Analyses were based on data from the eighth wave of SHARE, the Survey of Health, Aging, and Retirement in Europe (19). SHARE is a large pan-European social science panel study. The first wave of data collection started in 2004, including 12 countries. New waves of data collection were repeated every 2–3 years. Refresher samples were drawn to compensate for panel attrition. In total, about 530,000 interviews with people from 28 European countries and Israel were conducted. Topics covered in the interviews are, among others, social networks, physical and mental health, employment, retirement, financial situation, and social activities. The wave of 2020 was the first one that included data on the pandemic outbreak and was carried out in 27 countries. Based on population registers, SHARE used probability samples within the countries and includes non-institutionalized adults aged 50 years or older and, if available, their partners. Exclusion criteria were: being incarcerated, moved abroad, unable to speak the language of questionnaire, deceased, hospitalized, moved to an unknown address or not residing at sampled address (20, 21). The outbreak of the COVID-19 pandemic hit the data collection in the middle of the 8th wave, thus fieldwork was suspended and a specific COVID-19 questionnaire was developed to collect data with pandemic related topics. The fieldwork for this survey started in June 2020 and ended in September 2020.

The final SHARE sample consisted of N = 46.500 participants from 27 countries. As we were interested in the decline of mental health, we only considered those respondents who previously reported being depressed and/or being nervous or anxious (see measures). Thus, for the present study, two subsamples of persons with mental health problems were selected (people who reported being depressed, n = 9.240 or nervous/anxious, n = 10.551).

Additional data sources were used to build two macro indicators. The stringency in government response toward the COVID-19 outbreak was based on the Oxford COVID-19 Government Response Tracker (OxCGRT), a global panel database of pandemic policies that covered more than 180 countries (22). For each country at each day of the pandemic, measures to reduce the spread of the outbreak, like longer or stricter lockdown measures, reduced social contacts, and similar, were recorded and a daily index of government response for each country (stringency index) was calculated. Countries with higher stringency index have undertaken more intensive measures to reduce the spread of COVID-19. Data was filtered to include only those countries that were also present in this survey.

Data on numbers of confirmed cases of people infected with COVID-19 were taken from the COVID-19 Data Repository by the Center for Systems Science and Engineering (CSSE) at the Johns Hopkins University (23). Data on confirmed cases are collected from more than 190 countries. Based on these data, the percentage of the population infected with COVID-19 was calculated. Again, only data from countries that were also included in this survey was used.

Both data sources were continuously updated and thereby reflected the evolution of infection rates and government measures during the pandemic. To adequately represent the situation in the participants' countries at the time of the SHARE fieldwork, only those data for government response and confirmed cases that referred to the period from the beginning of the COVID-19 outbreak until the time of field work in each SHARE country was used (i.e., summer 2020).

Descriptive statistics were used to document the sample characteristics. For each of the two dependent variables indicating decline in mental health, three logistic regression models for complex samples were calculated, using quasi-binomial links to properly account for survey-weighting, disproportional sampling, and selective mortality. The country variable was used to define the strata in the survey-design, hence the regression models accounted for the fact that respondents were clustered within different countries. Robust Horvitz-Thompson standard errors are reported (25). The first model only included social factors, the second model only included the two macro indicators, and the third (full) model included both social factors and macro indicators. Models were checked for the presence of multicollinearity. All models had a variance inflation factor (VIF) below 2.8, indicating no severe collinearity issues (26). All analyses were performed using the R statistical package (27), including the packages “survey” (28), “ggplot2” (29), and “parameters” (30).

Each of the country classifications “low stringency,” “middle stringency,” and “high stringency” included nine countries (see Table 1). The average stringency index for low stringency countries varied from 34 to 39. Latvia, Slovenia and Bulgaria had rather low COVID-19 infection rates of 0.06, 0.07, and 0.08%, respectively. Countries with a middle stringency had a mean stringency index from 40 to 44. Overall, these countries showed the lowest COVID-19 infection rates from all three stringency classifications. High stringency countries had an average stringency index ranging from 45 to 56. The proportion of people infected with COVID-19 ranged from 0.11 to 0.60%, and overall, these countries had the highest infection rates.

Characteristics of the subsample of respondents who reported to be depressed or sad (n = 9,240) is shown in Table 2. The mean age of respondents was 72.3 years. 19.3% had high education. 70.0% were female. 36.5% of the participants mentioned that they had rare personal contacts. A share of 32.3% were living alone. 63.0% reported that they felt more depressed since the beginning of COVID-19.

Table 3 shows the characteristics of people who reported to be nervous and/or anxious. Their average age was 71.3 years. 37.9% were higher educated and 67.1% were of female gender. Rare personal contacts were reported by 37.8% of the respondents. 26.5% were living alone. 71.6% mentioned being more nervous and/or anxious since the pandemic outbreak.

Looking at the regression model with social factors included as predictors (model 1, see Table 4), higher age was associated with a lower likelihood of becoming more depressed (OR 0.87). Female respondents were more likely to feel more depressed since the outbreak (OR 1.50). Middle or higher educated persons, compared to people with low educational level, were less likely to feel more depressed (OR 0.61), although only middle education was statistically significant. People who lived alone in their household had a lower chance of feeling more depressed (OR 0.79). Rare personal contacts were not significantly associated with feeling more depressed. Regarding the model with macro indicators (model 2), in comparison to countries with low stringency index, respondents from countries with middle stringency index were less likely to feel more depressive after the COVID-19 outbreak (OR 0.66). The likelihood of feeling more depressed for people from countries with high stringency index, however, was higher in comparison to the low stringency countries (OR 1.37). An increased rate of people infected with COVID-19 was also associated with a higher chance of feeling more depressed (OR 3.24). In the full model (model 3), in which both social factors and macro indicators were included, most associations remained very similar compared to the first two models. The most noticeable change was related to the association with education, which has diminished in model 3.

In model 1 (see Table 5), feeling more nervous/anxious was less likely when respondents were older (OR 0.87). Female gender was associated with a higher likelihood of feeling more nervous/anxious (OR 1.54). The probability of feeling more nervous/anxious is higher for people with high education as compared to people with low education (OR 1.34). No statistically significant associations were found for middle education (OR 1.08), living alone (OR 0.99), and rare personal contacts (OR 1.07). While the stringency index was not significantly associated with feeling more nervous/anxious (model 2), the rate of people infected with COVID-19 significantly increased the chance of feeling more nervous/anxious (OR 3.95). In model 3, most associations remained similar compared to the first two models, with an exception for education. The likelihood of feeling more nervous/anxious increased with medium and higher education.

One finding was that the majority of older adults, who reported depressive symptoms or being nervous and/or anxious, showed a decline in mental health since the beginning of the COVID-19 pandemic. This holds true for respondents across all European countries. Although the proportion of people with a decline in mental health was higher in countries where the social life was affected by strict government response measures, even in countries with a milder response a majority of respondents reported an increase in depressive symptoms, nervousness, and anxiety. Regarding social factors, we found that increasing age and female gender were associated with a lower chance of a decline in mental health, while we found inconsistent or no associations for education, rare personal contacts and living alone. In terms of the macro indicators, there were positive associations with the proportion of infections in the population, but inconsistent results related to the stringency index.

Although the concrete impact of the COVID-19 pandemic on mental health is not yet fully understood, studies from former comparable outbreaks that required social restrictions or lockdown measures confirmed that the challenges and stress associated with such NPIs could trigger mental disorders, especially anxiety and depressive symptoms (11, 31). This included both, an increase in the prevalence of disorders as well as worsening mental health (18). Beyond the impact of NPIs on mental health, the fear of serious health consequences might exert a further negative influence on mental well-being. The fact that especially older people were at higher risk for adverse health outcomes due to COVID-19 might explain why—compared to surveys that also included younger populations—we found a higher proportion of decline in mental health, which did not only occur in countries with strict government response measures. This is in line with studies that showed the negative psychological impact of the pandemic on the older population due to increased morbidity and mortality risks (13, 32). In our study, we used infection rates of a population and the stringency index as two indicators on a macro level that reflect the pandemic threat and the rigidity of NPIs and found links between those indicators and a decline in mental health. Populations' infection rates were strongly positively associated with feeling more depressed and more nervous/anxious. This can be explained with the health concerns of elderly people, who were a particular risk group for adverse COVID-19 related health outcomes. The perceived threat of the pandemic increased with higher proportions of infected cases in the population. This fear was in particular present in risk groups such as older people (33). We found no evidence that the government response measures, indicated by the stringency index, were associated with feeling more nervous and/or anxious. Feeling more depressed, however, was more likely for respondents living in countries with higher stringency index. Studies that analyzed the impact of lockdown measures and comparable NPIs also reported negative mental health consequences. A longitudinal study from South Africa showed that the prevalence of depression and anxiety symptoms was higher during the time of social restrictions (34). Another study from Israel indicated that such measures were the main risk factor for depression and anxiety (35).

According to our results, increasing age was associated with a lower chance of a decline in mental health. This seemed to be somewhat counter intuitive given that in particular elderly people were one of the most vulnerable risk groups related to COVID-19 and their emotional response toward the pandemic outbreak was more apparent than in younger age groups (13, 36). Research on this topic revealed inconclusive results. While a documentary analysis indicated that mental health in older people was negatively affected by social restriction measures, another recent study found that adherence to such measures was not associated with a decline in mental health. One explanation might be that a key issue that serves as protective factor could be the social connectedness whose positive impact was rated higher than the negative impact of COVID-19 related worries (37, 38).

A clear association was found between gender and decline in mental health, with female persons being more likely to be affected. This is in line with previous research that described significantly higher scores of psychological distress for female gender during the pandemic (13). One study reported significantly larger increase in depressive symptoms for female persons (37). A meta-analysis on the prevalence of depression and anxiety among COVID-19 patients also showed higher proportions of female patients being affected. However, the total effect of gender was not statistically significant (39).

The associations of education and decline in mental health differed between the two subgroups of respondents who were more depressed and those who felt more nervous and/or anxious. In the models with social factors only, higher education showed weak evidence for a lower probability of decline in mental health. In the full model, the association between education and feeling more depressed diminished, while we found a significant positive association between higher education and feeling more nervous and/or anxious. The latter result was supported by other research results. Recent longitudinal studies on the change of mental health in the course of COVID-19 showed that a higher educational level was associated with increases in mental health problems (40). A reason might be that higher educated persons may feel greater concerns about the consequences of COVID-19, which was also reported in a study from the US (41).

Regarding social relations, living alone was associated with feeling more sad or depressed, but this association diminished after controlling for the macro indicators. We did not find clear associations for living alone and feeling more nervous and/or anxious. Rare personal contacts were not statistically significant in any model. Although studies showed a positive impact of social contacts on mental health (17), other findings suggested that there was no clear link between social contacts and nervousness and anxiety for older individuals (42). An explanation for the inconsistent associations between social relations and decline in mental health could be that older persons experience less life changes due to the pandemic, making social contacts more important for younger generations (43).

Some methodological limitations have to be considered that affect the interpretation of the findings. First, the SHARE data only provides some rather crude measures of decline in mental. An advantage of the mental health indicators we used is that they refer to acute and very recent mental health problems. On the other hand, it depends on a subjective perception of feeling depressed, nervous or anxious and is not measured by a validated assessment instrument for mental health disorders. Furthermore, the mental health indicators we used may only reflect a temporary condition and no longer lasting disorder. Nevertheless, we decided to use this indicator as it allowed us to measure a decline of mental health problems since the COVID-19 outbreak using cross-sectional data.

Another limitation is related to the measures of social relations, which might explain the inconsistent or unclear associations in our results. Only frequency of personal contacts, not telephone or digital contacts were assessed. Additionally, as rare personal contacts and living alone only refer to the quantitative dimension of social relations, they do not reflect qualitative or functional aspects. Furthermore, due to the age range in the sample, which defines older adults as persons aged 50 and older, respondents in the SHARE survey were a very heterogeneous group.

Finally, it is also important to mention that data collection was carried out in an early stage of the pandemic. Thus, our study refers to a rather short-term impact of NPIs on mental health, which might result in an underestimation of the associations between government measures and mental health. In addition, the impact of a pandemic outbreak on mental health is often influenced by many different factors. For instance, we could not include measures such as job security, loneliness, or other psycho-social factors. Nonetheless, one of the strengths of this study is the combination of a large dataset on an individual level combined with very well-prepared data on a macro level from other sources, which allowed us to gain more specific insights into the complex associations of social factors, infection rates, government response, and decline in mental health.