Background

The COVID-19 pandemic was unprecedented in modern history in terms of its global impact, speed of progression, and the extraordinary measures taken to mitigate its effects. This event took the world off guard—not only posing a direct threat to physical health [1] but ushered in economic uncertainties [2] and psychological challenges like social isolation, lifestyle disruptions, and health fears.

The mental health of many individuals suffered and studies suggest that at the pandemic’s onset anxiety, depressive symptoms, and psychological distress (PD) were on the rise with people affected to varying degrees, depending on e.g. gender, financial status, or pre-existing medical conditions [3,4,5,6]. For example research has shown that certain populations, particularly healthcare workers, faced heightened levels of anxiety and depression due to the demands of their roles during the pandemic [7]. Yet, mental health deterioration associated with the pandemic was not permanently present for the majority, including those with pre-existing mental health disorders (MHD; [8,9,10,11].

The ability to adapt to these new circumstances and to alleviate mental burden can broadly be explained by several factors: by those that are linked to external factors altering the perception of e.g. health risks, financial risks, or impairments due to measures taken to minimize the propagation of the virus [12,13,14], and by individual psychological factors serving as protective or risk factors for mental health [15]. Concerning the latter, Miglani and colleagues demonstrated that resilience or healthy coping strategies might be a valuable mitigation strategy to reduce psychological distress [16]. Lazarus and Folkman’s Transactional Model of Stress and Coping posits that coping is a dynamic process involving continuous appraisal and reappraisal of stressors [17]. According to this framework, individuals first engage in primary appraisal, evaluating whether a situation is irrelevant, benign-positive, or stressful. If deemed stressful, they then assess their available resources and options in the secondary appraisal phase, determining how to cope with the stressor. The response to the stressor is mainly determined by the individuals’ resources, capabilities, personal preferences, or experiences. According to the model by Lazarus and Folkman, emotion-oriented coping (i.e. regulating or modifying the reaction towards the stressor) is often used when a situation is perceived as uncontrollable. On the other hand, problem-oriented coping (i.e. attempting to change the situation) is used when the situation is perceived as controllable. This theoretical framework provides a basis for understanding how pandemic-related stressors, such as health risks and financial uncertainties, may influence individuals’ selection of coping strategies and the associations with PD.

Austria and Italy were impacted particularly badly at the beginning of the COVID-19 pandemic, creating a breeding ground for various stressors. In early 2020, the Italian healthcare system was stretched to its limits, unable to care effectively for the numerous patients [18] and making COVID-19 the leading cause of death during that time [19, 20]. Within that same period, a large SARS-CoV-2 cluster formed in the Austrian municipality of Ischgl in Tyrol, marking the beginning of the virus spreading in Austria and Northern Europe [21]. However, early implementation of a strict control policy (e.g. lockdown) proved effective, leading to fewer deaths compared to Italy [22] but those precautions also had negative consequences [23].

In that first phase, coping behaviour was studied around the globe among individuals from the general population [24,25,26,27,28]. Götmann and Bechtoldt explored how individuals from Germany did cope with COVID-19 associated restrictions of personal freedom. Their analysis revealed that problem-focused coping (e.g. active coping, positive reframing) was positively associated with well-being, whereas avoidant-oriented coping (e.g. distraction, denial) had the opposite relationship [29]. Additional evidence was provided by Rogowska and colleagues, showing that frequent usage of emotion-focused coping strategies (e.g. self-blame, venting) was associated with higher stress levels in Polish students, whereas individuals engaging in problem-focused coping reported lower perceived stress [30]. Although some studies also included Austria and Italy [31,32,33], they primarily focused on healthcare personnel in the latter [7, 34,35,36].

Firstly, according to most recent studies, we hypothesized that PD would decrease as pandemic restrictions were lifted. Secondly, previous research left an equivocal picture of people’s coping utilization, showing that it can be characterized by both malleable and stable components. Recent studies suggest that normative stability (mean-level change) and differential variability (intra-individual change) of coping responses are predominate in the context of extreme or life-threatening circumstances and the context of the COVID-19 pandemic [37, 38]. Hence, we assumed that the extent to which people utilized coping would not vary between the dates of measurements (indicated by mean-score stability) for most strategies. Moreover, we supposed high flexibility in intra-individual changes for most coping responses. Thirdly, we hypothesized that coping responses’ impact on PD would not differ between measurements.

Materials and methods

Design, participants and setting

This online questionnaire-based panel study was conducted over a period of three years involving three annual measurements. The study was set up in Tyrol (Austria) and South Tyrol (Italy), characterized by a mix of urban and rural areas, allowing for a diverse representation of participants in terms of geographic location and lifestyle. This study utilized a non-probability sampling approach, specifically convenience sampling. The sample consists of individuals of legal age (18 +) from the general population with valid data entries. Participants were recruited via multiple channels, including social media platforms, flyers distributed in public places, and print media advertisements in local newspapers. These recruitment methods were chosen to maximize reach and participation across various demographic groups, although the sample may not fully represent the entire population. Figure 1 depicts the participants flow through the study. In December 2020, Tyrol experienced a two-week light lockdown between two hard lockdowns expanding over the entire baseline assessment. These hard lockdowns were characterized by rules that applied during the day (e.g. curfew, closures of most shops, leisure facilities, and service providers for body-related services; [39]. In South Tyrol, comparable hard lockdown measures were taken during baseline assessment, i.e. from 5 February 2021 (three weeks) and from 15 March 2021 until mid-April 2021. The first COVID-19 vaccinations in Austria and Italy were rolled out from 27 December 2020. At the beginning of 2022, while the second survey was enrolled, measures changed in both countries with an increased focus on vaccinations and general hygiene measures such as face masks. By the third survey in 2023, COVID-19 measures had largely been lifted in both countries. Special precautions remained in facilities treating/housing vulnerable people, such as hospitals.

Fig. 1
figure 1

Study procedure flow chart

Full size image

Participants characteristics and flow through the study

Figure 1 depicts the study procedure chart. In total, N = 850 were assessed for eligibility and signed informed consent. 26 participants had to be excluded due to invalid data, unverifiable or low age and eventually, data of 824 people with at least one measurement during the study period was available. This yielded a dataset of n = 1805 assessments that contained (partially) complete information regarding demographic and clinical variables. However, another 114 participants had to be excluded from the primary analysis due to missing BSCL and Brief COPE data. In total, n = 1691 observations with complete data were used for the primary analysis.

Variables and measures

Sociodemographic and clinical data

Sociodemographic data collected various factors, including gender, age, place of residence, relationship status, educational background, current employment status, and living arrangements. Participants were also asked if they had been exposed to or experienced more violence since the outbreak of the pandemic. In addition, they were requested to share information on their COVID-19 vaccination status and if they were tested positive for COVID-19, as well as any severe physical health issues and whether, and if so, what type of treatment they were currently receiving due to MHD.

Coping responses

Coping responses were measured with the German and Italian version of the Brief COPE [40,41,42]. This short version of the multidimensional coping inventory (COPE; [43] consists of 28 items evaluating 14 coping responses derived from theoretical frameworks [44,45,46]: acceptance, active coping, behavioural disengagement, denial, emotional support, humour, informational support, planning, positive reframing, religion, self-blame, self-distraction, substance use, and venting. Participants were instructed to assess the alignment of each statement with their thoughts and actions during challenging or unpleasant situations. Responses were provided on a 4-point scale, with 1 indicating ‘not at all’, 2 representing ‘a little bit’, 3 signifying ‘medium amount’, and 4 implying ‘a lot’. Mean scores for each coping response were calculated resulting in a possible scale range of 1–4.

Psychological distress

PD was assessed using the German and Italian version of the Brief-Symptom-Checklist (BSCL; [47, 48]. This self-report instrument consists of 53 items that participants rate on a 5-point Likert-type scale ranging from 0 ‘not at all’, 1 ‘a little bit’, 2 ‘moderately’, 3 ‘quite a bit’ to 4 ‘extremely’. Participants were asked to indicate complaints they had had in the last seven days. The BSCL measures PD across nine sub-domains, each evaluated by specific subscales: anger-hostility, anxiety, depression, paranoid ideation, phobic anxiety, psychoticism, somatization, interpersonal sensitivity, and obsessive-compulsiveness.

The sub-domain scores were calculated by adding up the respective items’ values. Additionally, a composite score called the Global Summary (GS) was calculated by combining the sub-domains scores resulting in a possible scale range of 0–212. To allow meaningful interpretation, these raw scores were transformed into age- and gender-specific normative T-scores by using a standardization reference table. The average T-score lies within a mean of 50 ± one standard deviation (10 points). T-scores ≥ 63 indicate clinically significant PD.

Data collection

Data collection was realised through the Computer-based Health Evaluation System (CHES; [49]. For organizational reasons, the baseline assessment period (t0) differed between Austria (30 November 2020 to 24 January 2021) and Italy (8 February 2021 to 4 April 2021). The first follow-up assessment (t1) was conducted from 10 January 2022 to 21 February 2022, the second follow-up assessment (t2) took place between 9 January 2023 and 22 February 2023. During the assessments, participants were instructed to reflect specifically on their current circumstances.

Statistical methods

Statistical analysis

Statistical analysis was conducted with SPSS, version 29.0 [50] and R, version 4.2.3 [51] using the package nlme, version 3.1–163 [52] and ggplot2, version 3.4.4 [53]. The significance level was set to α = 0.05, conducted statistical tests were two-tailed. To adjust for multiple testing, Benjamini–Hochberg correction was applied with a false discovery control level of 5%. Participants without missing data regarding the Brief COPE and BSCL at least in one assessment were included in the primary analysis.

Sociodemographic data is described by mean, standard deviation, median, interquartile range, and counts. Sample properties in terms of coping responses (Brief COPE) and PD (BSCL) are given by means of scale’s reliability (McDonald’s ω; [54], proportions of response categories, and age- and gender-specific normative T-scores ≥ 63. Comparisons between men and women as well as between Tyrol (Austria) and South-Tyrol (Italy) regarding coping and psychological distress were provided in the Supplementary Material. By means of McNemar test marginal homogeneity regarding exposure to violence, increased propensity for violence, physical health problems and treatment due to MHD, and clinically significant PD T-Scores were analysed. Reported Cohen’s g and d effect sizes can be interpreted as g = 0.05 < 0.15; d = 0.20 < 0.50 small; g = 0.15 < 0.25; d = 0.50 < 0.80, medium; and large for g ≥ 0.25; d ≥ 0.80 [55].

First, potential within-group mean-level changes in PD sub-domain scores and GS as well as coping responses were analysed within a linear mixed model framework. Regarding coping responses, intraclass correlation coefficients (ICC) were calculated to assess differences in intra-individual change (i.e. differential stability). In case of non-normally distributed standardized residuals (assessed through skewness, kurtosis, QQ-plot and histogram), the BSCL variable was transformed appropriately (logarithmic) to provide robust standard errors and p-values. Since all statistical outliers fell within the range of the used scales, they were net excluded from analyses.

To test the third hypothesis a null-model including PD as dependent variable was initially specified. The unconditional ICC was then calculated, where τ2 is the between-subject variance and σ2 is the within-subject variance with values ranging from 0 to 1:

$$\text{ICC}=\frac{{\uptau }_{0}^{2}}{{\upsigma }^{2}+ {\uptau }_{0}^{2}}$$

Next, a series of univariate models (Model 1.1 – 1.14) with random intercept and fixed slope including each COPE sub-domain separately were established. Then the interaction term between coping responses × time was included (Model 2.1—2.14). Following univariate modelling we proceeded to a multivariable model including all COPE sub-domains attaining a p-value below 0.05 in the univariate model after Benjamini–Hochberg correction was applied. To account for possible differences both, the univariate and multivariable models additionally included age, gender and place of residence as covariate. Similarly to the univariate model building approach, the interaction between coping response × time was not included at first (Model 3), followed by a model including the interaction term (Model 4). In case the interaction term was not statistically significant, the model with parsimonious properties (i.e. fewer parameters) was chosen.

Multiple imputation and sensitivity analysis

A detailed description of the multiple imputation procedure and the sensitivity analysis can be found in the supplementary materials.

Results

Women accounted for almost three quarters (~ 72%) of the total sample. Participants’ mean age was in the mid-forties (baseline: Mean = 44.4, SD = 13.2), and nearly two thirds were from Austria (~ 62%). Over 76% reported having a stable relationship and living with another person in the same household. Children resided in 43% of households, of whom 47% were taught by their parents. More information regarding sociodemographic variables can be found in Table 1. At baseline 12.1% reported on having tested positive for COVID-19, increasing to 27.2% at first and 73.0% at second follow-up. 91.0% of participants were vaccinated at t1, 91.8% at second follow-up measurement.

Table 1 Sociodemographic data of participants at baseline (t0) and follow-up (t1 & t2)
Full size table

Clinically relevant PD was evident in almost 25% of the sample (Table 2). T-scores in the clinically relevant range were particularly found in terms of phobic anxiety at baseline (35.6%) and the first follow-up (34.5%) measurement. However, this prevalence decreased significantly at the third measurement (25.4%; p < 0.0001).

Table 2 T-scores, means, standard deviations, and reliability of BSCL scales at baseline (t0) and follow-up (t1 & t2)
Full size table

Supplementary Table 1 depicts participants’ exposure and propensity to violence, existing physical health problems, and treatment modalities due to MHD. According to McNemar’s test, the propensity for violence increased significantly between baseline (12.2%) and the second follow-up (18.6%). When analysing men and women (Supplementary Table 8) separately, this increase became not statistically significant anymore.

Primary analysis

The calculated ICC was 0.796, indicating that 79.6% of the variance of PD could be accounted for by between-subject differences and 20.4% by within-subject changes between consecutive measurements. Next, a significant decrease in the BSCL total score from first to second follow-up (β = -2.856, d = 0.08; Table 2) was evident. Sub-domain analysis indicated decreased scores for anger-hostility between t0 and t2 (β = -0.071, d = 0.12) as well as between t1 and t2 (β = -0.081, d = 0.13), depression between t1 and t2 (β = -0.069, d = 0.06), and phobic anxiety between t0 and t2 (β = -0.121, d = 0.16) as well as between t1 and t2 (β = -0.129, d = 0.19). Increased scores were found regarding somatization between t0 and t1 (β = 0.047, d = 0.10) and in terms of obsessive-compulsiveness between t0 and t1 (β = 0.085, d = 0.09). Interpersonal sensitivity increased between t0 and t1 (β = 0.065, d = 0.09) and decreased again between t1 and t2 (β = -0.063, d = 0.07).

Regarding mean-level changes in coping responses, behavioural disengagement increased between baseline and both follow-up measurements (βt0-t1 = 0.113, d = 0.17; βt0-t2 = 0.144, d = 0.24, Table 3). Positive reframing marginally shrunk between baseline and second follow-up (β = -0.082, d = 0.11), whereas the humour mean value was slightly lower at first follow-up compared to baseline (β = -0.081, d = 0.10). Calculated ICC indicated moderate differential stability for most coping responses ranging from 0.419 (behavioural disengagement) to 0.571 (self-blame). Higher stability for coping responses within persons across measurements was found in terms of humour (ICC = 0.621), substance use (ICC = 0.721), and religion (ICC = 0.785). The score distributions of coping responses at baseline and follow-up measurements are depicted in Supplementary Fig. 1.

Table 3 Response portions, means, standard deviations and reliability of Brief COPE scales at baseline (t0) and follow-ups (t1 & t2)
Full size table

Irrespective of the time of assessment (Model 1, Supplementary Table 2), acceptance, active coping, emotional support, humour, and positive reframing had in general a positive effect on PD (negative association). In turn, behavioural disengagement, denial, self-blame, self-distraction, and substance use had a negative effect on PD (positive association). The use of informational support, planning, religion, and venting were not significantly associated with PD.

After the interaction term with the time of measurement was included (Model 2), results indicated a significantly lower effect of behavioural disengagement and denial on PD at t2 compared to t0. Moreover, venting became statistically significant, showing a positive link with PD at baseline. At third measurement planning and venting were negatively associated with PD. Since the data was positively skewed, the same analysis was conducted after the dependent variable had been logarithmically transformed. Results then indicated a significant interaction for planning and venting only.

Model 3 was the multivariable model including all coping responses still attaining statistical significance after Benjamini–Hochberg correction was applied. Hence, use of informational support, planning, religion, and venting were excluded. Results indicated that humour was no longer significantly associated with PD (Model 3, Supplementary Table 3). The highest positive associations (i.e. negative effect) between coping responses and PD were evident for substance use and self-blame, followed by denial, behavioural disengagement, and self-distraction. Concerning positive effects (i.e. negatively associated) on PD, active coping and emotional support could be identified as most influential, followed by positive reframing and acceptance. Similarly to Model 2, Model 4 included the time of measurement as interaction term, which was not statistically significant concerning any coping response. Model 3 was more parsimonious compared to Model 4. Thus, it is primarily used for the discussion of the results.

Secondary gender and country specific analysis

Comparing men and women did reveal statistically significant differences at baseline for anger-hostility, anxiety, somatization and global severity, with higher scores for women (Supplementary Table 9). Interpersonal sensitivity was lower in men at all assessments. Both populations, from Tyrol and South-Tyrol did not differ regarding PD responses (Supplementary Table 11).

Men and women differed significantly (i.e. men scored lower) at all measurements concerning the following coping responses: emotional support, informational support, venting. Active coping (t2), religion (t2 & t3) was higher, and humour (t0 & t1), substance use (t1 & t2) lower in women, compared to men (Supplementary Table 10). Tyrolean and South-Tyrolean participants differed concerning emotional support, humour, religion, and substance use at t0, t1 and t2 (Supplementary Table 12).

Sensitivity analysis

According to the results by multivariable logistic regression analysis, none of the included predictor variables, i.e. age, gender, residence, relationship status, severe physical illness, and treatment due to MHD, were significant predictors for missing data (Supplementary Table 4). Little’s test of missing completely at random was statistically not significant (χ2[1399] = 1328.2, p = 0.9110).

The model comparison (non-pooled Model 3 vs. pooled Model 3) showed mostly consistent results in terms of effect size and confidence intervals. Point estimates regarding self-distraction, substance use, and denial showed highest deviations compared to the non-pooled model but were well within the range of the respective confidence intervals, thereby indicating no statistically significant differences between the two models. Furthermore, according to the pooled model point-estimate positive reframing had a higher positive impact on PD compared to emotional support. Detailed results of the pooled models can be found in the Supplementary Tables 6 and 7. The coping responses’ effects on PD for both models are displayed in Fig. 2.

Fig. 2
figure 2

Lollipop plot visualizing the effect of coping responses on psychological distress. Abbreviations. n.s. = not significant at the α = 0.05 level. Note. Plot shows model estimates with 95% confidence intervals. Variable order is based on the estimates’ size (highest positive to highest negative) in the non-pooled Model 3. Positive estimates are associated with increased psychological distress, negative estimates are associated with decreased psychological distress. Included covariates: gender, age, residence. Both models did not include the following coping responses as independent variables due to non-significant results in the univariate analysis: Informational support, planning, religion, and venting

Full size image

Discussion

The aim of our study was to investigate whether PD and coping behaviour changed over the course of the pandemic and to what extent coping responses are related to PD.

Contrary to our hypothesis, the results indicated only marginally decreased overall PD from first to second follow-up measurement. It was found that particularly the proportion of clinically relevant phobic anxiety values subsided.

As assumed, mean-level changes between measurements were only apparent for 3 out of 14 coping responses. Behavioural disengagement increased from baseline to subsequent measurements, whereas positive reframing and humour were slightly lower at follow-up compared to baseline. Differential stability was moderate for most coping responses and high for humour, substance use, and religion.

Substance use, self-blame, denial, behavioural disengagement, and self-distraction predicted increased PD, whereas active coping, emotional support, positive reframing, and acceptance, in turn, were predictors for decreased PD. These associations were stable over time, which was verified by sensitivity analysis. Results are discussed more thoroughly in the subsequent sections.

Psychological distress in the general population during and after the pandemic

Our analysis yielded decreased PD levels from first to second follow-up measurement. Considering the sample’s average scores in 2020/2021 (29.08), 2022 (31.01), and 2023 (28.70) it becomes evident that this change (effect size: dt1-t2 = 0.08) is more statistically significant than clinically relevant.

Although meta-analytic studies predominantly show that mental health deteriorated at the beginning of the pandemic and returned to the pre-pandemic levels over time, these findings are not homogeneous among samples [56, 57]. In fact, variability between subgroups is evident, but it was also highlighted that a major proportion of the population has high resilience capabilities and thus experienced no to very little mental health impairments during the pandemic [58, 59]. This may explainmarginal changes in PD in our sample.

The degree of psychological distress observed in our sample could also partly be explained on a sub-domain level, considering the constructs measured with the BSCL questionnaire. For instance, somatization scores were increased at first follow-up and measures among other symptoms e.g. ‘pains in the heart or chest’, ‘trouble getting your breath’, ‘hot and cold spells’, and ‘feeling weak in parts of your body’. Moreover, the obsessive-compulsiveness domain also includes symptoms related to ‘trouble remembering things’, ‘your mind going blank’ or ‘trouble concentrating’. These symptoms are commonly observed in patients during and after a SARS-CoV-2 infection [60]. Given the increasing numbers of individuals reporting on having a COVID-19 infection, it can be assumed that many of the respondents were currently experiencing or have recently experienced at least some extend of these symptoms. Moreover, we observed decreased mean and clinically relevant T-scores between consecutive measurements for phobic anxiety, measured by items like ‘Feeling afraid to travel on buses, subways, or trains’ or ‘Feeling uneasy in crowds’. Since their roll-out COVID-19 vaccines were quickly spread among the population with over 90% being vaccinated in our sample. This and the decrease in the life-threatening potential of the virus itself could have contributed to the fact that people's perception of the virus has changed.

Drapeau and colleagues [61] suggests that the prevalence of PD ranges from 5 to 27% in the general population but can reach even higher levels in certain population groupings or when someone is exposed to stressful events and life conditions. For example, some studies found that compared to men, women exhibited increased PD during and after pandemic-associated lockdowns [62, 63]. Yoshioka and colleagues found that, among women, caregiving burden, domestic violence and fear of COVID-19 were independently associated with higher levels of psychological distress [64]. According to a study by Kim and Royle, these periods were also associated with increased rates of domestic violence against women [65]. These findings are somewhat contrary to our results. Although, we found slightly higher overall PD (and on some sub-domains) in women, compared to men at baseline measurement (winter 2020 – spring 2021), exposure to domestic violence was not higher for women between measurements when compared to men.

The pandemic’s repercussions continue to be observed in the economy and labour markets. The tourism and hospitality sector make up much of the (South-) Tyrol’s economic strength and provided employment to almost 20% of study participants. These recovering industries were severely impacted, resulting in a shocked labour market where unemployment rates rose by nearly 100%, from 4.7% to 12.9%; [66, 67]. In line with findings from the US, related job insecurity can be expected to be associated with decreased mental health [68]. In addition, the events of the war in Ukraine exacerbated many of the economic difficulties and caused humanitarian catastrophes that affected people beyond Ukraine's borders [69, 70].

Variability of coping behaviour

In terms of normative stability, we observed that, with the exception of three responses, coping behaviour remained stable over the three-year survey period. Behavioural disengagement marginally increased from baseline to follow-up, humour slightly decreased from baseline to first follow-up, and positive reframing also slightly decreased between baseline and second follow-up. Although these differences were statistically significant, their magnitude and associated effect sizes were rather unsubstantial, thus favouring a trait-like disposition.

In line with our findings, Nielsen and colleagues [38] have shown that mean-levels of coping were hardly susceptible to change over a 2-year period in a sample of employed Norwegian adults. Next to positive reframing they identified small differences for informational and emotional support, self-distraction, and self-blaming. Except for active coping, venting, and acceptance, Godor and van der Hallen [37] found similar results in a sample from the Netherlands before and during the COVID-19 pandemic.

Looking at the covariance analysis in our study, at first glance there seems to be a contradiction as we obtained moderate differential stability results for most coping responses, providing arguments in favour of state-like disposition. Yet, previous research also supports these findings [38, 71, 72]. According to a study by Louvet and colleagues [73], this equivocality might be explained by possible subgroup differences that could have balanced each other out, concealing meaningful changes in individuals behind non-significant mean-level results.

Our sample consisted of a broad cross-section of society from North- and South Tyrol, including people with different educational backgrounds, resources, and obligations (e.g., childcare, homeschooling, treatment due to MHD, etc.). Given that coping behaviour is influenced by cognitive appraisals, situational factors, resources, previous experiences, and personal attributes [17], it is conceivable that a mixture of these factors has contributed to the heterogeneity in coping utilization across measurements. For instance, parents of young children could have perceived lockdown-related school closure as uncontrollable, very stressful, and an important event, which possibly led to adapted coping utilization. On the other hand, people without young children may have attached little importance to this restriction and may therefore have been able to maintain their coping strategies. These adults also may have had a harder time adjusting their coping styles during the pandemic, since their pre-pandemic coping strategies likely involved frequent social activities, such as dining out, attending bars, or engaging in other external leisure activities. When these outlets were suddenly unavailable due to restrictions, those without children may have struggled more to adapt, as their typical means of coping were heavily reliant on social and public environments.

In addition, resources such as social support were available to varying degrees in subgroups during the pandemic [74,75,76], which may have had an impact on susceptibility to situational factors. Interestingly, we observed both mean-level stability and high ICCs in religion and substance use coping with both responses being used to a low extent. In terms of clinical implications, this suggests that these strategies are far less adaptive in their nature compared to other coping responses.

Impact of coping behaviour on psychological distress

Our study’s findings highlight a crucial distinction between maladaptive and adaptive coping mechanisms. Positive associations with PD were evident for maladaptive coping strategies such as substance use, self-blame, denial, behavioural disengagement, and self-distraction. In contrast, adaptive coping strategies like active coping, emotional support, positive reframing, and acceptance were negatively associated with PD. Research across 30 countries reported similar associations for maladaptive coping with self-blame being the strongest predictor not only for lower mental health, but also for decreased physical health. Furthermore, humour, acceptance, and active coping predicted lower PD, with the latter also associated with increased physical health [31].

Conversely to our findings, emotional support was not [77, 78] or even positively [31] linked to PD in studies conducted during the pandemic’s initial phase. The wording used for emotional support in the Brief COPE includes statements like ‘I've been getting emotional support from others’, and ‘I've been getting comfort and understanding from someone’. It is possible that the relief experienced through supportive behaviour was intertwined with the level of symptoms and negative experiences already present in participants. Thus, individuals with high levels of distress may perceive this resource as not viable or even burdensome, while participants with lower levels of distress may find it helpful. This may have resulted in a bidirectional association with PD as described in the literature [79].

As previously demonstrated in a study by Avsec and collaborators [80], we also observed that the negative impact of avoidance coping (behavioural disengagement, denial, self-blame, and substance use) on PD was consistently higher than the positive impact of approach coping (e.g., acceptance, active coping, emotional support, and positive reframing). The latter strategies aim to reduce or eliminate the stressor. The changes and restrictions brought about by the pandemic were primarily the result of decisions not within the control of an individual. As a result, people were able to take only limited action against these restrictions, preventing these coping strategies from being fully effective. Another aspect is limited availability of resources and factors that can facilitate the application of certain coping behaviours. For example, it has been shown that next to social support, psychological flexibility also encourages the application of approach coping [81]. Nevertheless, the potential of coping strategies should not be ignored since positive reappraisal (comparable to positive reframing) was shown to be a protective factor with regard to PD during the pandemic [15].

Limitations and outlook

Our study is distinguished by the examination of coping behaviour and PD in a general population sample over the course of three years. Through sensitivity analysis we were able to confirm the results of the complete case analysis. Nonetheless, there are a few limitations to mention. Most importantly, representativeness of the sample may by impaired as attrition between consecutive measurements was not random. We also fail to answer the question of whether coping behaviour is a trait or a state construct. This question will likely not be answered singularly; rather, it is more probable that both are important to varying degrees [71]. Although we countered some of the criticisms raised by Carver and Connor-Smith [82], like a lack of longitudinal studies, or recording of coping categories that are too broad, further adaptations are necessary. They suggest investigating whether certain coping behaviours are perceived as more or less useful at different times and advocate for prospective studies within a laboratory setting facilitating standardised stressors and including daily surveys. Given that we did not conduct any intervention in our study, we cannot rule out the existence of reciprocal mechanisms of action. Studies suggest that mental health can influence coping behaviour [38]. Through standardized experimental procedures, researchers should be able to better investigate the cause-and-effect mechanisms as well as where differences in stress appraisal of individuals lie. In addition, the stable relationships between coping responses and PD shown in our study can also be taken into account when designing E-Health cognitive behavioural programmes (e.g. [83].