This paper was written according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement (Additional file 1: Table S2) [30].
Study design
We used data from a repeated cross-sectional population-based survey of social contacts (CONNECT—CONtact and Network Estimation to Control Transmission) conducted before and during the COVID-19 pandemic in Quebec, Canada. The detailed methodology of CONNECT has been described previously [3]. The first phase was conducted from February 2018 to March 2019, 1 year before the COVID-19 pandemic. In order to document the changes in social contacts during the pandemic, additional phases were undertaken (April 21–May 25, 2020, and July 3, 2020–July 4, 2021). The same methodology was used for all CONNECT phases.
Recruitment of participants
All non-institutionalized Quebecers without age limits were targeted for CONNECT recruitment. We restricted the current analyses to participants aged 18 years or older because they represent the vast majority of COVID-19 hospitalisations and deaths [7, 8]. We recruited participants by random digit dialling, using landline and mobile phone numbers. First, we explained the study to the respondent, verified the household eligibility, and documented the age and sex of all household members. Then, we used an age-stratified probability sampling to randomly select one individual per household to participate in CONNECT. We recruited new participants for each phase of CONNECT using the same procedure.
The CONNECT study was approved by the ethics committee of the CHU de Québec research center and participants gave their consent to participate in the study during the recruitment phone call. We commissioned the market company Advanis for the recruitment of participants and data collection.
Data collection
We used a self-administered online questionnaire for data collection, and the same questionnaire was used for all CONNECT phases. After the selected participants gave their consent to participate in the study, we sent them an email containing a secured individualized web link to the questionnaire and additional information about the study.
In the first section of the questionnaire, we documented key socio-demographic characteristics (e.g., age, sex, region, household size, education level, country of origin, race/ethnicity, main occupation) and the health condition of participants. Specifically, participants were questioned about any long-term health condition(s), which is expected to last or has already lasted 6 months or more, diagnosed by a health care professional they may have or have had. For each condition reported, they were asked whether they have had symptoms or taken medication in the past 12 months (Additional file 1: Example of questions S1). The second section of the questionnaire was a social contact diary, adapted from Polymod and other similar studies (Additional file 1: Example of questions S2) [31,32,33]. We assigned each participant two random days of the week (one weekday and one weekend day) to document every different person they had contact with between 5 am and 5 am the following morning. We defined contacts as either physical (handshake, hug, kiss) or nonphysical (two-way conversation in the physical presence of the person at a distance equal to or less than 2 m, irrespective of masking). Participants recorded in the diary the characteristics of the contact person (age, sex, race/ethnicity, and relationship to themselves) and the characteristics of the contact itself (location (home, work, school, public transportation, leisure, other locations), duration, frequency, and whether the contact was physical or not). When participants reported having more than 20 professional contacts per working day, we asked them general questions about these professional contacts (age groups of the majority of contact persons, average duration of contacts, and whether physical contacts were generally involved or not) rather than reporting each professional contact in the diary. Since the beginning of January 2021, we asked participants whether they have been vaccinated against COVID-19 and, if so, the date(s) of vaccination.
Main outcome and exposure variables
Our main outcome was the mean number of social contacts per person and per day, for all locations combined. We weighted contacts reported on weekdays (5/7) and the weekend (2/7) to represent the mean number of social contacts per day over a week. To distinguish between contacts at home with household members from the other contacts that could be influenced by physical distancing measures, we also stratified contacts into two groups: (1) contacts at home with household members and (2) contacts at home with visitors and contacts in all other locations. If a contact was reported both at home and in another location, we only considered the contact at home, where the risk of transmission is higher, to avoid counting multiple contacts with the same person. As in other previous studies that limited the number of contacts per day [31, 32], we truncated professional contacts to a maximum of 40 per day to eliminate contacts at low risk of infectious disease transmission and extreme values.
Our main exposure variable was the presence of an active physical comorbidity, that is a long-term physical health condition for which participants have had symptoms or taken medication in the past 12 months (excluding mental health conditions). We restricted the sensitivity analyses to active physical comorbidities that were shown to increase the risk of COVID-19 complications according to two analyses, one from the Institut National de Santé Publique du Québec (INSPQ) and another one from the National Advisory Committee on Immunization (NACI) (Additional file 1: Table S3 and Figure S1) [6, 34]. Active physical comorbidities are referred thereafter in the text as comorbidities.
For the vaccination status, participants were considered vaccinated if they reported receiving their first dose of the COVID-19 vaccine before or at their assigned day for the study. Although randomized control trials have shown high vaccine efficacy 14 days after vaccination [10, 18], we did not consider an interval after vaccination because anecdotal reports suggested that individuals may feel protected immediately after vaccination [35]. The two assigned days were considered separately for the vaccination status, meaning that if participants were vaccinated in the interval between their first and second assigned day, they were considered vaccinated only for the second day.
Analyses
We weighted CONNECT participants by age, sex, region (Greater Montreal and other Quebec regions), and household composition, using the 2016 Canadian census data of Quebec [36]. We performed all analyses using weighted generalized linear models with a Poisson distribution, and an identity link was used to obtain mean differences. To take into account the correlation between the two assigned days of each participant and data overdispersion, we used generalized estimating equations with robust variance [37]. Participants with inadequately completed social contact diaries or with missing values for comorbidities were excluded from the analyses.
For the main analysis, we compared the mean number of social contacts per day of individuals with and without comorbidities, for the pre-pandemic period and the first three waves of COVID-19. Periods were determined according to COVID-19 epidemiology in Quebec [38] and CONNECT data availability: pre-COVID in 2018–2019; the first wave from April 21 to May 25, 2020; summer 2020 from July 3 to August 22, 2020; the second wave from August 23, 2020, to March 20, 2021; and the third wave from March 21 to July 4, 2021. The second wave was further stratified to consider the Christmas holidays separately: before the holidays from August 23 to December 16, 2020; holidays from December 17, 2020, to January 8, 2021; and after the holidays from January 9 to March 20, 2021 (Fig. 1). We performed the analyses for all adults and stratified by age (18 to 65 and over 65 years old). We considered the following potential confounding variables: age, sex, region, household size, education level, main occupation, race/ethnicity, and country of origin. Age was the only confounding variable identified using the change in estimate method with backward selection [39, 40]. Therefore, we adjusted analyses for age (18 to 25, 26 to 45, 46 to 65, over 65 years old). Stratified analyses on the subgroup of participants aged over 65 years were not adjusted since there were too few participants for the adjustment. In sensitivity analysis, we replicated the same analysis, but using the classifications of comorbidities at risk of COVID-19 complications from INSPQ and NACI. Given that comparisons of social contacts between individuals with and without comorbidities represent a secondary analysis of the CONNECT study, we calculated the power using the number of participants available for these analyses and the proportion with comorbidities. For example, with a sample size of 1200 individuals, 22% of which having comorbidities, it is possible to detect a difference of 0.9 contact (3.0 vs 3.9) with 85% power and 5% two-tailed type I error, taking into account a design effect of 5 due to overdispersion and correlation between days (Additional file 1: Table S4).
For the secondary analysis, we compared the mean number of social contacts per day of individuals with and without comorbidities during the third wave, according to their vaccination status. This analysis was stratified by age and adjusted for time periods (March, April, May, June–July) since vaccination status and contacts varied substantially during the third wave as a result of the general population vaccination roll-out starting in March 2021 and gradual relaxation of physical distancing measures (Fig. 1). All statistical analyses were performed using SAS version 9.4.