The CI-DFU eCohort is part of a nationwide seroprevalence study coordinated by the Swiss School of Public Health (SSPH+) named Corona Immunitas [23]. All participants of the Corona Immunitas seroprevalence study were invited to join the CI-DFU eCohort. The present study is a secondary analysis of the CI-DFU eCohort study and comprised randomly selected adults living in Switzerland aged at least 20 years old, who provided informed consent in writing or online, had a valid e-mail address and had access to the internet. Individuals aged over 65 years were overrepresented by design. Participants were asked to complete weekly, later bi-weekly, online questionnaires on adherence to preventive measures using REDCap [24, 25]. The questionnaires were available in four different languages: German, French, Italian and English. Weekly participation rates ranged between 79% and 88%.

This study included participants of the CI-DFU eCohort living in six cantons (representing three language regions) of Switzerland (Basel-Landschaft, Basel-Stadt, Fribourg, Neuchâtel, Ticino and Zurich), who replied to the follow-up questionnaires between September 1, 2020, and November 30, 2021. Prior to completing the follow-up questionnaires, participants were asked to fill out a baseline questionnaire to assess their demographic and socioeconomic characteristics as well as their adherence to preventive measures at baseline. Adherence to preventive measures at baseline was not assessed in one canton (Ticino, n = 840). Participants completed the baseline questionnaire between June 3, 2020, and February 9, 2021. Supplementary Figure S1 (Supplementary Material) shows the timeline of the questionnaires’ administration together with the COVID-19 pandemic contexts. A flow chart of respondents’ inclusion is presented in Supplementary Figure S2 (Supplementary Material). We excluded participants who had no measurement of adherence to preventive measures during the follow-up period or no measurement for one of our main exposure variables (equivalised disposable income or education).

Our outcome was self-reported adherence to preventive measures, assessed from September 1, 2020, to November 30, 2021. We constructed an adherence to preventive measures score using the following variables, assessed through Likert scales: physical distancing during the previous 7 days, staying at home during the previous 7 days and wearing a mask during the previous 7 days. We considered these three preventive measures as a bundle because, despite the varying economic and social costs associated with each measure, they are all interconnected. Each answer was assigned a number on a 0–5 scale: never = 1, seldomly = 2, occasionally = 3, frequently = 4, always = 5. The score was computed by adding up the value of each variable and ranged from 3 to 15. The score was dichotomized using a cut-off of 12: a score above or equal to 12 was categorized as high adherence, i.e., having replied “frequently” or “always” to all 3 preventive measures’ questions; a score below 12 was categorized as incomplete adherence.

Equivalised disposable income (EDI) and educational level were used as proxies of SES. Our main predictor was the EDI measured at baseline. EDI is a measure of a household’s income, adjusted to account for the size and composition of the household [26]. We decided to use EDI instead of income to better capture participants’ actual economic availability. At baseline, respondents were asked to report their monthly gross household income as 8 categories (0–3000 CHF, 3,001–6000 CHF, 6,001–9000 CHF, 9,001–12000 CHF, 12,001–15000 CHF, 15,001–18000 CHF, 18,001–21000 CHF, >21,000 CHF). Participants’ EDI was then calculated by dividing the total household income by an equivalence scale that accounts for the number of household members and their ages. The first adult in the household was assigned a weight of 1, each additional adult was assigned a weight of 0.5, and each child was assigned a weight of 0.3 [26]. The score was stratified by quartiles. Participants within the first quartile had the lowest EDI, while participants within the fourth quartile had the highest. EDI.

Additionally, we assessed the association between education and adherence to preventive measures. Educational level comprised three categories: primary (i.e., no school certificate, mandatory school); secondary (i.e., apprenticeship, maturity, abitur, high school diploma); and tertiary education (i.e., higher technical school, university of applied sciences, university).

We described demographic, socioeconomic, health-related variables, and adherence to preventive measures of participants at baseline. Data were summarized as n (%) and median (interquartile range, IQR). We described adherence to preventive measures over time stratified by EDI quartiles and educational levels. We assessed whether participants’ EDI and educational level were associated with adherence to preventive measures using a multivariable mixed logistic regression model with person-specific random intercepts, reporting odds ratios (ORs) and 95% confidence intervals (CIs). We included the following covariates in our model based on findings of previous studies and background expert knowledge [7, 27–30]: sex, age, canton, work situation (retired, in training/studying, working part- or full-time, non-working, other), self-reported comorbidities (cancer, diabetes, immunological diseases, hypertension, cardiovascular diseases, respiratory diseases), body mass index (BMI), smoking status (smoking daily, smoking occasionally, former smoker, never smoked), worries about the risk of being infected with SARS-CoV-2 (Likert scale from 1 to 5; 1 not worried at all, 5 extremely worried). These variables were all assessed at baseline. As adherence to preventive measures is influenced by the pandemic context, calendar date of survey response was included in the model using a cubic spline variable with 5 knots. We also included a variable for time-updated vaccination status during the follow up period (participants who received at least one vaccine dose vs. non-vaccinated), measured monthly. Some participants (n = 598) did not report any information about vaccination status and were considered as non-vaccinated.

We decided a priori to stratify the analyses by two age groups and time periods. The stratification by age (below and above 65 years) was conducted to account for income differences between retired and non-retired participants. The stratification by time period (before and after June 30, 2021) was conducted to consider the varying pandemic situation, with the period after June 30, 2021, characterized by a more widespread adoption of vaccination, fewer COVID-19 restrictions and a likely lower risk perception. Indeed, the vaccination campaign for adults aged 20 to 54 in Switzerland began in May 2021, (for older adults, the vaccination campaign had started at the end of December 2020), leading to the relaxation or removal of various restrictions, including the lifting of travel restrictions and quarantine obligations for vaccinated individuals, the easing of indoor dining prohibitions, prohibitions on private gatherings, and restrictions on indoor events [31].

We conducted a sensitivity analysis excluding participants with missing values on vaccination status and a sensitivity analysis excluding the variable “worries about the risk of being infected with SARS-CoV-2” from the main models. Results have not been adjusted for multiple comparisons. Data analysis was conducted using Stata version 17 (Stata Corp, TX, 2021).

We included 4,299 respondents in our analytical sample (52% females), with a median age of 58 years (IQR = 46–69) and a median follow up time of 429 days (IQR = 364–446). Participant characteristics of the analytical sample at baseline stratified by adherence to preventive measures are reported in Table 1. Some 107 participants (2.4% of included respondents) were not included in the analytical sample due to having at least one missing value among the covariates used in the regression models. The characteristics of participants who did not meet our inclusion criteria or were not included in the model due to missing values are reported in Supplementary Table S1 (Supplementary Material). At baseline, 64% of participants with data on adherence at baseline reported high adherence to preventive measures. 61% of participants were aged between 20 and 64 years old and 39% above 65 years. 48% of participants had tertiary education, and 4% had primary education. Around 62% of participants had a gross monthly household income below 9,000 Swiss francs.

The analytical sample characteristics stratified by EDI quartiles are reported in Table 2. There were demographic differences across EDI quartiles. The proportion of older participants (65+ years) was higher in lower quartiles, with 72% of the participants in the fourth EDI quartile being in the 20–64 years group. Individuals in the fourth EDI quartile had also higher levels of education, had slightly better health conditions and were slightly less worried about the pandemic.

Trends of adherence to preventive measures by EDI quartiles are shown in Figure 1, along with the COVID-19 epidemiological situation. Trends by educational level are shown in Supplementary Figure S3 (Supplementary Material).

Adherence to preventive measures was high across all socioeconomic strata, especially during phases with high SARS-CoV-2 incidence and more stringent containment measures (before June 30, 2021). While small differences by educational level could be seen throughout the entire follow up period, differences by EDI quartiles started to emerge only after June 30, 2021.

After adjustment, we found no evidence for EDI- or education-associated differences in adherence to preventive measures when considering the whole sample throughout the entire follow up period. However, we saw a pattern where those with the higher EDI tended to adhere less to preventive measures and this association became stronger after June 30, 2021 (Table 3): compared to participants in the first EDI quartile, second EDI quartile (Adj. OR 0.56, 95% CI 0.37–0.85), third EDI quartile (0.38, 0.23–0.64), fourth EDI quartile (0.60, 0.36–0.98).

Similar results were observed when assessing educational level (Table 4); participants with higher education were less likely to adhere to preventive measures after June 30, 2021, compared to participants with a primary educational level: secondary education (Adj. OR 0.34, 95% CI 0.14–0.82), tertiary education (0.25, 0.10–0.62).

Stratifying by age groups, we observed that the relationship between EDI and adherence to preventive measures was mainly driven by participants aged over 65 years: second EDI quartile (Adj. OR 0.38, 95% CI 0.20–0.73), third EDI quartile (0.39, 0.16–0.96), fourth EDI quartile (0.22, 0.09–0.53). To ease the interpretation of these results, we re-ran the model for participants aged over 65 years after June 30, 2021 using individual preventive measures (physical distancing, staying at home and wearing a mask) as outcomes, instead of the adherence to preventive measure score (Supplementary Material: Supplementary Tables S2–S4). We observed a similar pattern of lower adherence among persons with higher income (compared with the lowest income group) for staying at home and physical distancing, but not for wearing a mask. Results for the overall sample and the whole follow-up period for all covariates are reported in Supplementary Table S5.

Sensitivity analyses yielded similar results to the main analyses (Supplementary Tables S6, S7).

Self-reported adherence to preventive measures was high across all socioeconomic strata during phases with high SARS-CoV-2 incidence and more stringent measures. Following the broad availability of vaccines, lower incidence rates, and the lifting of measures, differences across socioeconomic strata started to emerge. Despite several potential limitations such as selection or information biases, our study highlights the need for contextual interpretation when assessing the impact of SES on adherence to preventive measures and implementing interventions to improve adherence during a pandemic.