In this prospective cohort study, participants were recruited between March 28 and April 17, 2020, through social media and nationwide media coverage. Eligible study participants were volunteers not tested for SARS-CoV-2 at the time of recruitment 18 years or older, had a Norwegian identification number and electronic access to the secure national digital governmental identification service. All 122,453 participants signed an electronic consent form and completed an online baseline questionnaire detailing demographics, the use of public transport and other possible risk factors for SARS-CoV-2. The use of public transport was one of many other questions on risk factors. The initial lockdown period in March 2020 lasted 6 weeks and involved closure of kindergartens, schools, gyms, bars, restaurants and major cultural and sports events. After that, there were no additional national school closures, and the population was, largely, only advised on social distancing, with restrictions reinstated from November 5, 2020.
The use of public transport was defined as the use of public or commercial buses, trams, ferries and/or trains. All participants were asked how many times (1–3 times, 4–10 times and 11 times or more) during a 2-week period they used public transport (Additional file 3: Appendix). At baseline (March 28, 2020), there were separate questions regarding the use of public transport before and after March 12, 2020, whether participants had been standing (due to lack of seats) and if they travelled during rush hour. Norway’s initial lockdown started on March 12, 2020.
The outcome was a SARS-CoV-2 positive nasopharyngeal or oropharyngeal swab test determined by real-time polymerase chain reaction. The test was obtained from any accredited Norwegian microbiology laboratory and the test result was reported through the Norwegian Messaging System for Infectious Diseases (MSIS). We included only tests at a time point later than the date of the baseline questionnaire and before January 28, 2021. In Norway, it is mandatory to report all cases of SARS-CoV-2 infections to MSIS. The proportion of new positive SARS-CoV-2 tests by day in Norway at the different time points can be found in Additional file 1: Fig. S2.
We defined potential confounders to be age (5 years categories, missing), calendar time (date of questionnaire, continuous), sex (men, women, missing), income (NOK per household and year, below 299,999, 300,000–599,999, 600,000–1,000,000, more than 1,000,000, missing), fitness (very fit, fairly fit, in bad shape, missing), smoking habits (never, former, current, missing), underlying medical conditions (no, yes, missing) and municipality (358 different municipalities, missing).
We excluded participants with missing on the use of public transport at baseline. Missing on covariates was included as a separate category in each covariate.
In the current study all participants were untested at baseline in order to avoid recall bias and self-selection bias (difference in agreement to participate) between SARS-CoV-2 positive and non-positive participants. Outcome status was obtained from accredited laboratories in order to avoid misclassification of the outcome.
Because of the small losses to follow-up and the low percentage of SARS-CoV-2 infected [10, 11], cumulative incidence was used. The association between the use of public transport before and after the initial lockdown period and subsequent contraction of SARS-CoV-2 was investigated using logistic regression. All individuals who had not contracted SARS-CoV-2 by January 27, 2021, were included as controls. We estimated odds ratios (ORs) with 95% confidence intervals (CIs) adjusting for age, calendar time, gender, municipality, smoking habits, income level, fitness and underlying medical conditions. Trend test was performed by fitting ordinal values corresponding to exposure categories and testing whether the slope coefficient differed from zero. All analyses were performed using Stata (Stata Statistical Software, release 16, Stata Corp., College Station, TX) and R (version 3.6.2). A two-sided p-value of less than 0.05 was considered statistically significant. Sensitivity analyses were performed in health care workers and in non-health care workers, and by sex.