwerk für Universitätsmedizin (NaFOUniMedCovid19, FKZ: 01KX202; Project COVIM). AJ, JD received funding from the BMBF (Bundesweites Forschungsnetz "Angewandte Surveillance und Testung” (B-FAST); AP5: Anwendungsbereich Schulen und Kitas).

ETHICS STATEMENT

Ethics approval was obtained by the IRB of the Medical University Hannover (Ethical approval N°9609_BO_K_2021, Hannover). Data was collected by public agencies in all cases in line with public regulation on data privacy. Data was anonymised prior to transfer. Data was analysed anonymously. Data study complies with EU-GDPR as it is a data evaluation for research in the public interest (§ 89 EU-GDPR).