Introduction

The COVID-19 pandemic has placed substantial psychological demands on healthcare workers (HCWs) worldwide. Studies have pointed out hightened psychological distress of HCWs due to the pandemic, which was a major precursor of other mental health outcomes, such as depressive symptoms1. Psychological distress refers to a state of emotional suffering characterized by symptoms such as anxiety, depression, irritability, and feelings of being overwhelmed or unable to cope2. While it is a normal response to challenging situations, persistent psychological distress can interfere with daily functioning and may lead to mental health disorders if not addressed. As the pandemic drew to a close, psychological distress and other mental health issues decreased, returning to levels comparable to those observed at the pandemic´s start3. Behind such temporal pattern may also be changes in stressors experienced by HCWs during different points of the pandemic. Some stressors (such as lack of personal protective equipment) were especially high in initial phases, while other (cummulative exposure to deaths due to COVID-19) may have played a role at a later time point. Other stressors to HCWs have included fears of transmitting the virus to family members, redeployment to unfamiliar roles, increased risk of infection due to close contact with COVID-19 patients, stigmatization, and lack of organizational support2,4,5.

Despite the insights already gained, there remains a need for long-term research to understand the evolving nature of psychological distress and its predictors among HCWs. Relatively few studies have examined how levels of psychological distress and stressors evolve over time within this population. The longitudinal patterns of stress and mental health outcomes indicate that the predictors of psychological distress can shift over time, complicating efforts to support HCWs effectively6,7. A study of intensive care unit HCWs in Canada revealed that predictors of distress shifted across different phases of the pandemic, with gender differences in distress being particularly pronounced during the early stages6. However, some factors showed a consistent pattern. A systematic review of 139 studies further highlighted that being a woman, working as a nurse, experiencing stigma, and having direct contact with infected patients were risk factors for psychological distress among HCWs during the entire pandemic8.

Capitalizing on a unique sample of HCWs in the Czech Republic that has been followed since the start of the pandemic towards its concluding phase, the goal of our study was to prospectively follow the impact of COVID-19 related stressors on the dynamic change in psychological distress. A secondary aim was to determine if the relationships between the stressors and psychological distress differ for men and women.

Methods

Participants

We conducted our research with participants from the Czech arm of the international COVID-19 HEalth caRe wOrkErS (HEROES) Study. This global prospective cohort study aims to evaluate the impact of the COVID-19 pandemic on the mental health of HCWs across 22 countries on four continents, as previously described9. Data on demographics, mental health, support needs, and provisions were gathered via an online questionnaire. In the Czech Republic, data collection began in June 2020 (wave 0: June 24 to August 30, n = 1,778) after the initial peak of the pandemic in this country subsided. A follow-up survey was conducted in spring 2021 (wave 1: February 15 to April 31, n = 1,840) during the second peak of the pandemic. The last data collection took place in fall 2022 towards the end of the pandemic (wave 2: September 15 to November 15, n = 1,451). Eligible respondents included healthcare or social service workers such as physicians, nurses, paramedics, and social workers, as well as non-clinical, technical, and administrative personnel. Participants were invited through hospital administrators, scientific and professional societies. All methods were carried out in accordance with relevant guidelines and regulations. All participants provided an informed consent. The study was approved by the Ethical Review Board of the University Hospital Motol Prague, Czech Republic.

Distress

The outcome of this study, psychological distress, was assessed using the 12 items General Health Questionnaire (GHQ-12)10. The GHQ-12 focuses on the respondent’s current state and asks questions about experiences during the last week, covering aspects such as feelings of strain, depression, anxiety, social dysfunction, and loss of confidence. The GHQ-12 consists of 12 questions, each with four possible responses ranging from “not at all” to “much more than usual.” The answers were scored with the Likert scale method (0–1-2–3) and the final score was calculated as the sum of the individual answers (0–36 points), where higher scores indicate greater levels of psychological distress. GHQ-12 was found to effectively detect psychological distress in HCWs11. We are using this measure as a continuous variable in all analyses. From this point forward, we will refer to this construct as „distress“.

COVID-19 related stressors

The exposure in this study, a composite measure of stressors, was calculated similarly to our previous work3. We created a sum of six stressful situations the HCWs were exposed to during the COVID-19 pandemic. To address the varying numbers and character of stressors through the pandemic, we calculated the composite separately for each wave and we used only questions that occurred in all three waves. Each of the stressors was recoded as a binary variable (1 as exposed, 0 as unexposed). The stressors were summed together, yielding a maximum value of six in each wave. The individual stressors were as follows: (1) low trust in workplace: trust in the ability of the workplace to handle the pandemic (very low or low as 1, moderate to very high as 0); (2) assignment of new tasks: being assigned to a new team or new tasks from the beginning of the pandemic (yes 1, no 0); (3) prioritization of patients: having to prioritize patients due to COVID-19 (yes 1, options „no“ and „does not apply“ 0, option “prefer not to respond” from the wave 0 was coded as missing prior to imputation); (4) experience of death at work: close contact at work with someone who later died of COVID-19 (yes 1, no 0, option “I don’t know” that was only in wave 0 was coded as missing prior to imputation); (5) experience of discrimination, stigmatization or violence: feeling discriminated or stigmatized as HCW due to COVID-19 pandemic or experiencing violence during the pandemic (agree or strongly agree to any of the statements 1, disagree or strongly disagree 0); (6) contact with COVID-19 patients: close contact with suspected or confirmed COVID-19 disease in the last seven days (yes 1, no or don´t know 0).

Other characteristics

Other characteristics included baseline age (in years, centred at median), gender (man, woman) and occupation, which was originally coded as physician, nurse or other medical staff, management, and other. For the purpose of the model convergence in this analysis, we created dummy binary variables „physician“ (1 if the occupation reported at baseline was physician, else 0), and „nurse or other medical staff“ (1 if the occupation reported at baseline was nurse or other medical staff, else 0).

Statistical analysis

The Czech arm of the HEROES study consists of 3,624 respondents who participated in at least one wave of the Czech arm of the HEROES study, and 1,123 took part in at least two waves. We have then excluded participants who reported gender or age inconsistently across waves (n = 10) or categorised themselves as other gender (n= 3) due to the small size of this group. Prior to analysis, we performed multivariate imputation by chained equations in R12 (see details in the Supplement). The missing data in the GHQ were imputed if a minimum of 11 out of 12 items were filled for a given respondent, so the total GHQ score could be calculated. We kept in the final analytical sample the participants with available data on the outcome in all three waves, leading to the final sample of 264 respondents (Supplementary Figure S1). Comparing the analytical sample with not included participants who had at least two observations, we found that the respondents in the studied sample had a slightly larger proportion of physicians and slightly lower level of distress in wave 1 (Supplementary Table S1). In our main analysis, we employed a path model to investigate the relationships among the outcome (distress), exposure (stressors) and covariates (see the Supplement). The analysis was conducted using a structural equation modelling technique available in R with the lavaan package, version 0.6–18 with a maximum likelihood estimator. Our model included autoregressive paths, which account for the influence of previous observations on subsequent outcomes. Additionally, we specified paths from exposures (stressors) to outcomes (distress), incorporating both the direct effects of exposures on concurrent outcomes and the lagged effects of exposures from previous observations on recent outcomes. This approach allowed us to capture the longitudinal effects of exposure on the outcome over time. Covariates (age, gender, occupation) were included in the model to control for their potential confounding effects, with each covariate being assessed at baseline. The resulting path model was visualised with a dedicated plotting tool in R, specifically semPlot, version 1.1.6. In the end, we tested effect modification by gender in the associations between stressors and distress by including an interaction term. In addition, we performed a secondary analysis by repeating the path model with individual stressors as exposures to identify, which specific stressors may have had particular impact. All analyses were conducted using R version 4.4.1.

Results

We studied 264 participants (74% women, median baseline age 46 years), of whom 38% worked as physicians and 36% as nurses or other medical staff (Table 1). The distribution of stressors was lowest in wave 0 (mean 1.0 ± SD 1.0), almost doubled in wave 1 (mean 1.8 ± 1.4) and then decreased in wave 2 (mean 1.2 ± 1.1). The distribution of distress followed the same pattern: it was lowest in wave 0 (mean 11.7 ± 4.7), highest in wave 1 (mean 13.8 ± 5.4) and then lower again in wave 2 (mean 11.3 ± 4.7). Women reported slightly higher levels of distress but lower levels of stressors throughout all waves when compared to men.

Table 1 Baseline characteristics of the sample (n = 264).
Full size table

Table 2; Fig. 1 present summary results of the path model, while Supplementary Table S1 presents complete infomation on the model results. The following regression paths were estimated: greater distress in wave 1 was significantly predicted by distress in wave 0 (β = 0.53, standard error (SE) = 0.06, p < 0.001). Similarly, greater distress in wave 2 was significantly predicted by distress in wave 1 (β = 0.35, SE = 0.05, p < 0.001). Greater distress in wave 0 was significantly associated with stressors in wave 0 (β = 0.82, SE = 0.28, p = 0.004). In line with that, greater distress in wave 1 was significantly related to stressors in wave 1 (β = 0.85, SE = 0.24, p < 0.001), but not to stressors in wave 0 (p = 0.302). Distress in wave 2 was not significantly predicted by stressors in wave 1 (p = 0.488) or related to stressors in wave 2 (p = 0.329).

Table 2 Results of the path model.
Full size table
Fig. 1
figure 1

Path model visualisation. Coloured lines (blue, orange) represent statistically significant results with blue suggesting negative value of β and orange suggesting positive value of β. Grey lines represent statistically insignificant results. Note: W0, W1, W2 = wave 0, wave 1, wave 2. Nurse = nurse or other medical staff.

Full size image

The covariates also showed varying associations: being a man was a significant predictor of lower distress in wave 0 (β = −1.59, SE = 0.65, p = 0.014) and wave 1 (β = −1.72, SE = 0.66, p = 0.009), but not wave 2 (p = 0.449). Neither age or occupation predicted distress at any wave. We did not detect any interaction between stressors and gender in any wave.

Additionally, we ran linear regression models to estimate total effects of stressors on distress in each wave (Supplementary Table S2). In all waves, the exposure’s total effect was statistically significant. In contrast with direct effects in the path model, we found that greater level of stressors in wave 2 was significantly associated with greater distress in wave 2 (β = 0.56, SE = 0.27, p = 0.038) and that being a man was to some extent related to lower distress in wave 2 (β = −1.24, SE = 0.67, p = 0.063).

A secondary analysis using individual stressors as exposures in the path model (Supplementary Table S3) demonstrated that psychological distress was associated with the experience of discrimination, stigmatization or violence (all waves), low trust in workplace (wave 1 & wave 2), and to some extent with prioritization of patients (wave 1). However, experience of death at work and having contact with COVID-19 patients were not statistically reliable predictors of distress (Supplementary Table S3).

Discussion

The results of this study, capitalizing on a longitudinal sample of HCWs in the Czech Republic, revealed a dynamic pattern in the relationship between stressors related to the COVID-19 pandemic and psychological distress among HCWs over time. Across the three waves of data collection, the distribution of stressors and distress followed a similar trajectory: both were lowest at baseline in 2020, after the initial wave of the pandemic subsided, highest at the second data collection point in 2021, then decreased at the third data collection in 2022 towards the end of the pandemic. This fluctuation suggests that participants experienced a surge in stressors and distress, which later returned to levels from the beggining of the pandemic. The path model further elucidates these relationships, indicating that distress in subsequent waves was strongly predicted by distress in the previous wave, demonstrating a carry-over effect of psychological distress over time. Examining intra-individual differences, we found that stressors in the first two waves significantly predicted concurrent distress. Further, the stressors towards the end of the pandemic did not seem to influence concurrent distress. The impact of stressors on psychological distress was similar for men and women. Gender emerged as a significant predictor of distress, with men experiencing lower levels of distress in the first two waves, but this pattern was not sustained towards the end of the pandemic, again reflecting intra-individual rather than inter-individual variations within the sample. Neither age nor occupation had a significant impact on distress, highlighting that other factors may be more critical in understanding the psychological experiences of HCWs in this context.

Secondary analyses revealed several stressors of particular importance. Most importantly, experiences of discrimination, stigma, or violence were associated with distress and were the only factor with carry-over effects from 2021 to 2022. Stigma, discrimination, and violence have been previously found to have significant negative consequences for mental health and psychological distress13,14,15. A mediating element in the relationship between stigma and psychological distress may be social isolation and the use of maladaptive emotion regulation strategies such as rumination or suppression13. These elements were present during the pandemic, which was an unprecedented health threat not only for health professionals, with a lack of information, prevalent fear in society, and measures based on isolation and distance. Additionaly, prioritization of patients at the height of the pandemic in Czechia in 2021 predicted concurrent distress, and low trust in workplace predicted distress in 2021 and 2022. These findings align with international research from the HEROES study16, where trust in workplace‘s and the goverment’s ability to manage the pandemic emerged as a predictor of mental healh across 22 studied countries. The timing of the role of these two predictors also aligns with the pandemic trajectory and lived experience of HCWs, as resource scarcity was highest in 2021. Low trust contributes to high-stress working environments and has a negative effect on the mental health and wellbeing of employees17. Factors of low trust in the workplace include non-benevolent behaviors, low competence of leaders, dishonesty of colleagues, unsafe environment, lack of openness and nepotism18.

Our study uncovered several novel aspects of the trajectory of psychological distress, which can be understood within the framework of the “recovery” response to stressors. This model describes a gradual return to baseline distress levels following an initial increase in distress during heightened adversity19. Our findings align with a longitudinal study of a representative sample of the U.S. general population, which reported an increase in psychological distress following a positive COVID-19 test, with distress returning to baseline levels after a month20. That study also noted that psychological distress was more severe among individuals with more intense and prolonged COVID-19 related symptoms. The stressors among HCWs as well as in general population evolved over time in response to pandemic measures and the epidemiological situation, exerting temporary effects on psychological distress. This pattern is consistent with the “stress reaction model”21, which posits that work-related stressors impact psychological functioning only over the duration of exposure. Nevertheless, we observed a carry-over effect of psychological distress over time, perhaps due to the influence of pre-existing mental health conditions, which are among the most robust predictors of mental health problems during and after pandemics and disasters7,22. It is also plausible that changes in other factors, such as increase in the out-of-workplace stressors (e.g., child caring, home schooling) may have contributed to the observed carry-over effect.

Our study uniquely found that COVID-19 related stressors during the last wave of data collection in 2022, towards the end of the pandemic, did not predict concurrent psychological distress. By including autoregressive paths in our analysis, we accounted for the impact of earlier observed factors on later outcomes. In contrast to the lacking direct effects, our additional analysis did reveal significant total effects in linear regression models. This suggests that, towards the end of the pandemic, it was primarily the same individuals whose distress increased due to the stressors, rather than the stressors causing new individuals to experience heightened distress. Therefore, these findings suggest that identifying individuals at hightened risk earlier on and working with them in interventions could have lasting impact on their mental health.

We identified gender differences in the descriptive part of our study, but the lack of effect modification suggests that gender did not significantly alter the association between pandemic-related stressors and psychological distress. Previous research, including a systematic review, has shown that female HCWs are more likely to experience psychological distress during infectious disease outbreaks, including COVID-198. Our findings from the first two waves, showing lower psychological distress among men, align with this. There are several possible reasons for the higher rates of distress among women during the pandemic. Women are generally more likely to report mental health issues, a trend that may have been amplified by the pandemic. A systematic review and meta-analysis of 134 studies found that mental health declines during the pandemic were more pronounced among women23. Additionally, women faced increased domestic responsibilities, such as childcare, especially during school closures24,25. In the Czech Republic, where schools were closed for an extended period, these added burdens may have contributed to higher distress among female HCWs. Interestingly, in our study, women did not report a higher distribution of COVID-19 related occupational stressors, contrasting with other studies that found that female HCWs experienced some work-related stressors to a higher degree than men. For example, the global HEROES study showed that women were less likely to receive support from colleagues and more likely to perceive personal protective equipment as insufficient - both factors linked to higher distress26. By the time of our final data collection in 2022, many of the pandemic’s pressures had eased: schools were open, vaccinations were widespread, and COVID-19 patient management was more established. These changes may have reduced gender-specific stressors. The absence of significant gender differences in distress during the later stages of the pandemic in the path model likely reflects intra-individual rather than inter-individual variations in these associations. This means that the lack of direct effects in the path model, in contrast to the total effects from linear regression, suggests that the same women who experienced higher distress compared to men in earlier waves continued to do so, rather than new women experiencing increased distress relative to men. Additionally, participants´ characteristics, such as temperament, resilience, personality or attachment style27,28,29, may have played a role beyond gender and gender-associated factors. For instance, individuals with higher resilience may adapt more effectively to prolonged stressors, potentially buffering against distress. Similarly, traits like emotional stability or flexibility could play a critical role in modulating stress responses over time. Future research should consider integrating these and other psychological factors to better understand the interplay between gender and intraindividual variations during prolonged crises.

Our study found no evidence that age or profession affect the psychological distress of HCWs. The literature on this topic presents mixed results. A systematic review of 139 studies8 found that 72 of them examined age as a predictor of psychological distress. In 39 studies, younger age was identified as a significant risk factor, while eight studies found that older HCWs were at greater risk, and the remaining studies reported no significant association between age and distress. Regarding the type of profession, in the systematic review by Sirois et al.8, 34 studies examined occupational roles, with 18 studies reporting that nurses experienced higher stress levels, while 16 found no difference, and five indicated that physicians were at greater risk.

The study has several strengths, including the examination of multiple COVID-19 related stressors, the inclusion of diverse healthcare professions in the sample, and the representation of all regions of the Czech Republic through the original HEROES study. Additionally, the longitudinal design, with data collected across three waves from 2020 to 2022, allows for the tracking of psychological distress over time, providing valuable insights into the evolving impact of the pandemic. However, there are limitations, notably a large dropout rate, which reduced the sample size and limited the ability to examine individual factors in detail. A significant limitation of this study lies in the attrition rate, with less than 20% of the original sample providing complete data across all three waves. We found that the participants who remained in the study had a slightly larger representation of physicians and slightly lower levels of distress when compared to those who dropped out. It is also possible that these two groups systematically differed in other unmeasured characteristicsl. Consequently, the generalizability of our findings to the full study cohort may be limited.

Another limitation of our study is the relatively small number of participants within certain subgroups, such as specific professional roles, gender categories, and socioeconomic levels. This constraint limits the statistical power and precision of our estimates for these subgroups, potentially masking meaningful differences or interactions. As a result, the generalizability of findings to these subgroups is restricted, and interpretations should be made with caution. Future research should aim to recruit larger, more diverse samples that allow for robust subgroup analyses and ensure adequate representation across professional, gender, and socioeconomic categories. Furthermore, the study did not account for stressors outside of work or personal factors that could contribute to heightened risk for psychological distress, potentially overlooking key influences on the mental health of HCWs.

The implications of our findings are important for the development of mental health support strategies for HCWs during pandemics and other crises. The dynamic nature of psychological distress observed in our study suggests that interventions should be adaptable, with heightened support during peak stress periods and ongoing resources as distress levels taper off. The early gender differences in distress indicate that successful programs should consider gender-specific challenges that go beyond occupational stressors. Additionally, the lack of significant effects of age and occupation on distress suggests that mental health interventions should be broad-based, addressing a wide range of HCWs rather than focusing solely on specific groups. Moreover, the carry-over effect of distress over time underscores the importance of long-term mental health monitoring and support, even after the immediate crisis has subsided and providing support to those in whom distress levels rise. These insights can inform more effective mental health policies and practices, ultimately improving the well-being and resilience of HCWs in future public health emergencies.

While our study identified key predictors of distress among HCWs, future research could benefit from longitudinal monitoring of specific factors to better understand stress dynamics over time. Factors such as resilience, coping strategies, emotional regulation, and work-life balance could provide deeper insights into stress adaptation processes. Additionally, organizational-level variables, including staffing adequacy, workplace support, and access to mental health resources, merit continuous evaluation to assess their impact on stress levels. By examining these elements in tandem, future studies could develop more targeted interventions to mitigate stress and promote well-being among HCWs, particularly during prolonged or recurring crises.