A cross-sectional online survey was used to investigate the physical and psychological effects of the coronavirus pandemic in the general population in Germany. We collected data during the recovery phase from the first wave, from May 12 to September 29, 2020, a phase with low numbers of daily new infections and COVID-19–associated deaths. Participants were recruited primarily through press releases (print, online), social media platforms (Twitter, Facebook), the websites at the Central Institute of Mental Health (CIMH), including a COVID-19 mental health support page, and the universities of Mainz and Konstanz. The Ethics Committee of the University of Mainz agreed to conduct the study (2020-JGU-psychEK-S010).

The inclusion criteria of the study were a minimum age of at least 16 years and written informed consent. The exclusion criteria included incomplete processing of the questionnaire and an unrealistically fast total survey completion time (DEG time < 100). In total, 2,224 people started the online survey and 2,160 participants completed it in a realistic processing time. Each participant was asked to report their gender, age, country of birth, highest level of education, employment status, and living situation. We also asked about their experiences with COVID-19 (current or past infection, COVID-19 symptoms, COVID-19–related risk factors, and fears of and perceived risk from COVID-19).

To assess somatic symptoms, psychological distress, and health anxiety both for the time before and after the beginning of the COVID-19 pandemic, participants were instructed to answer the same symptom measures twice: first for the current period (T1), then retrospectively for the period before the onset of the pandemic (defined as “the period between the end of February and beginning of March 2020”; T0; this comparatively brief time period was chosen for reasons of standardization and anchor point fixation between participants and in order to use a timeframe that is compatible with the PHQ-4 instruction regarding the previous two weeks).

First, we conducted descriptive analyses to describe sample characteristics. Second, we investigated changes in psychological and somatic distress and health anxiety between T0 and T1 using dependent sample t-tests for dimensional variables. Third, we conducted chi-square tests for categorical variables to investigate the effects of gender on the prevalence of depression, anxiety, and psychological distress. Fourth, we conducted Pearson correlations to explore associations between predictor and outcome variables. Finally, structural equation modeling was used to explore the independent relationships of predictor variables with outcome variables (current levels of distress as assessed with the PHQ-15 and PHQ-4, health anxiety, and COVID-19 fears). Two of the four outcome variables (PHQ-15 T1 and PHQ-4 T1) were modeled as latent variables. For the PHQ-15-T1 measurement model, a general-factor model with correlated error terms capturing the symptom-specific variance was applied. In case of the PHQ-4, we used a general factor model with correlated error terms between the two anxiety symptoms and the two depression symptoms, respectively. The following 16 predictor variables were entered simultaneously as manifest variables into a latent regression model: pre-existing symptom levels (somatic symptoms, anxiety and depression scores at T0), socio-demographic variables (age, gender, education), personality traits (extraversion, neuroticism, openness, conscientiousness, and agreeableness), pre-existing levels of health anxiety and avoidance behavior (illness information avoidance at T0), and COVID-19 variables (number of COVID-19 risk factors, COVID-19 risk perception, days after the peak of the first COVID-19 infection wave, current/past COVID-19 infection). The analysis was conducted in MPlus Version 7.3 (Muthèn and Muthèn, 2010) using the robust mean and variance adjusted weighted least squares (WLSMV) procedure. Because the chi² test is known for its sensitivity regarding sample size and model complexity, we additionally used common absolute (RMSEA) and comparative (CFI, TLI) fit indices for model fit evaluation. The remaining analyses were performed using SPSS Statistics Version 23 (IBM, Armonk, NY), and the level of significance was set at p ≤ 0.01 because even unimportant effects can be significant in large samples. Cohen’s d was calculated as effect size for t-tests (d ≥ 0.30 small effect, d ≥ 0.50 medium effect, d ≥ 0.80 large effect) (Cohen, 1988). A total of 2,160 participants completed the survey; however, 36 participants had missing values for the “education” variable and 10 participants reported a diverse sex and were excluded, leaving 2,114 participants included in the following analyses. For the predictor analysis, the variables gender and education were dichotomized (gender: female vs. male; education: less than 12 years of schooling vs. more than 12 years of schooling).

Of the 2,160 participants who completed the questionnaire, 74.8% were female, 24.7% were male and 0.5% were of diverse sex. The average age was 42.75 years (range: 16–86). Concerning their living situations, 20.9% reported living alone, 71.9% reported living with a partner, family or someone else, and 7.2% reported living in a shared apartment. With regard to education, 78.0% reported having at least 12 years of schooling, and 48.0% had a university degree. With regard to coronavirus, 0.2% said they were currently infected with the coronavirus, and 1.1% said they had been infected with the coronavirus in the past. Thus, the mean infection rate with the Corona virus was 1.3% in our sample compared to 0.24% in the German population (data reported by the Robert Koch Institute on July 21, the median of our survey period). 1.2% reported that a person close to them is currently infected with the coronavirus, and 12.2% reported having a close person who was infected in the past. At least 21.5% of the respondents reported having medium-to-severe fears about coronavirus infection, at least 22.8% expected their fears to remain moderate over the next four weeks, and 23.8% expected their fears to remain moderate to severe over the next eight weeks. At least 30% of the respondents estimated that they were at least 50% likely to become infected with the coronavirus in the future, 31.3% expected with a probability of at least 50% to become a carrier themselves, and 41.4% reported a medium-to-severe fear of becoming a carrier. 19.5% stated that they have been moderately to severely affected by the COVID-19 pandemic in their daily lives. With regard to COVID-19 risk factors, 12.9% were older than 60 years, 18.6% reported that they were smokers, 10.7% stated that they were extremely overweight, 12.7% had cardiovascular disease (e.g., coronary heart disease and hypertension), 5.6% had chronic lung disease [e.g., chronic obstructive pulmonary disease (COPD)], 1.5% had chronic liver disease, 3.0% had diabetes mellitus, 1.8% had cancer, and 7.0% said they had a weakened immune system (e.g., due to illness or regular medication).

Current research has revealed clear evidence that the “first wave” of the COVID-19 pandemic and the subsequent restrictive measures adopted to slow the spread of the virus are related to increased levels of depression, anxiety, and general distress in different populations around the world. However, changes in these symptom measures over different phases of the pandemic, especially levels of psychological and somatic distress in the recovery period, are heterogeneous. Accordingly, the present study examined the extent of psychological and somatic distress in the recovery period between the first and second waves of the COVID-19 pandemic in Germany. To investigate possible changes in distress levels in the German population, participants answered the same symptom measures twice: first for the current recovery period (T1), then retrospectively for the period before the onset of the COVID-19 pandemic (T0).

In our study, we observed elevated levels of psychological and somatic distress and health anxiety in the recovery period compared to before the pandemic. On average, participants rated their current symptoms of depression, anxiety, somatization, and health anxiety significantly higher than before the onset of the pandemic. Applying established cutoff scores for at least moderate levels of symptoms, approximately 25% of participants experienced psychological and somatic symptoms. Our prevalence rates before the outbreak (T0) are similar to those observed in representative German population samples in the years before the pandemic (Kocalevent et al., 2013; Hajek et al., 2020). In these nationally representative validation studies of the PHQ-4 and PHQ-15 screenings, the prevalence was 10.4% for depression, 9.8% for anxiety, and 9.3% for somatization syndromes. Interestingly, women had higher risk of both anxiety and somatization in these representative samples, which has also been shown in the present study. Our results are based on a convenience sample recruited online who were mostly women (75%) which may explain the slightly higher prevalence of anxiety and somatization at T0 compared to representative samples (Kocalevent et al., 2013; Hajek et al., 2020).

Furthermore, our results agree with findings from a German longitudinal observational study with four stages of online data collection from March 27, the punctum maximum of the first wave (phase two), to June 15, 2020, the beginning of the recovery phase (phase three) (Bendau et al., 2020). The authors observed only a slight decrease in psychological distress (PHQ-4 scores) from March to June with prevalence rates of 31.0% (T1), 25.9% (T2), 22.1% (T3), and 22.6% (T4). Their last assessment interval overlaps with the beginning of our study period and provided nearly identical prevalence rates to those assessed in the following months of the recovery phase in our study (from May to September 2020) using a comparable sample of the German general population. Another online survey that collected data over a 50-day period after the onset of the COVID-19 outbreak in Germany found a similar pattern of results: while COVID-19 fear decreased within six weeks to the level before the lockdown, generalized anxiety remained elevated over time (Hetkamp et al., 2020), indicating no return to the pre-pandemic level.

The study’s second aim was to identify significant predictors of distress and COVID-19 fears during the first recovery phase of the pandemic. As expected, and in line with prior research on the impact of pre-existing mental conditions on current mental health status (e.g., Fiorillo et al., 2020; González-Sanguino et al., 2020; McCracken et al., 2020; Newby et al., 2020), the strongest predictor of current levels of psychological and somatic distress was the domain-specific pre-existing distress level, retrospectively assessed for the period before the onset of the pandemic. Similar strong domain-specific associations were found for current health anxiety, which was best predicted by past health anxiety, and for COVID-19 fears, which were best predicted by higher levels of perceived risk of infection.

Furthermore, younger age was associated with higher psychological and somatic distress. These findings are consistent with previous research on distress during the early phase of the pandemic (e.g., Balsamo and Carlucci, 2020; Bäuerle et al., 2020a; González-Sanguino et al., 2020; Jia et al., 2020; Newby et al., 2020; Nwachukwu et al., 2020; Pierce et al., 2020; Ran et al., 2020; Rossi et al., 2020; Shevlin et al., 2020; Solomou and Constantinidou, 2020). Possible reasons include generally higher social mobility among younger people, little experience with socioeconomic or major life events or pandemics, and higher perceived threat of their academic, social occupational, and economic prospects compared to older people over 25 years (Huang and Zhao, 2020; Wang et al., 2021). As expected, high levels of neuroticism were significantly associated with all outcome measures in this study. Higher neuroticism scores predicted higher levels of psychological distress, somatic distress, health anxiety, and COVID-19 fears, whereas higher extraversion scores were significantly associated with lower levels of COVID-19 fears. These findings are consistent with the vulnerability model, which postulates that neuroticism is an important vulnerability factor for the development of unspecific mental distress and common mental disorders, including anxiety and depression (Jeronimus et al., 2016). In addition, extraversion has shown a robust positive link with subjective psychological well-being in a recent meta-analysis of the links between personality traits and well-being (Anglim et al., 2020).

In line with previous findings from the first acute phase of the pandemic (Jungmann and Witthöft, 2020; Sauer et al., 2020), pre-existing health anxiety predicted current health anxiety and COVID-19 fears but not distress. Moreover, the tendency to avoid illness-related information (e.g., documentaries about illnesses), which is often used as avoidance behavior by health-anxious individuals (Weck et al., 2013), was significantly associated with current health anxiety and with higher levels of current psychological and somatic distress and COVID-19 fears. These findings are consistent with previous research, which has shown that health anxiety, neuroticism, and coronavirus anxiety were significant predictors of depression, generalized anxiety, and death anxiety experienced during the COVID-19 crisis in the United States (Lee et al., 2020). Approximately half (46%) of the current sample surveyed had one or more risk factors to suffer a more severe course of COVID-19. Our results confirm the impact of self-reported biological COVID-19 risk factors (i.e., age > 60 years, smoking, overweight, cardiovascular and other somatic diseases) on psychological and somatic distress and COVID-19 fears. Furthermore, higher subjective risk perception, including the risk of contracting the virus or infecting someone else, was associated with higher COVID-19 fears, but not with higher psychological distress or health anxiety. The number of days after the peak of the first COVID-19 infection wave was also significantly associated with higher COVID-19 fears and elevated levels symptoms of depression and anxiety, which seems consistent against the background of rising infection numbers again in Germany from September onward. Finally, a current or past coronavirus infection predicted lower covid-19 anxiety, which seems plausible against the background of an expected temporary immunization against further Coronavirus infection, as discussed in Germany at that time (Robert Koch Institute, 2021).

The major strength of this study relates to the examination of a variety of different variables that predict coronavirus-related psychological distress, somatic symptoms, health anxiety, and COVID-19 fears using structural equation modeling. In addition, to the best of our knowledge, this is one of the first studies to investigate the possible effects of COVID-19 on mental health during the complete recovery phase between the first and the second “waves” in Germany. Nonetheless, several limitations of the present study that may limit the interpretation of our findings must be considered.

The first limitation is that the data at T0 were assessed retrospectively, which might have introduced a memory bias in terms of remembering disorder-specific symptoms, stress and behaviors. There are studies that indicate that retrospective reports of change in mental health are prone to substantial bias during the Corona pandemic (Hipp et al., 2020) and beyond [e.g., (Ben-Zeev et al., 2009; Van den Bergh and Walentynowicz, 2016)]. Despite this substantial limitation of the significance of our findings, the comparability of our retrospective results with cross-sectional prevalence data assessed in the years before the pandemic (Kocalevent et al., 2013) argues against the presence of such a bias. Furthermore, there is evidence that potential bias at the individual survey level is reduced by aggregating data (Jaspers et al., 2009). Additionally, various empirically recommended criteria for ensuring the highest possible reliability and validity of the study’s statements despite retrospective questioning were followed in our study; these include the use of short, easy-to-understand questions, as well as the use of fixed anchor points (for recommendation of retrospective survey questions in COVID-19 studies see Hipp et al., 2020). Furthermore, there are no studies to date that have investigated the validity of our measurement instruments in retrospective use. Due to the sudden spread of COVID-19, it was not possible to implement a longitudinal design with our participants before the outbreak; we decided to adopt an economical solution by applying these partly unvalidated self-report questionnaires. Future studies should incorporate newly published instruments to assess COVID19 related fear (e.g., Ahorsu et al., 2020; Lee et al., 2020; Taylor et al., 2020).

Second, our sample was recruited as a convenience sample mainly through social media and a COVID-19 mental health support page of the CIMH, which might have led to a sample bias. People who have easy access to or are familiar with social media might have been overrepresented in this study. Furthermore, as previous studies indicate (Bendau et al., 2020), people who experience a relatively high level of COVID-19–related psychological distress and who are looking for support might be particularly attracted by social media and might have been more likely to participate in our study. Additionally, the number of men was lower than that of women. This selection bias may partly overestimate the symptom severity and impact of COVID-19, especially given past studies have shown worse impact of pandemics on those with pre-existing mental illness, of younger age and in the female gender (Balsamo and Carlucci, 2020; Bäuerle et al., 2020a; González-Sanguino et al., 2020; Hajek et al., 2020; Jia et al., 2020; Newby et al., 2020; Nwachukwu et al., 2020; Pierce et al., 2020; Ran et al., 2020; Rossi et al., 2020; Shevlin et al., 2020; Solomou and Constantinidou, 2020).

Third, there are several differences between the demographics in our sample and the general population in Germany. Our sample consisted of 74.8% female participants. Since previous studies on the 12-month prevalence of mental disorders in Germany indicate a higher prevalence in women (33% versus 22% in men), the rate and intensity of depressive, anxious, somatization or health anxiety symptoms might be biased (Jacobi et al., 2014). Regarding sociodemographic variables relevant to corona risk, the proportion of participants 60 years and older was lower in our sample than in the general German population [12.7% in our sample versus 29% in the German population; (Statistisches Bundesamt, 2020)]. In contrast, a larger proportion of people than in the general German population were affected by conditions that, according to the Robert Koch Institute, increase the likelihood of corona infection or a severe course [e.g., coronary heart disease; 12.7% in our sample versus 9.3% in the German population; (Gößwald et al., 2013)]. Since our results suggest that younger people but at the same time those with more corona risk factors might be affected by anxiety, depression, health anxiety or somatization symptoms, these differences between our sample and the general German population represent limitations in the generalizability of the findings of this study. In addition, the prevalence rates of corona risk factors were solely based on self-report (e.g., when asked about a weak immune system), which may be subject to social desirability, self-report errors and poor recall and must be considered when interpreting the results. Furthermore, the proportion of people being currently or in the past infected with the Coronavirus was about five times higher among our study participants compared to the German population. Given the Structural Equation Model (SEM) finding that current or past coronavirus infection was associated with lower COVID-19 fears, our results may represent lower levels of COVID-19 fears compared to the general population, limiting the generalizability of the findings.

Finally, although we selected potential predictor variables based on previous studies examining psychological distress under COVID-19, additional important predictors might exist that should be examined in future research.