We conducted repeated cross-sectional online surveys in Vaud, a French-speaking canton of 823,000 inhabitants (2021) in Switzerland. We launched 4 surveys between April 2020 and December 2021: survey 1 between April 17 and May 14, 2020 (4 weeks); survey 2 between May 15 and June 22, 2020 (5.5 weeks); survey 3 between October 30 and December 1, 2020 (4.5 weeks); and survey 4 between June 18 and December 30, 2021 (28 weeks, Table 1). Some of the survey items were adapted or replaced to capture changes in federal measures. We followed the CHERRIES (Checklist for Reporting Results of Internet e-Surveys) guidelines [21]. Self-reported worry was an outcome of the study of worry levels and an exposure variable for the study of associations between worry and self-reported adherence during these 4 COVID-19 pandemic periods.

Table 1 presents the COVID-19 waves and their duration in the French-speaking part of Switzerland. The first and second forms were distributed at the end of the first wave, corresponding to the gradual emergence from confinement. The third form was distributed over a longer period, which included the second wave and the resumption of restrictive measures. The fourth form was distributed once vaccination was available for the entire population.

Using a weblink, we distributed the surveys on the social media platforms of multiple community organizations to collect a convenience sample of the population. These organizations were a regional consumer organization, regional disease leagues for cancer and diabetes, the association of senior citizens as well as the cantonal websites for the COVID-19 testing and vaccination decisions. These cantonal sites were used by large portions of the population. The organizations advertised the study through links on their websites and some social media accounts. The links were accompanied by a short explanation of the study and its purpose. No incentives to participate were given. The online interface for the survey was created in REDCap (Research Electronic Data Capture, Vanderbilt University).

The development and testing of the survey are described in our previous publication with results from survey 1 [20]. Each survey was submitted to 5 nonmedical persons to test the understandability of questions. There was no review step for this short questionnaire. The first, second, and fourth surveys had 20 items, and the third survey had 25 items. A total of 12 items remained unchanged throughout all surveys. We included demographic data, such as age, sex, number of persons per household, canton of residence, level of education, literacy, and whether the respondent had been tested for the COVID-19 pandemic. For the literacy question, we used a validated item from Chew et al [22]. Employment status was included in surveys 2-4. Respondents rated (1) worry about the pandemic situation, (2) self-reported adherence to government restrictions, and (3) perceived adherence of others to government restrictions, on visual analog scales from 0 to 100 (0=not at all; 100=in all situations). Items were not randomized. Participants could go back to earlier questions at any time. Adaptive questioning was used for several items. The first 2 surveys took up 6 screens, and the 2 latter, 7. REDCap automatically generated a completeness variable if participants went all the way to the end of the survey. Anyone who opened the survey generated a response. We did not determine unique site visitors, establish view or participation rates, or IP address checks, as surveys were entirely anonymous and the risk of repeating them was low. The time used to fill them in was not registered. No cookies were used, and there were no other techniques to analyze the log file of our database. Statistical corrections were not used. The 4 surveys can be found in Multimedia Appendices 1-4.

The item “Level of education” was dichotomized into “university or college education” or other. Health literacy was dichotomized according to ease of answering a medical form on one’s own: low literacy (“never,” “rarely,” or “sometimes” at ease) and “high literacy” (“often” or “always”). This was based on a validated item [22] and our previous article [20].

We limited our analysis to complete questionnaires for 2 reasons. First, the survey was distributed using an online link on government websites and a large number of persons clicked the link but only completed 1 or 2 questions. Second, no single question had many missing responses and we preferred to maintain consistency across surveys. We calculated means with SD and frequencies with IQR as appropriate. Independence between surveys was tested with the t test for continuous variables (eg, age), and with the chi-square test for gender, education, and health literacy. We performed linear regressions to analyze associations between the 4 periods and level of worry (model A), self-reported adherence (model B), and perceived adherence of others to restrictions (model C). First, we performed univariate linear regression, followed by multivariable linear regression controlling for age (grouped as 18-39 years, 40-64 years, and 65 years or older), sex (dichotomized male-female), level of education (dichotomized university education or other), and health literacy (dichotomized high or low health literacy). We subsequently used the margins command in Stata (StataCorp) to report absolute differences in the predicted levels of worry or adherence with each model. The level of significance was set to P<.05. Statistical analyses were performed in Microsoft Excel and Stata (version 16.1).

According to the Cantonal Commission on Ethics in Research Involving Human Beings of the Canton of Vaud, Switzerland the study was exempted from ethical review because it did not qualify as human subjects research and all data collection was anonymous (2024-010901).

Citizens completed 7106 surveys between April 17, 2020, and December 20, 2021. After the exclusion of minors, persons living outside the Canton, and incomplete questionnaires, 4729 (66%) surveys remained. The number of questionnaires per period ranged between 563 and 2675 (Figure 1).

The participants were more often women (3025/4729, 63.96%) and between 40 and 64 years old (2442/4729, 51.64%). Furthermore, 2526/4729 (53.42%) of participants had attended university or college, and 4510/4729 (95.37%) reported high health literacy. The participants were younger in survey 3 (mean age 43.2, SD 14.6 years; P<.001) than in the other surveys (survey 1: mean age 47.9, SD 14.6 years; survey 2: mean age 47.3, SD 15.6 years; and survey 4: mean age 48.7, SD 15.3 years; Table 2).

We conducted online surveys during different phases of the COVID-19 pandemic to describe the evolution of worry levels and to assess how these were associated with adherence to government restrictions. In Switzerland, the self-reported worry was highest during the first and second pandemic waves, corresponding to survey 1 and survey 3, at times of many COVID-19–related hospitalizations and deaths, and when a vaccine was not yet available. Women and younger people reported higher levels of worry than men and older people. Education did not influence worry levels, while lower health literacy was associated with higher worry. Higher worry levels were associated with higher self-reported and perceived adherence of others to federal restrictions.

We found elevated worry levels during the more dramatic phases of the pandemic. In a systematic review covering 204 countries in 2020, higher anxiety was associated with higher COVID-19 incidence [23], a finding that was confirmed by Salanti et al [13]. In March 2020, Fitzpatrick et al [24] found that in a national sample in the United States, worry was higher in the regions with high COVID-19 incidence. In Ontario, Canada, COVID-19–related worry in young persons also increased during the early phases of the pandemic and then again in the autumn of 2020, when the incidence was higher [25]. So, anxiety and worry varied during the pandemic and increased repeatedly with the rising incidence of COVID-19. This is in line with findings on decreased mental health on a larger scale during the pandemic [11,26].

In periods of increased worry, we found higher self-reported adherence to government restrictions. Similarly, a study in Saudi Arabia described an association between higher anxiety levels and preventive practices among health care workers [27]. Another study identified fear as a predictor of behavioral change [28]. The association between worry and adherence in our study could indicate that worry was not overwhelming and that citizens felt in control of risks by respecting restrictions. We cannot exclude that this might have been different with higher anxiety levels.

Considering population subgroups, young adults were often the most anxious despite being less at risk of hospitalization or death [17,18,23,29]. Young people were worried about social isolation and develop depressive symptoms during school closings [30]. In Switzerland, students were concerned about whether they would be able to finish the 2020 university year [31], and lockdowns as well as their socioeconomic consequences were stressful for students [15]. Apart from concerns about the future, young people were not only worried about their own health but also about that of relatives. For example, in a study in Zurich, Switzerland, students were more concerned about the health of their parents and grandparents than their own [32]. Also, our finding of higher worry levels in women echoes several publications [17-19,33]. General factors potentially contributing to worry were the increasing risk of unemployment or loss of income, as well as loneliness and feeling less productive. One-third of respondents had lost part of their income by surveys 2 and 3 (31%, and 34%, respectively), with a slightly better situation in survey 4 (21%). The reported feelings of isolation, loneliness, and being less productive could further contribute to worry in some respondents. For example, studying for exams through online classes only, without any campus life, can be a source of worry compared with when stress from studying and exams is compensated by in-person interactions with teachers and colleagues. Four years after the pandemic, Sayed et al. [34] insist on the importance of addressing mental health of children and young adults during global crises and of recognizing long-term impacts. They further emphasize the need for research and public health prioritization of these important topics.

Overall, our findings are in line with publications that highlight the importance of addressing the many individual and collective aspects that influenced mental health during the COVID-19 pandemic, such as isolation, loneliness, and fear [35]. Many individuals demonstrated remarkable resilience, allowing society to avoid a general increase in loneliness [36]. However, population estimates may mask individual heterogeneity; loneliness is indeed a major public health concern and must be considered as a negative determinant of health [35]. Even though the pandemic is over, we must not forget its long-term effects on mental health and public health authorities should consider the differing impact of governmental decisions on the general population versus on individuals with pre-existing mental health conditions [37]. Our results of the worries of citizens and adherence to pandemic measures can be useful in preparing for future pandemics, for example, in considering criteria for and potential impact of restrictions on different subgroups of the population.

The strengths of our study were repeated surveys with similar questions and our relatively large sample size, allowing us to examine subgroups of the population.

Concerning limitations, data were collected through the online distribution of surveys in a simple, cost-effective, and feasible way during the rapidly evolving pandemic. Participation was more attractive to women and younger people with higher literacy and education. Also, participation was variable during data collection periods. The participant sample was more representative in the last and longest period (survey 4). This selection and variation need to be considered in the interpretation of our data which are prone to desirability, information, and selection bias. Different distribution channels and methods are probably needed for disadvantaged populations, as we showed in a recent study using our survey in a population of refugees and migrants [38]. Finally, another inherent limitation of our anonymous data collection is that we could not follow a cohort of persons throughout the pandemic.

Worry reached moderate levels and varied with COVID-19 incidence during the pandemic. Higher worry levels were associated with increased self-reported and perceived adherence of others to government restrictions. Younger people and women reported higher worry levels. Authorities should take population worry levels into account in planning and designing pandemic communication. Adapting communication to population subgroups should be considered for future health crises.