Mental Health Status of Healthcare Workers During the COVID-19 Outbreak: An International Study

Context and Objectives

The COVID-19 pandemic has represented a massive global health crisis that exerts enormous physical and psychological pressure on entire populations. Healthcare workers (HCWs) face particularly challenging circumstances, including increased patient numbers, construction of COVID-19 units, shortage of intensive care beds and ventilators, and scarcity of protection equipment. Previous studies have documented that mental health of the general population was compromised during the pandemic, with high prevalence of psychological stress, depression, boredom, insomnia, and poor sleep quality. However, research on mental health in HCWs during COVID-19 has been limited mainly to national contexts and Asian countries. This study aimed to determine mental health outcomes in HCWs related to COVID-19, compare them with the general population, and identify predictive factors of psychological impact during the first pandemic wave in an international context.

Method

Participants and Design The study is part of the "COVID-19 IMPACT Study," a larger international research project that employed an online survey. Random sampling followed by snowball sampling was used to collect data over a two-month period (between July 7 and June 7, 2020). The total sample of the COVID-19 IMPACT project included 9,563 participants from 78 countries/regions. For this specific study, a sample of 1,856 HCWs from 45 countries was used. Inclusion criteria required minimum age of 18 years and ability to understand one of 18 languages offered in the survey (English, Greek, German, French, Spanish, Turkish, Dutch, Latvian, Italian, Portuguese, Finnish, Slovenian, Polish, Romanian, Hong Kong Chinese, Hungarian, Montenegrin, and Persian). HCWs who reported not being healthcare workers were excluded from comparisons between HCWs and general population. Countries with the largest samples were Italy (n = 236), Austria (n = 210), Latvia (n = 122), Cyprus (n = 121), Switzerland (n = 120), France (n = 89), Portugal (n = 80), Hungary (n = 63), Hong Kong (n = 60), and Germany (n = 54). The majority of HCWs came from European countries (n = 1,325, 85%).

Intervention/Conditions A 20-minute self-administered online survey was employed via a secure Google platform. Recruitment was through multiple channels: local press (newspapers, newsletters, and radio), social media (Facebook, Instagram, Twitter, WeChat), professional platforms, local hospitals, health centers, email lists of HCW professional groups, and university institutions. Ethics approval was obtained from the National Cyprus Bioethics Committee and approvals from different research teams involved in data collection.

Measurement Instruments

Predictors: Sociodemographic information was collected including age, sex, marital status, having children, living situation, employment status, and educational background.

Perceived Stress: Measured with the Perceived Stress Scale (PSS) by Cohen et al. (1983), the most widely used instrument. It includes 10 items assessing how people perceive stressful situations during the last month, using a Likert scale from 0 (never) to 4 (very often). Raw scores range from 0 to 40. Categorization: 0-13 = low perceived stress, 14-26 = moderate perceived stress, 27-40 = high perceived stress. Cronbach's alpha = 0.89.

Depressive Symptoms: Evaluated using two items from the disengagement subscale of the Multidimensional State Boredom Scale (MSBS) by Fahlman et al. (2011). This subscale was selected because it was considered relevant to the unique situation of lockdown. The items measured: (a) willingness to do pleasant things but without enjoying them (boredom) and (b) person's time wasting. Scored on a 4-point Likert scale, from 1 = very little to 4 = extremely, with higher scores indicating greater depressive symptoms. The MSBS disengagement subscale showed significant correlation with measures of neuroticism trait, depression, and Borewell Promises Scale (r = 0.61), and with Center for Epidemiological Studies Depression Scale (r = 0.66).

Sleep Changes: One structured, closed question assessed sleep changes, asking how sleep had changed since lockdown, with options: I sleep more, I sleep less, and I sleep about the same.

Perceived Social Support: Measured using the Oslo Social Support Scale (OSS) by Berkman & Kawachi (2000), consisting of three items assessing the number of people an individual can count on for important personal problems, interest of others in what happens, and difficulty in obtaining help from neighbors. Scored on a two-point Likert scale, with three levels of social support: low (3-8), moderate (9-11), and high (12-14). Cronbach's alpha for OSS = 0.54.

Statistical Analysis Statistical analyses were performed using SPSS version 26. Data followed normal distribution. Descriptive statistics (frequency, percentage, mean, standard deviation) were used to describe sociodemographic data, perceived stress, depressive symptoms, perceived social support, and sleep changes. All predictor variables of sociodemographic data and social support were treated as categorical variables, except age treated as continuous variable. Outcome variables of perceived stress and depressive symptoms were treated as continuous variables, and sleep changes as categorical variable. Univariate analyses (ANOVAs) were performed for categorical predictors and linear regression for numerical predictor (age). Games-Howell post-hoc analysis or independent-sample t-tests were used for numerical variables, and chi-square tests for categorical variables to examine differences between HCWs and general population. Multivariable associations between each mental health outcome variable of HCWs and all predictors were investigated through multiple linear regression for variables of perceived stress and depressive symptoms, and ordinal logistic regression for the sleep changes variable. The strength of associations between predictor and dependent variables was assessed by computing standardized regression coefficients and Cramer's V. Values for standardized effects were: small from 0.05 to 0.10, medium 0.20 to 0.30, large 0.30 to 0.40, very large greater than 0.40. Statistical significance was defined as p < 0.05.

Results

Descriptive Data - Sociodemographic Characteristics of Predictors The mean age of HCWs was 40.52 years (SD = 11.57). The majority were female (n = 1,307, 84%), working full-time (n = 1,540, 82.8%), with undergraduate education (n = 775, 49.8%), married (n = 728, 46.8%), living with their own family (n = 1,092, 70.2%), and having children (n = 818, 52.6%). Sociodemographic characteristics of HCWs differed significantly from general population by sex, where the percentage of women was higher among HCWs than in general population (ES = 0.67). In particular, the mean age of general population was 36.12 years (SD = 13.44). The majority were female (n = 5,980, 76.3%), working full-time (n = 3,873, 50.8%), with undergraduate education (n = 2,453, 31.4%), married (n = 2,641, 33.8%), living with their own family (n = 3,966, 50.7%), and not having children (n = 4,820, 61.6%).

Perceived Social Support The mean score of perceived social support among HCWs was 10.39 (SD = 2.25), indicating that HCWs had moderate level of perceived social support. Approximately half of HCWs (n = 813, 52.2%) had moderate level of perceived social support, one-quarter (n = 478, 30.7%) had high level, and less than one-fifth (n = 265, 17%) had low level of perceived social support. Participants from general population sample demonstrated low levels of perceived social support (25.9%), and also more participants had high level of perceived social support (21.8%, ES = 0.1) than among HCWs.

Perceived Stress Most HCWs (n = 887, 57%) demonstrated moderate levels of perceived stress, some HCWs (n = 562, 36.1%) showed low levels of perceived stress, and the remaining HCWs (n = 107, 6.9%) showed high levels of perceived stress. The general population reported higher levels of perceived stress than HCWs, with a statistically significant mean difference (95% CI [-1.58, -0.77]).

Depressive Symptoms HCWs had moderate symptoms of depression (M = 6.14, SD = 2.12). Their average depressive symptom score was lower than the general population, with a statistically significant mean difference (95% CI [-0.65, -0.41]).

Sleep Changes During lockdown or self-isolation, half of HCWs (n = 800, 51.4%) continued to sleep about the same, while 383 (24.6%) slept more and 373 (24.0%) slept less. Regarding sleep quality, 925 HCWs (59.4%) reported having fairly good or very good sleep, 340 (21.8%) neither good nor bad, and 291 (18.7%) fairly poor or very poor. Similar patterns were observed with reported sleep changes in general population (p > 0.1). In particular, approximately one-third continued to sleep about the same (n = 3,594, 46.0%), one-third slept more (n = 2,554, 32.7%), and one-fifth slept less (n = 1,671, 21.4%). Regarding sleep quality, almost none reported their sleep quality as fairly good or very good (n = 4,298, 54.9%), 26.3% (n = 2,055) reported neither poor nor good, and minority indicated that their sleep quality was fairly poor or very poor (18.7%, n = 1,464).

Bivariate Analyses

Perceived Stress: The results of bivariate analyses with perceived stress as outcome showed that predictors associated with higher perceived stress among HCWs included: female sex (95% CI [-2.67, -0.80]), not having children (95% CI [-1.83, -0.46]), and living with parents (95% CI [0.63, 3.06]). An additional predictor of higher perceived stress was lower level of perceived social support. In particular, HCWs with perceived social support low had significantly higher perceived stress than HCWs with low level of perceived social support, Welch's F(2, 661.639) = 42.404, p < 0.00. There was no other statistically significant association between PSS score and sociodemographic characteristics.

Depressive Symptoms: The results of bivariate analyses with depressive symptoms as outcome are available in the article (https://doi.org/10.23668/psycharchives.5071). One predictor was lower education (95% CI [3.43, 7.18]), living with parents (95% CI [7.4, 1.548] = 10.836, p < 0.01), not having children (-0.53, 95% CI [-1.074, -0.32], p < 0.01), and lower level of perceived social support. It was predicted that these would be predictors of higher levels of depressive symptoms among HCWs. No other sociodemographic characteristics appeared to have a statistically significant association with prevalence of depressive symptoms.

Sleep Changes: The results of bivariate analyses with sleep changes as outcome are available at https://doi.org/10.23668/psycharchives.5071. There was a statistically significant association between sleep changes and age (p < 0.01), marital status (χ²(10) = 20.123, p = 0.028), having children (χ²(2) = 27.555, p < 0.001), and living situation (χ²(8) = 16.721, p = 0.033). In particular, across marital status widowed, widowed HCWs (33.3%) appeared to sleep more hours. At older age, older HCWs (M = 42.24, SD = 9-11.90), not working at that moment (29.3%), having children (25.1%), and living with parents (28.7%) were sleeping fewer hours as well. Educational status studies did not demonstrate a statistically significant association with sleep changes.

Multivariate Analyses

Perceived Stress: The results of multivariate analyses for perceived stress outcome are available in Table 2. The strongest predictor of perceived stress was low perceived social support (95% CI [-4.65, -3.65]) and moderate (95% CI [-2.92, -1.84, -2.00]) of perceived social support proved protective against perceived stress. Other protective factors of perceived stress were male sex (95% CI [-1.81, -2.73, -0.90]), younger age (95% CI [-0.12, -0.16, -0.09]), living with own family (95% CI [-1.68, -2.60, -0.16]), and having children (95% CI [1.10, 0.12, 2.73]). Statistically significant differences were found in results when countries with more than 50 participants (n = 10) were added to the analysis as covariate (https://doi.org/10.23668/psycharchives.5071). The hierarchical multiple regression analysis indicated that the full model of sociodemographic characteristics and social support was statistically significant in predicting perceived stress R² = 0.10, F(2, 1.532) = 7.867, p < 0.01 (see https://doi.org/10.23668/psycharchives.5071). The effect of social support in predicting perceived stress led to a statistically significant increase in R² of 0.089. The addition of social support to prediction of perceived stress led to a statistically significant increase in R² of 0.067, F(2, 1.532) = 58.443, p < 0.001.

Depressive Symptoms: The results of multivariate analyses for depressive symptoms outcome can be seen in Table 2. Predictors associated with lower depressive symptoms were lower education (95% CI [-0.03, -0.03]) and living with friends or roommates (-0.04, 95% CI [-0.62, -0.55]). Perceived social support was also protective against depressive symptoms with high (95% CI [-1.68, 95% CI [-1.99, -1.38]]) and moderate levels (95% CI [-0.96, 95% CI [-1.24, -0.68]]). No statistically significant differences were observed in results when countries with more than 50 participants (n = 10) were added to the analysis as covariate (https://doi.org/10.23668/psycharchives.5071). According to hierarchical multiple regression analysis, the full model of sociodemographic characteristics and social support was statistically significant in predicting depressive symptoms R² = 0.12, F(2, 1.532) = 9.992, p < 0.01, adjusted R² = 0.113. The addition of social support to prediction of depressive symptoms led to a statistically significant increase in R² of 0.067, F(2, 1.532) = 58.443, p < 0.001.

Sleep Changes: The results of multivariate analyses for sleep changes outcome can be seen at https://doi.org/10.23668/psycharchives.5071. The positive predictor of sleep changes was working full-time (0.46, 95% CI [0.107, 0.85]), and the negative predictor was having children (-0.43, 95% CI [-0.70, -0.17]). Additionally, the effect of depression was controlled in the analysis, and no significant changes were observed.

Discussion and Conclusions

The COVID-19 pandemic represents a public health emergency of international concern that poses an enormous challenge to both HCWs and the general population. In line with the hypothesis and previous studies, this study found that, at the height of the first lockdown, almost half of HCWs demonstrated moderate levels of perceived stress, and 7% showed high levels. Participants also reported moderate depressive symptoms with the majority continuing to sleep about the same, having fairly good or very good sleep. The strongest perceived stress was perceived social support, with low and moderate levels being a protective factor. Other protective factors of perceived stress were male sex, younger age, not having children, living with own family, and having children. HCWs who were widowed, older, not currently working, having children, and living with parents reported sleeping fewer hours.

These results suggest that there are subgroups among HCWs suffering psychological problems and may be at risk for developing future mental health difficulties. The current findings corroborate findings from previous studies, demonstrating the need to improve mental health of HCWs. However, they do not reveal more serious mental health problems being observed especially in China where the initial COVID-19 outbreak emerged. It appears that Chinese HCWs reported worse mental health being first suddenly affected with an unknown and highly contagious virus. Healthcare systems and HCWs had to struggle with the rapid spread of Coronavirus and an unprecedented number of deaths, as healthcare systems do not appear to have been prepared for this pandemic. The rapid increase in infected patients increased an enormous workload and psychological problem of medical personnel. Some studies examining frontline healthcare personnel who worked directly with infected COVID-19 patients found higher symptoms of depression, anxiety, somatization, and insomnia than non-frontline personnel or general population. Such differences in psychological outcomes suggest variations in outcomes among different subgroups of HCWs and among different countries in association with individual differences and other coping factors.

HCWs, especially those at the frontline, are susceptible to mental and physical exhaustion and sleep disturbances. This was not surprising, as HCWs tend to work under stressful situations that probably make them generally more resilient to stress, so they may have been better adapted to stressful situations. The present finding that half of respondents continued to sleep the same in one of their reported sleep disturbances was consistent with a previous report. HCWs presenting with sleep problems were those who also reported higher levels of perceived stress and depressive symptoms. Sociodemographic characteristics such as being widowed, older, not working at that time, having children, and living with parents predicted sleeping fewer hours.

In line with our hypothesis and previous studies, HCWs presented lower perceived stress and depressive symptoms than the general population. This is not surprising, as HCWs tend to work under stressful situations that probably make them more resistant to stress, so they may have been better adapted to stressful situations. HCWs as a group tend to have higher education, possibly being a protective factor against stress and depression. Perceived social support emerged as a strong protective mental health factor in this study. HCWs reported significantly greater perceived social support than the general population, which probably means better means of protection against mental health problems, rather than HCWs being better than the general population in handling stressful situations. HCWs potentially received support from their social environment since those who lived with friends or roommates reported fewer depressive symptoms and those who lived with family and children showed lower stress and sleep disturbances. A previous study found that HCWs rely extensively on social support and contact with family and friends as a way of coping with stress. Perceived social support was found to contribute to reductions in anxiety and stress with increased self-efficacy among HCWs.

Female HCWs who live with parents and children without working were found to present higher levels of perceived stress and depressive symptoms. As a group, female HCWs were reported in the literature to be more vulnerable to stress, anxiety, and depression than male HCWs. The present findings that half of respondents continued to sleep the same in one of their reported sleep disturbances was consistent with a previous report. HCWs presenting with sleep problems were those who also reported higher levels of perceived stress and depressive symptoms. Sociodemographic characteristics such as being widowed, older, not working at that time, having children, and living with parents predicted sleeping fewer hours.

The study had several limitations. First, sample composition was an important limitation of the present study. Although we obtained an international sample, we did not distinguish subgroups of HCWs in terms of types of workers (physicians, nurses, emergency HCWs, etc.), types of medical or nursing specialties, years of clinical experience, etc. Therefore, we were unable to assess how mental health outcomes differed between various subgroups of HCWs. Furthermore, in many countries, the sample was very small at less than five participants, which could bias the findings of the study. Second, the results are based on cross-sectional and correlational analysis, so causation cannot be inferred and any delayed impact of the pandemic and lockdown on mental health of HCWs was not captured. Third, we used online self-report questionnaires, which can be subject to retrospective response bias. Further research is required, applying qualitative analysis of richer data from different subgroups of HCWs in longitudinal studies with multiple time points to confirm or refute the results of the present survey. The current survey was conducted during the first wave of the pandemic and before pandemic fatigue may have set in. Longitudinal studies would be beneficial, especially if they examine long-term fluctuations in stress and depressive symptoms among HCWs. Finally, country-specific measures of case incidence and lockdown measures may vary across countries, all of which may bias the results of the study and should be considered in future research for better understanding of the impact of the COVID-19 pandemic on the health status of HCWs.

Importance and Contribution

This international research study documents the mental health outcomes and sleep quality changes among a broad sample of HCWs during the global COVID-19 pandemic crisis. It is important that the public health system, especially where there are low levels of perceived social support, lower education, no children, and residence with parents were predictors of higher levels of perceived stress and depressive symptoms among HCWs. The present findings highlight the need for implementation of appropriate supportive measures and early treatment of COVID-19-related mental health problems in HCWs. As the pandemic continues, healthcare systems need to continue refining the psychological support system, providing early psychological interventions targeted to vulnerable groups both in HCWs and in the general population.

---