Abstract
Objective:
Although the COVID-19 pandemic has affected mental health, understanding who has been affected most and why is incomplete. We sought to understand changes in mental health in the context of transmission numbers and pandemic (social) restrictions and whether changes in mental health varied among population groups.
Methods:
We analyzed data from 92 062 people (aged ≥16 years and able to read Dutch) who participated in the Corona Behavioral Unit cohort study at the National Institute for Public Health and the Environment, the Netherlands, from April 17, 2020, through January 25, 2022. Participants self-reported mental well-being through multiple rounds of surveys. We used a multivariable linear mixed-effects model to analyze loneliness, general mental health, and life satisfaction.
Results:
As strictness of pandemic prevention measures and social restrictions increased, people’s feelings of loneliness increased and mental health and life satisfaction decreased. As restrictions were relaxed, loneliness decreased and general mental health improved. Younger people (aged 16-24 y) versus older people (aged ≥40 y), people with low (vs high) education levels, and people living alone (vs living together) were more likely to have negative well-being outcomes. We observed that trajectories over time differed considerably only by age, with participants aged 16-24 years affected substantially more than participants aged ≥40 years by pandemic social restrictions. These patterns were consistent across multiple waves of SARS-CoV-2 infection.
Conclusions:
Our findings suggest that the social restrictions imposed by the Dutch government during the study period were associated with reduced mental well-being, especially among younger people. However, people appeared resilient as they recovered during periods when restrictions were relaxed. Monitoring and supporting well-being, in particular to reduce loneliness, may help younger people during periods of intense social restrictions.
Global measures have been introduced to reduce the spread of SARS-CoV-2, to prevent COVID-19 and its associated morbidity and mortality, and to reduce the strain on health care systems. 1 COVID-19 prevention measures have varied from hand washing to school closures and banning of all public gatherings. 2 In the Netherlands, the first SARS-CoV-2 infection was detected in February 2020. Between February 2020 and December 2022, in a population of >17 million people, >8.5 million positive cases and >23 000 deaths were registered, and 3 lockdowns were imposed. 3
Social restrictions are effective in reducing the spread of COVID-19, but they may also result in increasing loneliness, financial struggles, stress from balancing work and home schooling, and mental health problems.4,5 Among studies on mental well-being during the COVID-19 pandemic, to our knowledge, only a few longitudinal studies have been conducted; these studies used a relatively short period into the pandemic, 6 accounted for only 1 lockdown, 7 or did not compare sociodemographic groups. 8 In the Netherlands, a few longitudinal studies on mental well-being have been published; these studies included only a specific region, 9 used a relatively short period of the pandemic, 10 or investigated just 1 demographic subgroup.11,12 With few studies on mental well-being during the COVID-19 pandemic, limited knowledge is available on who has been affected most and why.
Among 18 meta-analyses, high proportions of psychophysiologic stress were reported in the general population. 13 However, although some studies showed that mental health deteriorated,14,15 others found only minor negative changes in mental health 16 or no changes.17-19 One possible explanation for these varying results is that population subgroups respond differently to the pandemic. For example, younger people versus older people or people who live alone versus those who live together could be more strongly affected by extended periods with social restrictions, whereas older people might be more affected by preexisting health conditions than their younger counterparts.
A systematic review and meta-analysis of 65 studies reported that a significant but small increase in mental health symptoms occurred in the early period after the start of the COVID-19 pandemic (March through April 2020) as compared with before the pandemic, with larger and persistent increases in depressive symptoms than in anxiety symptoms. Mental health symptoms were especially elevated in people with preexisting physical health conditions but were not related to age, sex or gender, or preexisting mental health conditions, 20 showing that COVID-19 had varying effects among different population subgroups. The increase in mental health symptoms observed in the first months of the pandemic decreased toward the summer, which the authors proposed could indicate an initial response to an unforeseen and traumatic event, followed by adjustment and consequent improvement of mental health. 20 An alternative explanation could be that, as the social and mobility restrictions on people’s daily lives were relaxed, well-being improved. Testing this hypothesis, however, requires data during multiple surges in SARS-CoV-2 infections or pandemic waves. A longitudinal study on the changes in the well-being of people resulting from COVID-19 policies during an extended follow-up period, during which long periods of strict measures alternated with extended periods of relaxation, could help explain the different effects among population subgroups.
For this study, we used 18 rounds of cohort data during the first 22 months of the COVID-19 pandemic (April 17, 2020, through January 25, 2022) in the Netherlands. Our objective was to examine mental well-being in the context of multiple pandemic waves in the Netherlands and whether trajectories in well-being differed among population subgroups.
Methods
Study Design and Setting
In April 2020, which corresponded with the first months of the COVID-19 pandemic in the Netherlands, the Corona Behavioral Unit at the National Institute for Public Health and the Environment, the Netherlands, launched a dynamic cohort study to monitor the Dutch population, including the well-being of people over time. 21 We invited the participants of existing panels of the 25 regional Public Health Services in the Netherlands to participate in a survey. To compensate for dropout, for several rounds we recruited additional respondents. To reduce the length of the survey, we randomized one-third of the respondents to a subcohort on well-being (the other two-thirds were randomized to other topics). We included participants who were aged ≥16 years and able to read Dutch.
We asked participants who completed their first survey whether they would be willing to participate in a cohort study and complete the survey on a regular basis. We sent emails to participants who consented during follow-up measures. We informed participants that they could withdraw from the study at any time. We collected informed consent from participants. We outsourced data collection to a research agency (Research 2Evolve).
The Centre for Clinical Expertise at the National Institute for Public Health and the Environment, the Netherlands, determined that the cohort study did not meet the requirements as laid down in the Law for Research Involving Human Subjects and exempted the study from formal ethical review (study G&M-561).
Outcome Variables
We assessed loneliness by using the short scales for emotional and social loneliness from the validated 6-item De Jong Gierveld Loneliness Scale. 22 Participants rated items based on the previous week (eg, “I miss the pleasure of the company of others”), with 3 response options (yes, more or less, no). Following the recommended conversion, we calculated a total score from 0 (lowest loneliness score) to 6 (highest).
We assessed general mental health by using the Mental Health Inventory–5, 23 a brief validated 5-item scale for measuring mental health in adults. Participants rated items such as “How much of the time, during the past week, have you felt downhearted and blue?” on a 6-point Likert-type scale, ranging from 1 (all the time) to 6 (never). Following the recommended conversion, we calculated total scores from 0 (lowest mental health score) to 100 (highest).
We assessed life satisfaction with a single 10-point item 24 : “For some participants the corona measures may have little impact on their lives. For others the impact may be greater. What score would you give your life at the moment?” Scores ranged from 1 (very poor) to 10 (excellent).
Independent Variables
We examined the association between each of the 3 outcome measures and the following self-reported sociodemographic characteristics among participants: sex (male, female), age at the time of enrollment (≥70, 55-69, 40-54, 25-39, and 16-24 y), education level (lower: primary, middle: secondary or vocational, and higher: higher professional or university), living alone or together, and having a physical health condition (eg, a heart condition or asthma) that increases the likelihood of severe COVID-19 when infected (yes, no).
We also included 2 other independent variables: round of data collection (the consecutive surveys completed by participants at different times during the study period) and COVID-19 stringency index. We calculated the COVID-19 stringency index as a mean score of the stringency level of the Dutch government’s response to control SARS-CoV-2 infections on any given day, ranging from 0 (no measures) to 100 (strictest measures). 1
Statistical Analysis
We used a multivariable linear mixed-effects model to estimate trajectories in the outcomes over time (loneliness, general mental health, and life satisfaction), with sociodemographic variables and round of data collection as independent adjusting variables. To determine the effect of the strictness of the imposed COVID-19 prevention measures, we used separate models in which we included the stringency index but excluded the round of data collection. We used the function “lmer” of the lme4 package (version 1.32) in R version 4.2.3 (R Foundation for Statistical Computing) for model fitting. We accounted for missing data by using complete case analysis.
We used restricted cubic splines to model the effect of time (round of data collection) with 4 degrees of freedom. We determined the optimal number of degrees of freedom by comparing the Bayesian information criteria for models using splines with different numbers of degrees of freedom. Because our observations were clustered within rounds of data collection and participants, we performed cross-classified multilevel analyses with random intercepts for participant and round of data collection, to account for dependencies in observations, and with a random slope for round, to allow for the effect of time to vary by participant.
To test for mean differences among categories of the independent variables, we estimated marginal mean differences among independent variable categories for each outcome. We calculated estimated marginal means by using the emmeans package version 1.8.5 in R version 4.2.3. Because of multiple testing and the large sample size, we defined associations of interest as those with an estimated β coefficient that suggested an average difference in the outcome variable of ≥5 percentage points (ie, ±0.30 on the 6-point loneliness scale, ±5.0 on the 100-point general mental health scale, and ±0.5 on the 10-point life satisfaction scale) and P < .01.
We also investigated whether the trajectories over time differed among various subgroups (eg, age categories). To limit the number of interaction terms per multivariable model, for every outcome, we fitted separate models with the interaction between round of data collection (as cubic splines) and the independent variable. We considered variables as relevant if the marginal means of the subgroups differed by ≥5 percentage points and the corresponding significance for this variable in the multivariable model was P < .01.
We visualized overall longitudinal patterns and patterns per subgroup by plotting the estimated marginal means of values over time. To describe the stages of the COVID-19 pandemic, we plotted the COVID-19 stringency index. To describe the threats as posed by SARS-CoV-2 infections, we plotted the number of confirmed infections and the number of COVID-19–related hospital admissions. 25 Out-of-hospital testing for SARS-CoV-2 infection became available in July 2020.
Results
From April 17, 2020, through January 25, 2022, an overall 92 062 participants entered the cohort and completed ≥1 survey on their well-being (Table 1). The majority of participants were aged ≥40 years (73.3%), 57.9% were highly educated, 16.9% were living alone, 24.4% had a physical health condition, and 34.9% responded to ≥2 surveys.
Characteristics of participants in a subcohort on well-being in the Corona Behavioral Unit cohort, National Institute for Public Health and the Environment, the Netherlands, April 17, 2020, through January 25, 2022 a
All values are number (percentage) unless otherwise indicated. Discrepancies in totals are due to missing responses.
Lower education level represents elementary school or less. Middle education level represents secondary education (academic, vocational, and technical education). Higher education level represents postsecondary education (college, university, professional, vocational, and technical).
Includes participants who provided updated responses in ≥2 rounds of well-being surveys.
Multiple waves of SARS-CoV-2 outbreaks occurred during the study period, resulting in multiple peaks in the number of hospital admissions and related changes to government COVID-19 prevention measures in the Netherlands (Figure 1).
Trends in COVID-19 stringency, number of cases of COVID-19 infection, hospital admissions due to COVID-19 in the Netherlands, and mental well-being among participants of the Corona Behavioral Unit cohort, National Institute for Public Health and the Environment, the Netherlands, April 17, 2020, through January 25, 2022. (A) COVID-19 stringency index, which measured the stringency level of the Dutch government’s response to control COVID-19 infections on any given day, ranging from 0 (no measures) to 100 (strictest measures). Data source: Oxford COVID-19 Government Response Tracker. 1 Numbers represent the data collection rounds of well-being surveys from Dutch people who participated in the Corona Behavioral Unit cohort, National Institute for Public Health and the Environment, the Netherlands, April 17, 2020, through January 25, 2022, and circled numbers represent key moments of the COVID-19 pandemic in the Netherlands (February 2020 through January 2022). (B) Number of confirmed cases of COVID-19. Data source: National Institute for Public Health and the Environment. 25 (C) Number of hospital admissions due to COVID-19. Data source: National Institute for Public Health and the Environment. 25 Trends in well-being, as self-reported by Corona Behavioral Unit study participants, for (D) loneliness, (E) general mental health, and (F) life satisfaction, controlling for sociodemographic factors. Shaded areas indicate 95% CIs. Data source: Corona Behavioral Unit, National Institute for Public Health and the Environment, the Netherlands.
The Netherlands government implemented various COVID-19 prevention measures during the first 22 months of the COVID-19 pandemic (Table 2). In April 2020, the Netherlands government ordered the first lockdown; in summer 2020, prevention measures became substantially less stringent, including the possibility of increased outdoor activity. With a second surge in SARS-CoV-2 infections in autumn 2020, the Netherlands government ordered an extended (partial) lockdown from October 2020 through February 2021. The government gradually relaxed measures again starting in March 2021 and through summer 2021 and increased allowances for outdoor activity. In autumn 2021, the government increased prevention measures, including an evening lockdown and eventually a third lockdown in winter 2021-2022.
Main COVID-19 prevention measures imposed by the Dutch government before and during rounds of well-being surveys of Dutch people who participated in the Corona Behavioral Unit cohort, National Institute for Public Health and the Environment, the Netherlands, April 17, 2020, through January 25, 2022 a
COVID-19 stringency index was calculated as a mean score of the stringency level of the Dutch government’s response to control COVID-19 infections on any given day, ranging from 0 (no measures) to 100 (strictest measures). Data source: Oxford COVID-19 Government Response Tracker. 1
Circled numbers are depicted in Figure 1A.
3G Corona pass: people received a QR code and were allowed access to a location if they had been vaccinated, recovered from COVID-19, or tested negative in the past 24 hours.
Average loneliness levels among participants were high during the first lockdown (April 2020) and then decreased when COVID-19 prevention measures were relaxed during summer 2020 (Figure 1D). Loneliness levels peaked again in winter 2020, when the government intensified restrictions up to a second lockdown on December 15. As the vaccination campaign progressed (starting at the beginning of 2021), the pressure of COVID-19 on the Netherlands’ health care system decreased, and the prevention measures were relaxed step-by-step. During this period, self-reported loneliness among participants also decreased, although the decrease in loneliness levels occurred more slowly as compared with the decrease in summer 2020. With the arrival of the Omicron variant as well as the Delta variant (late 2021), restrictions were intensified again but not at levels as stringent as during prior lockdowns, with participants reporting only a modest increase in loneliness during this period.
General mental health showed an inverse relationship to loneliness: it improved when loneliness decreased and vice versa (Figure 1E). We noted 2 differences. First, during the third COVID-19 lockdown, although the level of loneliness did not increase substantially, the average level of self-reported mental health declined. Second, although participants reported feeling considerably less lonely during the third lockdown than during the second lockdown, levels of general mental health reached the same low points during the third and second lockdowns. We observed a similar inverse relationship with loneliness for scores on life satisfaction (Figure 1F).
The models showed that participants, on average, recovered at times when measures were relaxed and after an extended period of stringent COVID-19 prevention measures. However, the rate of recovery after the first lockdown was much more rapid than after the second lockdown. Moreover, self-reported mental health levels were as low during the third lockdown as during the second lockdown, despite the less stringent COVID-19 prevention measures and the lower self-reported levels of loneliness. The trough and peak in general mental health were considerably larger for the first lockdown than for the second and third lockdowns.
We observed differences between the demographic characteristics and the 3 mental health dependent variables (Figure 2). Participants who were younger (aged 16-24 y) versus older (aged ≥40 y), who were living alone (vs living together), and who had a lower (vs higher) education level were, on average, more likely to be lonely. For general mental health and life satisfaction, only age showed a relevant between-subject association, with lower general mental health and life satisfaction scores among participants aged 16-54 and 16-39 years, respectively, as compared with participants aged ≥55 and ≥40 years. Additional investigations into the association between stringency index and the 3 outcomes revealed that, overall, increased stringency in COVID-19 prevention measures was significantly associated with increased loneliness (β = 1.6; P < .001), decreased general mental health (β = –6.1; P < .001), and decreased satisfaction with life (β = –1.1; P < .001).
Results of multivariable models to analyze the main outcomes of loneliness (A), general mental health (B), and life satisfaction (C) among participants of the Corona Behavioral Unit cohort, National Institute for Public Health and the Environment, the Netherlands, April 17, 2020, through January 25, 2022, in the context of Dutch COVID-19 policies. Because of large numbers, 95% CIs are relatively small. Negative and positive numbers indicate negative and positive associations with the outcome, respectively. Models included 90 881 participants, accounting for 306 763 records with complete data on all included adjusting variables. Data source: Corona Behavioral Unit, National Institute for Public Health and the Environment. *** Significant at P < .001 based on cross-classified multilevel analyses with random intercepts for participant and round of data collection, and with a random slope for round.
Among outcome trajectories for the various sociodemographic subgroups over time, younger participants (particularly those aged 16-24 y) reported substantially lower levels of mental health during the second lockdown than older participants. This finding was evident from data collection round 9 and later (Figure 3A). This period had a curfew and extended closures of institutions of higher education. An inverse pattern, but to a lesser extent, was also visible for loneliness and life satisfaction, with the highest and lowest scores, respectively, observed during the second lockdown. Differential patterns were not as pronounced for the other predictors.
Subgroup patterns of loneliness (A), general mental health (B), and life satisfaction (C) among participants of the Corona Behavioral Unit cohort, National Institute for Public Health and the Environment, the Netherlands, April 17, 2020, through January 25, 2022, in the context of Dutch COVID-19 policies. Subgroups were based on sex, age group, education level, living situation, and physical health condition. Scores represent total scores, ranging from 0 (lowest) to 6 (highest) for loneliness, from 0 (lowest) to 100 (highest) for mental health, and from 1 (very poor) to 10 (excellent) for life satisfaction. Numbers on the x-axis refer to the dates of data collection rounds of the Corona Behavioral Unit cohort. See Table 2 for corresponding dates of data collection. Data source: Corona Behavioral Unit, National Institute for Public Health and the Environment.
Discussion
In the current study, we analyzed self-reported survey data from the first 22 months of the COVID-19 pandemic in the Netherlands (April 17, 2020, through January 25, 2022), among participants of a cohort study conducted by the Corona Behavioral Unit at the National Institute for Public Health and the Environment, the Netherlands. We found that COVID-19 restrictions implemented during the pandemic, which primarily targeted reducing social contacts and mobility, coincided with increased loneliness and decreased general mental health and life satisfaction. Self-reported general mental health declined during extended periods of stringent COVID-19 restrictions, particularly among younger participants. When COVID-19 prevention measures were relaxed, self-reported levels of loneliness declined and self-reported levels of general mental health and life satisfaction improved to previous levels. Of note, the trajectory was different for younger participants: their well-being throughout 22 months of the COVID-19 pandemic remained substantially lower than during summer 2020, the period after the first lockdown. The fluctuations in well-being during the first SARS-CoV-2 surge were considerably larger than those during consecutive surges, suggesting that although participants on average appeared to be affected less by the pandemic further into it, recovery from periods of lockdown was also slower and well-being never reached the peak that occurred during summer 2020 after the first lockdown.
Periods of stringent COVID-19 prevention measures have been associated with deteriorating mental health.8,26 In contrast to our findings, a study in the United Kingdom did not report recovery from deteriorated mental health levels among participants after the first 2 lockdowns were lifted, indicating a sustained worsening of mental health during the pandemic as compared with prepandemic levels. 6 Robinson et al 20 proposed that increased mental health symptoms in the first months of the pandemic, and the consequent decrease toward summer 2020, could indicate a response to an unforeseen and traumatic event. However, our findings suggest that this pattern could be a consequence of strict COVID-19 restrictions during the first wave of SARS-CoV-2 infections, as patterns were similar during consecutive waves of SARS-CoV-2 infections (albeit less extreme, except for younger participants). In contrast to Robinson et al, we did not observe poorer mental health among participants with a physical health condition as compared with those without such a condition, which could have placed them at increased risk for severe COVID-19. An explanation for this difference could be that our multivariable models included potential confounders of this association. Instead, in our models, age, education level, and living situation were risk factors for mental health issues.
Pandemic restrictions did not affect all participants equally. Although older adults and people with a physical health condition may experience COVID-19–related stress as a result of their personal health risk, 27 we found that periods of increased COVID-19 restrictions did not have a stronger effect on the well-being of older adults and people with physical health conditions than other subgroups in our analysis. Rather, longitudinal patterns of loneliness, general mental well-being, and life satisfaction showed that, during extended periods of increased social restrictions due to COVID-19, well-being declined, particularly among younger participants, in line with previous international 6 and Dutch 9 studies. However, several Dutch studies have concluded that mental health symptoms during the first months of the pandemic were stable in the general population 11 and among adolescents. 10 Differences in outcomes between those studies and our study may have resulted from differences in the number of measurements and length of the study period.
The difference in mental well-being between younger (aged 16-24 y) and older participants still existed when restrictions were temporarily lifted. The decreased self-reported general mental health among younger people may have resulted from limitations in their skills and resources (eg, small living spaces) to cope with social restrictions and COVID-19 pandemic stressors. In addition, the working and social lives of younger participants may have been more drastically altered by closures of schools, universities, pubs, and night clubs in a period of their lives when social contacts are particularly important. 28
Future analyses can help answer questions about how well-being outcomes will develop over time, as the pandemic evolves into an endemic stage in which few to no social restrictions will be implemented. Future research is needed on other mediators between COVID-19 and mental health outcomes, including job security, work–home balance, perceived threat of the virus, or coping strategies.
The societal effects of the COVID-19 pandemic and COVID-19 prevention measures concern several domains, including the social and mental well-being of people, as addressed in this study, and whether certain sociodemographic subgroups are more affected by the effects of socially restrictive pandemic measures than other subgroups. Policy makers should consider the differential effects of these measures when making policies for future pandemics and consider any interventions that could help people, particularly younger people, cope with the negative social effects of COVID-19 restrictions. Young adults are still in a vulnerable neurodevelopmental period, when mental health problems are most likely to begin to arise. 29 Therefore, programs that improve social connections, with an aim to reduce loneliness, may be relevant for younger people. 30 Increasing social support by encouraging a sense of belonging through group activities may reduce experiences of loneliness in times of social isolation during a pandemic. 31 Some promising public health efforts include the use of a positive psychology application to improve the quality of existing social relations 32 and specific mind–body practice techniques (eg, meditation) to address anxiety and depressive symptoms. 33 Another option is to create safe environments (eg, well-ventilated spaces) where people can meet and continue with in-person education or sports activities.
Strengths and Limitations
This study had several strengths, including the long follow-up time during multiple stages of the pandemic in a large group of respondents, the use of multivariable models to examine subgroup differences over time, and the availability of contextual data for interpretation (SARS-CoV-2 infections, COVID-19 hospitalizations, and strictness of COVID-19 policies). This study also had several limitations. First, people with low literacy skills and people who were immigrants were underrepresented in this cohort. Because people with low literacy and immigrant populations may be disproportionately affected by the pandemic, our results may have underestimated the effects of the COVID-19 pandemic on mental well-being. 34 Second, because this cohort study was initiated during the pandemic, we had no prepandemic measurements. Third, we did not control for SARS-CoV-2 infection; being infected with SARS-CoV-2 may have also negatively affected the mental health of people through social isolation or because of biologic mechanisms. 35
Conclusion
In a general population cohort in the Netherlands, COVID-19 restrictions, which affected social activities and mobility, resulted in decreased levels of mental well-being, especially among younger people. However, when social and mobility restrictions were lifted, levels of mental well-being increased, showing that people recovered over time. For younger people, we suggest continued monitoring and support.
Footnotes
Acknowledgements
The authors thank colleagues at GGD GHOR (Association for Public Health and Safety in the Netherlands) and the 25 regional public health services (GGD) for their assistance with data collection and colleagues at the Corona Behavioral Unit, especially Maarten Schipper, PhD, for statistical assistance.
Disclaimer
The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the article. Because of the General Data Protection Regulation, a European Union privacy regulation, personal data cannot be shared publicly. For academic collaborations and publishing in scientific journals, we have initiated a Behavioral Science Consortium (“BePrepared”) with researchers working at universities as well as the National Institute for Public Health and the Environment, which allows collaborative work on further analyses of the data and publication of future articles.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Corona Behavioral Unit initiative has been made possible by funding from ZonMW (Netherlands Organization for Health Research and Development), the Dutch Ministry of Health, Welfare and Sport, and the National Institute for Public Health and the Environment.
