Musician members of professional concert or opera orchestras and professional concert or opera choirs, as well as control subjects, were recruited from participating institutions throughout Germany in a three-arm prospective cohort study. Data collection took place from October 1, 2020, to June 30, 2021, thus mostly in line with the 2020/21 performance season.

Data was collected prospectively via online survey that could be completed via smartphone, tablet, or personal computer. Participants received letters with information about the study and invitations to participate in both German and English. Consent to study participation was given via a link to the online consent form. After giving their consent, the participants received a baseline questionnaire and subsequently follow-up questionnaires at weekly intervals for up to 9 months. To ensure data protection, questionnaires could only be opened via a personalized link that participants received individually by email.

Exposure was defined as working as a professional orchestral musician or choir singer. Exposure was a proxy variable for individuals being exposed to potentially infectious aerosols and droplets generated by, among other things, wind instruments or singing during regular rehearsals and concert performances in professional orchestras or choirs.

The control group included all individuals not exposed to potentially infectious aerosols or droplets caused by orchestral playing or singing. The primary endpoint was the number of SARS-CoV-2 infections confirmed by positive testing during the study period. Secondary endpoints were the incidence of influenza, flu, or other upper respiratory tract infection, and the number of days of sick leave.

All participants had to be at least 18 years old and consent to study participation. Orchestra musicians had to be members of a professional concert or opera orchestra, and singers had to be members of a professional concert or opera choir. Control subjects had to be non-musician employees of the same participating institutions as those of the exposed subjects, and not present in the room during the rehearsal and concert activities of the orchestral and choir musicians. This included various professional groups from administration with no regular contact to the musicians during rehearsal and concert activities but mostly office work. The stage personnel had direct contact with the musicians but regularly not during rehearsal and concert activities. Exclusion criteria were the existence of known infection with SARS-CoV-2 at study inclusion, activity in a string only orchestra, or activity as music student or temporary employee.

At baseline, sociodemographic data (age, sex, number of adults and children in the household, occupation including instrument or voice specialty), and health-specific variables being risk factors for a severe course of the illness, height, weight, e.g. chronic diseases, smoking status and also vaccinations status were obtained [19,20,21].

Subsequently for up to 38 weeks, the occurrence of symptoms of SARS-CoV-2 infection, influenza, flu, or upper respiratory tract infection were collected on a weekly basis for each of the last 7 days. Furthermore, results of testing for SARS-CoV-2 infection (if testing was done), the number of days absent due to illness, and vaccinations given for SARS-CoV-2, influenza, or pneumococcus were recorded. Participants reported the frequency and type of private social contacts and the use of general protective measures in everyday life and at work, as well as specific protective measures of individual musician groups. Furthermore, the length, frequency, location, and sequence of music-making in rehearsals and concerts, possible teaching activities, and tours were assessed. Study participants who tested positive for SARS-CoV-2 were subsequently contacted via email and telephone to obtain information on the suspected infection source and the clinical course of illness, graded following the National Institutes of Health [22].

The highly variable exposure due to the pandemic containment measures was accounted for by the calculation of a weekly professional exposure risk score combining all infection risks of the professional activity, including location, sequence of music-making in rehearsals and concerts, application of general protective measures at the workplace, specific protective measures of individual groups of musicians, possible teaching activities, touring, as well as hygiene concepts at the workplace (room size, ventilation, ensemble set-up, audience concepts, etc.). This score was weighted by each subject’s weekly rehearsal and/or concert time.

A weekly private risk score (range 0 to 28 points) was calculated from the recorded possible confounders, and consists of two parts: 1) confirmed contacts with SARS-CoV-2 positive individuals (contact risk), and 2) other potential risks from the private environment or public space (general risk), including data on vaccinations, frequency and type of contacts with persons at increased risk for SARS-CoV-2 infection, contacts without mouth-nose protection with others, household size and regular contacts with daycare or school children, health care workers and professional teachers, use of public or other transportation, and other general personal risk behavior related to SARS-CoV-2 infection risks. Calculation of the professional exposure risk score and the private risk score and references for the rationales apart from our expert consensus, partly based on former risk calculations [23], are given in detail in the supplementary material table S1.

In addition, the respective hygiene concepts and modalities of performance in the participating ensembles were collected.

All data from the questionnaires were recorded pseudonymously.

For each study group and in total we calculated baseline characteristics for categorical (n; %) and continuous (mean; SD) variables, the cumulative incidence and incidence rate (per year) of SARS-CoV-2 infection, as well as the time at risk (in weeks and years). Cumulative incidence curves were calculated (1 minus survival function) of SARS-CoV-2 infections per study group. The weekly mean of the private and professional exposure risk scores per study group was calculated, as was the weekly proportion of participants with influenza, flu, or other respiratory symptoms and sick leave.

A mixed effects cox proportional hazards model was used to model the effect of the exposure on SARS-CoV-2 infection. The time scale was calendar time and the fixed effects in the model were the exposure group (orchestra, choir, controls) and the mean private risk score (mean score over the entire study period for each subject). The ensembles were considered to be random effects. As a post-hoc secondary analysis, the crude overall exposure (orchestra, choir, control) was stratified by exposure intensity; the orchestra/choir exposures were split into high or low intensity groups based on their respective medians of the mean professional exposure risk score (higher/lower than 22.3 for orchestral musicians, higher/lower than 19.6 for choral singers).

The contact risk has a potentially large confounding effect and adjusting for it helps differentiate between SARS-CoV-2 infections caused by the exposure (orchestra, choir) and infections from non-exposure sources (private contacts). Therefore, to test how sensitive the results are to changes in the contact risk, we included additional information into the contact risk from a follow-up survey performed with SARS-CoV-2 positive individuals that provided information on the suspected source of infection. In further sensitivity analyses, we adjusted either for the contact risk only, for the general risk only, or for both risks separately (two separate variables). All sensitivity analyses used a mixed effects cox proportional hazard model as described above.

The incidence of influenza, flu, or upper respiratory tract infection and the number of sick leave days were modelled using a linear mixed effects model (same fixed/random effects as described above). The respective outcome was defined as the percentage of weeks (with respect to individual study participation time in weeks) that a subject reported influenza, flu, or upper respiratory tract infection as well as the percentage of days that a subject reported sick leave.

All statistical analysis was performed using the R software version 4.1.1. [24].

As part of stakeholder involvement, the seven Berlin concert and opera orchestras (orchestra boards, artistic directors) and the radio orchestras and their choirs were consulted in the planning and development of the study. They helped design the survey instruments and participated in recruiting participants.