Longitudinal analyses of depression and anxiety highlight greater prevalence during COVID-19 lockdowns in the Dutch general population and a continuing increase in suicidal ideation in young adults

Introduction

The ongoing pandemic of the coronavirus disease 2019 (COVID-19) has a major impact on societies and led to increases in major depressive disorder (MDD), generalized anxiety disorder (GAD), and suicidality across the world[1–3]. These conditions are severe and disabling and represent major contributors to the global burden of disease and mortality[4]. As most studies on depression, anxiety, and suicidality so far were conducted over the first months of the pandemic and were cross-sectional in their design, they do not provide temporal insights on mental health consequences over time nor on the impact of later COVID-19 lockdowns or other governmental restrictions. Continued longitudinal monitoring of changes in mental health in the population is therefore important as it can help identify groups who are at risk as well as inform government actions to reduce spread of the virus while also prioritizing the mental health of residents.

Here, we investigate the development of (symptoms of) MDD, GAD, and suicidal ideation during the COVID-19 pandemic in Lifelines, a large population-based cohort in the North of the Netherlands. Between April 2020 and July 2021, MDD, GAD and suicidal ideation were repeatedly assessed in >76,000 participants, which represents the largest longitudinal cohort on pandemic-related impact on mental health in the population[5]. Using detailed self-reported longitudinal data, we estimated the prevalence of these outcomes across the first fifteen months of the pandemic and three nation-wide lockdown periods. We furthermore investigated differences in risk across age, sex, and lifetime history of MDD/GAD.

Methods

Full details on cohort information, digital questionnaires, and our analytical strategy can be found in the supplemental materials.

Results

The COVID-19 pandemic and subsequent lockdowns in the Netherlands

During different phases of the pandemic, the Dutch government enacted a total of three nationwide lockdowns, each defined by specific measures and characteristics (Figure 1). Using data collected between March 2020 and June 2021, we next estimated the longitudinal trajectories of (symptoms of) MDD/GAD and suicidal ideation across the three lockdown periods.

• Download figure
• Open in new tab

Figure 1. Timeline of the COVID-19 pandemic and subsequent government measures in the North of the Netherlands.

Shown are the total number of daily reported infections (black bar graphs) and average change in mobility (red line) in the three Northern provinces in the Netherland (Groningen, Friesland, Drenthe) across three nation-wide lockdowns over time. The number of daily infections were downloaded from the website of the National Institute for Public Health and the Environment (RIVM) specifically for these three provinces. The change in mobility was downloaded from Apple Mobility Trends Reports also for these three provinces. The change in mobility is expressed as the percentage change compared to a baseline on January 13th, 2020. The three enacted nation-wide lockdowns are visualized by the grey vertical rectangles. The first known case of a SARS-CoV-2 infection was reported in the Netherlands on February 27 and the first reported death on March 06, 2020. The first lockdown, from March 12 - May 31, 2020, to reduce spreading of the virus was defined by targeted measures to restrict social interaction such as closing of public spaces, bars and restaurants, and work from home recommendations. The targeted lockdown reduced mobility in society by 80% but was not a hard lockdown[16]. After the subsequent summer, a new partial lockdown was announced that started on October 14 and lasted until December 14, 2020. During this period of partial lockdown, bars and restaurants were closed and limitations on social gatherings and house visits were recommended. The partial lockdown transitioned into a hard lockdown that lasted for four months which included mandatory closure of all non-essential stores and public space, closure of schools, and an evening curfew, among other measures. The national vaccination program started on January 6, 2021. Easing of restrictions of the hard lockdown were introduced on April 21 and set in motion starting April 28, 2021.

Longitudinal trajectories of depression, anxiety, and suicidal ideation

Depressive symptoms

We found a significant non-linear trajectory for depressive symptoms over time (Figure 2, Table S3). Depressive symptoms were high at the start of the pandemic and declined as the first targeted lockdown progressed. Participants reported a lower number of symptoms during mid-summer, which then increased again in August 2020. Symptoms plateaued during the second partial lockdown in November and increased again during the hard lockdown after December 2020. During this third lockdown, reported symptoms reached their peak mid-March 2021 and declined again as the lockdown ended. Comparing the end of April 2021, when restrictions of the hard third lockdown started to be lifted, with the end of April 2020, participants reported more depressive symptoms a year later (0.54 compared to 0.46). The non-linear symptom trajectories were similar in sensitivity analyses in the full cohort as well as after taking into account individual- and family-specific effects (Figure S4 and Table S7, S11).

• Download figure
• Open in new tab

Figure 2. The longitudinal trajectory of MDD and GAD outcomes and suicidal ideation during the COVID-19 pandemic.

Shown are the scores of reported depressive and anxiety symptoms and point prevalence of MDD, GAD, and suicidal ideation in the population over time. These trajectories were estimated by GAMs applied to 36,106 Lifelines participants. The x-axis denotes time with the corresponding month and year shown. The grey rectangles highlight the three different nationwide lockdowns in the Netherlands.

Major depressive disorder

We found a significant non-linear trajectory for the point prevalence of MDD (Figure 2, Table S3). The prevalence of MDD was relatively stable during the months of 2020 and increased rapidly at the start of the third hard lockdown. We observed a peak in MDD prevalence of 2.88% (95% CI: 2.71%–3.06%) at the beginning of March 2021, which then declined again the months after. Comparing the end of April 2021 with the end of April 2020, the prevalence of MDD was higher one year later (2.41% versus 1.78%;). The trajectory of MDD was similar across sensitivity analyses (Figure S5 and Table S7, S11).

Anxiety symptoms

Anxiety symptoms showed a significant non-linear trajectory with the highest number of symptoms reported at the immediate start of the pandemic (Figure 2, Table S3). Reported anxiety symptoms declined as the first lockdown progressed and were lower during mid-summer 2020 when government restrictions were eased. The prevalence of symptoms increased again from August and plateaued during the second lockdown. Anxiety symptoms moderately increased during the third lockdown but did not reach the level of symptoms reported at the start of the pandemic. The prevalence of symptoms declined as the end of the third lockdown approached in May 2021.

Comparing the end of April 2021 with the end of April 2020, the number of reported anxiety symptoms one year later was slightly higher (0.65 versus 0.59). Trajectories observed in our sensitivity analyses aligned with our findings from our main analyses (Figure S6 and Table S7, S11).

General anxiety disorder

For GAD, we found a significant non-linear trajectory. The prevalence of GAD was high at the beginning of the pandemic and declined as the first lockdown progressed, reaching its lowest prevalence at the start of July 2020. The months after, the prevalence had a roughly linear increase reaching its peak prevalence of 2.92% (95% CI: 2.76%-3.08%) at the beginning of March 2021 and declined again toward the end of the third lockdown. Comparing the end of April 2021 with the end of April 2020, the prevalence of GAD was higher one year later (2.71% versus 2.16%). The non-linear trajectories were similar in sensitivity analyses in the full cohort as well as after taking into account individual- and family-specific effects (Figure S5 and S7 and Table S2, S6, S10). The GAD trajectory did show a more smoothed pattern in the selected 5K subsample (Figure S5), which was likely due to lower statistical power. In the subsample of only younger participants, we observed a more similar trajectory as that of our main analysis.

Suicidal ideation

We observed a significant linear increase in the prevalence of suicidal ideation (Figure 2 and Table S3). At the beginning of April 2020, the prevalence of suicidal ideation was 0.59% (95% CI: 0.56%-0.62%) which increased to 0.95% (95% CI: 0.89%-1.01%) during mid-June 2021. This represents a 1.61x increase in reported suicidal ideation in the population. Sensitivity analyses in the full cohort yielded similar results. After taking into account individual-specific and family-specific variation the trajectory of prevalence was flat over time, indicating that most individuals in the population did not experience an increase in suicidal ideation and that specific individuals or subgroups may be more at risk (Figure S8 and Table S3, S7, S11).

Longitudinal trajectories across age, sex and history of MDD/GAD

As (symptoms of) MDD and GAD are known to be more prevalent in younger adults, women, and individuals with a previous diagnosis, we next investigated how the observed longitudinal trajectories differed by age, sex, and lifetime history of MDD/GAD in our sample.

Longitudinal trajectories by age

Younger participants reported significantly more depressive and anxiety symptoms, as well as a higher prevalence of MDD and GAD than older participants across all time points (Table S4, Figure S9-S12). The relative risk of young subjects compared to older subjects did not change across time for these four outcomes nor did they in our sensitivity analyses (Tables S4, S8, S12).

Suicidal ideation

At the start of the pandemic, younger participants also reported significantly more suicidal thoughts than older participants (e.g. 1.12% versus 0.52% in 20-versus 60-year old participants). Moreover, younger participants also reported a steeper increase in suicidal ideations over time as indicated by a significant interaction effect between time and age (Figure 3 and Table S4). By mid-June 2021, 20-year-old participants had 4.14x more reports of suicidal ideation than in March 2020 (4.64% (CI: 3.09%-6.96%) versus 1.12% (CI: 0.76%-1.66%)), whereas this increase was lower for older participants (i.e. 1.98x, 1.17x, 0.59x, 1.08x for 30-, 40-, 60-, and 80-year old participants, respectively). This interaction effect between time and age was also significant in our sensitivity analyses of the youngest 5,000 study participants, but not in our randomly selected subsample (Table S8 and S12, Figure S14 and S15). As suicidal ideation is a rarer phenotype, including more young adults increased our statistical power to detect such an effect.

• Download figure
• Open in new tab

Figure 3. The trajectory of suicidal ideation by time and age during the COVID-19 pandemic.

Shown are the results of interaction between time and age on reported suicidal ideation in our main sample of 36,106 participants (age 18-45). The left panels show the trajectory of suicidal ideation for specific ages over time. The right panels show the trajectory of specific time points across age. The legends at the top of the graph denoted color coding of groups. The gray rectangles highlight the three different nationwide lockdowns in the Netherlands.

Longitudinal trajectories by sex

As expected, the prevalence of MDD and GAD and their symptom scores were on average significantly higher in women than in men (similarly to findings in[10]), but the development of these outcomes over time did not differ between the sexes (Figure S16-S25 and Table S5, S9, S13). The prevalence and development of suicidal thoughts across time did not differ between women and men, which we also observed in our sensitivity analyses.

Longitudinal trajectories by lifetime history of MDD/GAD

Study participants with a lifetime history of MDD/GAD reported significantly more symptoms and a higher prevalence of MDD/GAD and suicidal ideation during the pandemic than participants without a previous diagnosis (Figure 4, Table S6). We found a significant difference in trajectories over time between participants with and without a history for MDD or GAD for depressive or anxiety symptom scores, respectively, while not for the other three outcomes (Table S6). For both depressive and anxiety symptom scores, the difference between participants with and without a lifetime diagnosis was greatest during periods out of lockdown when reported symptoms were lowest in the general population (Figure 4). Sensitivity analyses in the full cohort yielded similar findings, as did correcting for individual- and family-specific variation (Figure S26-S35 and Table S6, S10, S14).

• Download figure
• Open in new tab

Figure 4. The trajectory of depressive and anxiety symptom scores between study participants with and without a history of MDD/GAD during the COVID-19 pandemic.

Shown are the results of interaction between time and a lifetime history of MDD/GAD on symptom scores based on GAM analyses in our main analysis sample. The left panels show the trajectory of the mean score of reported depressive and anxiety symptoms between participants with (green) and without (blue) a history of MDD/GAD. The right panels show corresponding difference plots that visualize the trajectory of the difference between the two groups. The gray rectangles highlight the three different nationwide lockdowns in the Netherlands.

Discussion

To investigate the impact of the COVID-19 pandemic on mental health, we described the development of current (symptoms of) MDD, GAD and suicidal thoughts in a longitudinal sample from the Northern Dutch general population who were followed for more than a year during the COVID-19 pandemic across three nation-wide lockdowns.

In general, we observed a greater prevalence of (symptoms of) MDD and GAD in the population during periods of lockdown and a declining prevalence during periods of eased or no restrictions, which suggests that the pandemic and government measures enacted to restrict spreading of the coronavirus indeed impacted the mental health of the general population. We observed a high prevalence of symptoms of MDD and GAD at the start of the pandemic followed by a rapid decline in the months after, which is in line with observations from other countries[3, 17–19]. A new finding of our study is that the prevalence of symptoms of MDD and GAD increased again during later lockdowns alongside a high prevalence of MDD and GAD during the third lockdown which peaked in March 2021. The third lockdown in the Netherlands, unlike the first and second lockdown, was a hard lockdown and characterized by stricter measures, including an evening curfew. Interestingly, we observed a plateauing of reported symptoms during the second lockdown, which was a shorter partial lockdown, where schools, sports, libraries and museums, unlike during the third lockdown, stayed open. Together our results suggest that lockdown measures impacted mental health in the population, although we submit that this is an observational study of multifactorial psychiatric conditions.

Women, younger adults, and subjects with pre-existing mental health conditions, were more at risk for developing MDD and GAD symptoms and disorders, which is in line with observations from previous studies as well[17–19]. However, apart from a difference in intercept, we observed no differences in the development of MDD/GAD over time across sex or age. We did find a difference in the trajectory of reported MDD/GAD symptoms between participants with a history of MDD/GAD compared to participants without a history, but this difference was not found for MDD and GAD. Participants with a history of MDD/GAD reported increased symptom severity during the end of the first lockdown and months after, which is in contrast to a smaller Dutch study that found no difference at the start of the pandemic[20] and highlights the importance of continued longitudinal measurements in large cohorts.

In contrast to the non-linear trajectories of (symptoms of) MDD and GAD, the prevalence of suicidal ideation showed an increasing linear trend among young adults while we did not observe this effect in the overall population. There were no sex differences in prevalence rates of suicidality, but subjects with a lifetime history of MDD reported more suicidal ideation. A meta-analysis of suicidality during the pandemic found increased rates of suicidal ideation and behaviors with a higher incidence in younger individuals as well[2]. As only 14% of studies on suicidality during the pandemic included children or young people[21], there is an urgent need for more epidemiological studies, like the Lifelines COVID-19 study, that includes these groups to investigate what factors are driving the observed increase. As psychological distress, low perceived social support, and loneliness are known factors that increase suicidality in adolescents and young adults[22, 23], government measures to reduce the spread of the virus by social distancing measures likely contributed to the increased prevalence. Adolescents and young adults were furthermore significantly impacted by unemployment at the start of the pandemic [24]. As unemployment increases risk of suicide[25], our observation of increased suicidal thoughts may translate to suicide attempts and mortality[2]. Indeed, the Dutch Suicide Prevention Center reported an increase in completed suicides among young adults in January and February of 2021 compared to previous years [26]. These alarming findings warrant for alertness in psychiatric care services and urge governments to consider the long-term impact of pandemic measurements on young people.

A key strength of our study is the high-quality and high-resolution longitudinal data collected using a validated structured diagnostic interview throughout the first fifteen months of the pandemic spanning three nationwide lockdowns. This allowed for the application of sophisticated nonlinear statistical models to investigate the development of MDD, GAD and suicidality across time. However, our findings should be interpreted considering several limitations. First, due to the large computational resources required to run GAMs with random effects, we were only able to account for individual- and family-specific variation in a subsample of our cohort. While this subsample was similar to our larger sample in its main characteristics, we cannot exclude that our analyses may have missed important insights due to limited statistical power, especially for more rare phenotypes. We did show that analyses on the youngest study participants is worthwhile to prioritize as a significant proportion of the variation in MDD/GAD outcomes lies in younger adults, which increases statistical power to identify group differences. Second, we fitted each outcome in a separate model and thus could only compare the prevalence within a single trajectory and not between trajectories of different outcomes. As depression and anxiety are known to have comorbidity, how changes in trajectories relate among outcomes is an important question to investigate in future research. Third, we did not have information on the prevalence of MDD, GAD and suicidality in the year before the start of the pandemic and thus could not account for that, nor for seasonal effects. Fourth, we assessed current symptoms of GAD within the past seven and fourteen days. We therefore did not assess GAD according to the DSM-IV criteria that requires symptoms to be present for at least six months. Fifth, we did not account for corona infection status of study participants. As the number of participants who reported to have tested positive for the coronavirus during the study was relatively low (11.2%), we expect infection status and its accompanying symptoms to have minimal impact on our findings, if any at all, given that the observed trajectories in prevalence of MDD/GAD and suicidal ideation also do not follow the number of reported infections over time (Figure 1). Sixth, Lifelines participants are more often female, middle aged, married, and Dutch native compared to the population in the North of the Netherlands [27]. We therefore cannot exclude that there are population sub-groups who may be at greater risk of declining mental health that are under-represented in the Lifelines COVID 19 study. Finally, our findings should be interpreted within the societal context of the study. The Netherlands had three lockdown periods with different characteristics and is furthermore on average a rich country with a social welfare system.

In summary, we investigated the development of (symptoms of) MDD and GAD and suicidal thoughts in the Northern Dutch population during the COVID-19 pandemic and observed higher prevalence during periods of lockdown, in particular the third hard lockdown. We furthermore found an alarming linear increase in suicidal thoughts among young adults that warrants for alertness in psychiatric care services. Further studies are needed to investigate mechanisms underlying these rising prevalence rates. Our findings provide important insights into the impact of the pandemic on the mental health of the population, which can help guide policy makers and clinical care during future lockdowns and epi/pandemics.

Funding support

HvL was supported by a NARSAD Young Investigator Grant from the Brain & Behavior Research Foundation and a VENI grant from the Talent Programme of the Netherlands Organization of Scientific Research (NWO-ZonMW 09150161810021). The funding organizations had no impact on study design, data collection, analysis, or interpretation, or decision to submit the manuscript.

Conflicts of interest

The authors report no conflicts of interest.

Author contributions

All authors contributed to design of the study. The Lifelines Corona Research Initiative, AO and HvL were involved in data collection. AO, HvL and MW designed the study. AO performed the statistical analyses. AO drafted the manuscript; all other authors interpreted results and provided feedback on analyses and drafts of the manuscript.