This study used data of individuals aged 20+ from the Corona Immunitas digital follow-up cohort (eCohort) which is part of the Corona Immunitas seroprevalence study, described elsewhere [30, 31]. The source population for the eCohort consisted of participants from the Corona Immunitas study who were randomly selected from the general Swiss population. At enrollment into the Corona Immunitas seroprevalence study, participants completed an entry assessment covering socio-demographic and health factors. Enrollments into Corona Immunitas occurred between June 2020 and February 2021, depending on site [31]. Enrolled participants aged 20+ were also systematically invited to join the longitudinal eCohort if they had a working email address and internet access. Compared to the Corona Immunitas cohort, participants enrolled in the eCohort were more likely to be older, to have a Swiss citizenship, to have a university diploma, and to have greater household income than the general Swiss Population [31]. Further details on eCohort recruitment, survey response and survey topics can be found elsewhere [31].

Our analysis used data from 11 study sites representing all language regions in Switzerland (i.e., Bern; Basel-Landschaft; Basel-Stadt; Fribourg; Graubünden; Luzern; Neuchâtel; St. Gallen; Ticino; Vaud; and Zürich). We included information from the Corona Immunitas entry assessment, as well as data from the monthly eCohort follow-up surveys. These follow-up surveys collected information on SARS-CoV-2 infections, mental and physical health, healthcare utilization, questions on perceived consequences of the pandemic, perceptions of personal and societal risks concerning the pandemic; and use of the Swiss digital proximity tracing app (i.e., SwissCOVID) [32]. Starting in February 2021, the monthly surveys also included additional questions about SARS-CoV-2 vaccination status, as well as questions on attitudes and hesitancy towards vaccination, which were only presented to people who had not yet received a vaccine dose. Additional data on personality traits were collected using a separate survey at a single time point between March 2021 and April 2021, depending on eCohort site.

This analysis included participants who filled out at least one vaccine status assessment as part of an eCohort follow-up survey between 1 September 2021 and 31 December 2021 (Figure 1). The analysis baseline was set for each individual at the first available vaccination status survey. Baseline information included age, gender, education, number of health conditions, work status, income, and language region from the entry assessment for Corona Immunitas, which needed to be non-missing for inclusion in analysis. Moreover, attitudes toward vaccination and vaccine hesitance were taken from the monthly survey that defined the individual analysis baseline. The earliest possible baseline date was 1 February 2021 (when the first monthly vaccination survey was released), and the maximally possible analysis follow-up duration was 11 months (i.e., from February to December 2021).

All characteristics were self-reported.

Vaccination status (1 = Non-vaccinated; 0 = Vaccinated) was self-reported by participants at each monthly follow-up. We created a variable Vaccinated which means having received at least one vaccine dose by the end of 31 December 2021.

Socio-demographic and personal characteristics were reported in the Corona Immunitas entry assessment. Age was expressed in years. Gender comprised three categories: men; women; other. Education comprised three categories: primary (i.e., No school certificate; Mandatory school); secondary (i.e., Apprenticeship); and tertiary education (i.e., Higher technical school; University of applied sciences; University). Household monthly gross income comprised four categories: Swiss Francs, CHF: 0–6,000; >6,000–12,000; >12,000–18,000; >18,000. Language spoken in the study site was coded as German (Bern; Basel-Landschaft; Basel-Stadt; Graubūnden; Luzern; St. Gallen; Zürich; and 1/3 of the sample in Fribourg); French (2/3 of the sample in Fribourg; Neuchâtel; Vaud); and Italian (Ticino). Work status comprised three categories: 1 = Employed; 2 = Retired; 3 = Not employed and not retired.

For both vaccinated and non-vaccinated participants, we descriptively analyzed socio-demographic and personal characteristics, health-related variables, vaccination-related beliefs and attitudes, vaccination acceptance, and personality traits at baseline. We used univariable and multivariable logistic regression models to estimate the likelihood of remaining non-vaccinated until 31 December 2021 according to analysis baseline socio-demographic and personal characteristics, health-related variables, vaccination acceptance, and personality traits. Based on evidence from other studies, age, gender, education, and number of health conditions were included a priori as predictors of vaccination status in the multivariable logistic regression model [12–19]. Each additional variable from demographic and personal characteristic, health-related variables, vaccination acceptance, and personality traits was added to this model sequentially and maintained in the model if the Akaike’s Information Criterion (AIC) decreased by two or more units [38]. As some questions were only answered by a subsample of participants, we calculated three multivariable models for socio-demographic and personal characteristics (model 1/basic model); health-related variables and vaccination acceptance (model 2, building on top of model 1), personality traits (model 3, building on top of model 2). In each model complete case analysis was used. A sensitivity analysis was also conducted using the sample size of model 3 to repeat estimation of models 1 and 2 (see Supplementary Tables S1, S2). Descriptive statistics reporting the socio-demographic characteristics of the overall study sample and of the subsample of participants with full data in all study variables are reported in Supplementary Table S3. Univariable logistic regression models were also fit to report the association of vaccination status with each of the investigated variables. We report odds ratios (OR) and 95% confidence intervals (CIs). Statistical significance was set at a two-tailed value of p < 0.05. For categorical variables in the models, the first category was chosen as reference. Data analysis was conducted using Stata version 17 [39].

Of the 13,348 eCohort participants, 10,642 (79.7%) had included at least one survey with vaccination questions between 1 September 2021 and 31 December 2021, thereby contributing data to this study. Of these, 1,323 (12.4%) were non-vaccinated by 31 December 2021 (Figure 1).

For our complete case analysis, 9,582 participants were included in Model 1 (basic model) as they had non-missing information on age, gender, education, number of health conditions, work status, income, and language spoken in the study site questions. Model 2 included a subsample of 6,596 individuals who answered all questions assessing worry for one’s health, prior positive SARS-CoV-2 test in the participant, and vaccination acceptance. Model 3 included a more limited subsample of 2,028 individuals who answered all questions assessing personality traits, which was assessed in the eCohort only in March and April 2021 and hence not available for all study participants.

Vaccinated participants had, at baseline, a mean age of 57.71 years (Standard deviation, SD = 15.94 years; Table 1) whereas non-vaccinated participants had a mean age of 46.10 years (SD = 15.15 years). Men represented 45.5% of vaccinated participants and 40.4% of non-vaccinated participants. Compared to non-vaccinated individuals, a lower proportion of vaccinated participants completed secondary education (45.7% of vaccinated vs. 52.0% of non-vaccinated), but a higher proportion completed tertiary education (49.5% of vaccinated vs. 43.1% of non-vaccinated). A gross monthly income of CHF 6,000 to 12,000 was reported by a higher proportion of vaccinated individuals (47.1%) compared to non-vaccinated individuals (43.8%). The mean number of chronic health conditions were 0.59 (SD = 0.89) in the vaccinated sample and 0.35 (SD = 0.68) in the non-vaccinated sample. Among vaccinated participants, 30.4% were retired and 43.4% were employed whereas among non-vaccinated participants, 15.1% were retired and 60.6% were employed.

Amongst all the socio-demographic and personal characteristics variables investigated, in the multivariable logistic regressions (model 1, analytic sample N = 9,582), compared to vaccinated individuals, non-vaccinated individuals were statistically significantly more likely to be younger (OR = 0.97; 95% CI: 0.96; 0.97) and to have fewer chronic health conditions (OR = 0.86; 95% CI: 0.78; 0.94). Non-vaccinated individuals were also less likely to be retired (OR = 0.56; 95% CI: 0.44; 0.71) and less likely to have a higher monthly income (p < 0.001). Education did not differ significantly between the two samples.

On average, the non-vaccinated subgroup reported lower levels of vaccination acceptance as indicated by their score on the Vaccination Acceptance Index (mean = 14.09; SD = 4.62; Table 2) than the vaccinated group (mean = 19.93; SD = 3.98).

In multivariable logistic regressions (analytic sample N = 6,596; Table 2) adjusted for age, gender, education, number of health conditions, working status, and income, those who were worried for their own health at baseline were more likely to be non-vaccinated than those not at all worried about their own health (p < 0.001). Participants who reported having a previous infection determined by a positive SARS-CoV-2 test were more likely to be non-vaccinated than those who did not report a positive test (OR = 2.18; 95% CI: 1.49; 3.19). Non-vaccinated individuals were also more likely to obtain a lower score on the vaccination acceptance index (OR = 0.77; 95% CI: 0.76; 0.79). Based on the final model, we post-hoc also explored interactions of the vaccination acceptance index with age and sex to examine influences of differences in general attitudes towards vaccinations in different subpopulations. While there was no evidence for an interaction with sex (p > .05), we observed an interaction between vaccination with increasing age (OR = .98; 95% CI: .97; .99; p = .01), suggesting that those who were older were less likely to be non-vaccinated when they had more positive attitudes towards vaccination (higher score on the vaccination acceptability index).

The multivariable logistic regression model (model 3, N = 2,028; Table 3) shows that higher levels of conscientiousness were related to greater likelihood of being non-vaccinated (OR = 1.14; 95% CI: 1.01; 1.30).

Table 4 reports descriptive statistics for vaccination-related beliefs and attitudes at baseline based on the overall study sample (N = 10,642). More than double of non-vaccinated individuals (65.5%) reported (i.e., answered either strongly agreed or agreed) preferring to wait before being vaccinated until more is known about the vaccine’s effectiveness compared to vaccinated individuals (29.1%). Similarly, more than double of non-vaccinated individuals (67.8%) reported preferring to wait before being vaccinated until more is known about the vaccine’s safety compared to vaccinated individuals (31.8%). About half (51.7%) of non-vaccinated individuals feared vaccination side-effects compared to one-fourth of vaccinated individuals (26.7%).

Non-vaccinated individuals were more likely to distrust public health authorities and vaccination in general. For example, 29.7% of non-vaccinated individuals compared to 68.4% of vaccinated individuals trusted manufacturers or pharmaceutical companies. Moreover, whereas 42.5% of non-vaccinated individuals trusted the Federal Office of Public Health, this institution was trusted by 76.5% of vaccinated individuals.

Among non-vaccinated participants, 31.9% thought that SARS-CoV-2 vaccination protects them from SARS-CoV-2 infection and 41.9% thought that vaccination protects them against a severe course of disease. These proportions amounted to 68.6% and 76.8%, respectively, in vaccinated individuals. Moreover, whereas among non-vaccinated participants, 24.3% thought that vaccination protects them against transmission of the SARS-CoV-2 infection to others, 48.7% of vaccinated individuals thought so. Finally, 55.3% of non-vaccinated individuals preferred natural immunity (i.e., infection-induced immunity) against the SARS-CoV-2 infection over vaccine-induced immunity and 39.5% preferred protecting themselves by other means (e.g., wearing a mask and hand hygiene); these proportions amounted to 16.6% and 14.3%, respectively, in the vaccinated group.

In Supplementary Tables S1, S2 we report sensitivity analyses where all study analyses were conducted on the subsample with full data in all variables of interest. Results of sensitivity analyses were consistent with main study analyses except for the association between the number of chronic health conditions and vaccination status which was no longer statistically significant in the sensitivity analyses. As outlined in Supplementary Table S2, the subsample with full data had, on average, the same number of health conditions as the overall study sample. However, the subsample with full data included a higher proportion of individuals who were employed and who had higher household income.

This study has several limitations. First, the proportion of non-vaccinated individuals in this study by end of December 2021 (12.4%) is lower than that reported in Switzerland in January 2023 (about 20%) [4] and therefore overrepresented the vaccination rate of the population. It may be that due to social desirability bias, non-vaccinated individuals were more likely to withdraw from the study. However, representativeness of the sample to the overall population was not a requirement for the current study as its main aim was to identify individual-level factors related to non-vaccination. Moreover, the number of non-vaccinated individuals was high enough to ensure sufficient analytical power in all the tested models. Second, direct comparisons between our models cannot be drawn because of the varying analytic samples due to missing values for some exposure variables. However, in the Supplementary Material we report sensitivity analyses where all study analyses were conducted on the subsample with full data in all variables of interest. Although the subsample with full data included a slightly higher proportion of participants who were employed and with better income compared to the overall study sample, sensitivity analyses did not materially alter our conclusions. Only the association between the number of chronic health conditions and vaccination status was no longer statistically significant in the sensitivity analyses. Third, the baseline for participants was set at different timepoints (depending on the completion of the first vaccination survey), and information on efficacy and safety of vaccines may have changed with time. Fourth, vaccination was self-reported by participants and no confirmation was obtained from health authorities.

Nonetheless, this study has several strengths. First, it included extensive assessments of factors related to non-vaccination. Second, the analyses are based on a large sample of individuals aged 20+ years from a representatively selected source population, although subsequent selection effects (e.g., due to selective acceptance of eCohort study enrollment or follow-up participation) cannot be fully excluded. Whereas several research studies focused on the exploration of factors related to willingness to vaccinate against COVID-19 prior to the availability of the vaccine, this study collected data in a period when the SARS-CoV-2 vaccine was available to all adult age groups in Switzerland and the vaccination campaign was ongoing.

In Switzerland, individuals who are younger, healthier, have a lower income, and/or were previously infected with SARS-CoV-2 were more likely to be non-vaccinated. Individuals who decided not to vaccinate generally had lower confidence in the vaccine’s effectiveness against infection, transmission, and a severe course of the SARS-CoV-2 infection, as well as in the Federal Office of Public Health, and vaccines in general. These findings could inform public health authorities and healthcare providers to target/tailor future vaccination campaigns.