High levels of psychosocial distress among Australian frontline healthcare workers during the COVID-19 pandemic: a cross-sectional survey

Participants and study design

A nationwide, voluntary, anonymous, online survey was conducted between 27 August and 23 October 2020, concurrently with the second wave of the pandemic. Australian HCWs, comprising medical, nursing, allied health, medical laboratory, administrative and other support staff, who self-identified as frontline HCWs in secondary or primary and community care, were invited to participate. Participants did not need to have cared for people with COVID-19 to participate. Over 8 weeks, 9518 survey responses were received, with complete data from 7846 (82.4%) participants reported in this article (figure 1).

Request permission

Participants were recruited through multiple strategies. Information regarding the survey was emailed to chief executive officers and departmental directors of frontline areas (emergency medicine, critical care, respiratory medicine, general medicine, infectious diseases, palliative care and hospital aged care) of all public hospitals throughout Victoria, and to multiple hospitals around Australia. Hospital leaders were asked to share the survey information with colleagues. Thirty-six professional societies, colleges, universities, associations and government health department staff also disseminated the information about the survey across Australia. In addition, the study was promoted through 117 newspapers, 8 television and radio news items and 30 social media sites.

Data collection

Each participant completed the survey once, with no longitudinal data collected. Participants completed the online survey either directly or via a purpose-built website (https://covid-19-frontline.com.au/). Before commencing the survey, participants provided online consent to participate. Data were collected and managed using REDCap electronic data capture tools.26

Information collected included demographics, home life details, professional background, work arrangements, the impact of the pandemic on employment and finances, organisational leadership, workplace change, exposure to COVID-19 and health and recreational habits (online supplemental file 1). Most questions were in a single-choice or multiple-choice format, with free text questions enabling more detailed answers. Five validated psychological measurement tools were completed to assess symptoms of mental illnesses: anxiety (Generalized Anxiety Disorder Scale-7 (GAD-7)),27 depression (Patient Health Questionnaire-9 (PHQ-9)),28 post-traumatic stress disorder (PTSD) (abbreviated Impact of Events Scale-6 (IES-6))29 and burnout (abbreviated Maslach Burnout Inventory (MBI),30 with subdomains of emotional exhaustion (EE), depersonalisation (DP) and personal accomplishment (PA)). Resilience was assessed using the abbreviated two-item Connor-Davidson Resilience Scale-2.31 Burnout on the MBI is indicated by higher scores on the EE and DP, and lower scores on the scale of PA. Cut-off scores for validated scales were as follows: depersonalisation: 0 to 3=low, 4 to 6=moderate, 7 to 18=high; emotional exhaustion: 0 to 6=low, 7 to 10=moderate, 11 to 18=high; personal accomplishment: 0 to 12=high, 13 to 14=moderate, 15 to 18=low 32; IES is categorised as 0 to 9=none/minimal and >9=moderate-severe29; GAD-7: 0 to 4=none/minimal, 5 to 9=mild, 10 to 14=moderate, 15 to 21=severe anxiety27; PHQ-9: 0 to 4=none/minimal, 5 to 9=mild, 10 to 14=moderate, 15 to 19=moderately severe, 20 to 27=severe.28 In addition, participants were asked to report if they subjectively believed they had experienced anxiety, depression, PTSD, burnout or other mental health issues in order to determine their insight into their mental health. Ethics approval was provided by the Royal Melbourne Hospital Human Research Ethics Committee (HREC/67074/MH-2020).

Statistical methods and data analysis

A power calculation for general linear models was computed using RStudio.33 With an expected medium to large effect size and a power of 0.95, and significance level of 0.05, a sample of 6348 participants was required. To account for missing or incomplete data, a minimum sample size of 7000 responses was chosen. Data analysis was performed using SPSS V.26.0 statistical software (IBM). Demographic and socioeconomic characteristics were reported descriptively. Predictors of mental illness symptoms were identified through univariable logistic regression then entered into a multivariable logistic regression model. Covariates examined in univariable analyses included age; gender; state; occupation; number of working years since graduation; living situation (living alone, living with children, living with elderly); frontline area; practice location; working with patients with COVID-19; anticipating working with patients with COVID-19; having received personal protective equipment (PPE) training; worry that their role will lead to COVID-19 transmission to family; worry regarding being blamed by colleagues, close friends or relatives infected with COVID-19; changed relationships with partner or friends or family or colleagues; changed household income; concerns regarding household income and pre-existing mental health diagnoses. For each mental illness scale, outcomes were merged into dichotomous categories (no or minimal symptoms vs moderate to severe symptoms) in the regression model. Associations between mental illness symptoms and predictor variables are presented as ORs with 95% CIs. Multicollinearity of predictor variables was examined using the variance inflation factor criterion. The Spearman coefficient (r) was calculated to evaluate the correlation between self-reported and objective evidence of mental illness symptoms. For all statistical tests, significance was indicated by p≤0.05.

Demographic characteristics and workplace environment

More than half (n=4110, 52.4%) of the participants were younger than 40 years, and 6344 (80.9%) were women (table 1). Most participants were nurses (n=3088, 39.4%), doctors (n=2436, 31.0%) or allied health staff (n=1314, 16.7%) with 523 participants working in other health organisation roles including food services and security. The medical staff group comprised 389 general practitioners, 1221 senior medical staff, 745 junior medical staff and 81 students. More than one-quarter of participants (n=2250, 28.7%) had caring responsibilities at home, and 2133 (27.2%) participants had children who were being homeschooled.

Participants worked in primary care or community roles (n=1250, 15.9%), medical specialty areas (n=1205, 15.4%), emergency departments (n=1146, 14.6%), anaesthetics or surgical areas (n=824, 10.5%) or intensive care units (n=745, 9.5%) (table 2). More than three-quarters (n=6158, 78.5%) had been tested for COVID-19, 136 (1.7%) had been infected with COVID-19 and 77 (0.9%) had been previously quarantined because of unprotected exposure to COVID-19. Three-quarters (n=4551, 76.4%) were worried or very worried that their role could lead to them transmitting COVID-19 to their families, and almost two-thirds (n=4949, 63.1%) were worried about being blamed by colleagues for not taking adequate precautions if they contracted COVID-19.

Relationship changes and prevalence of mental illness symptoms

More than three-quarters of participants (n=5994, 76.4%) reported that the pandemic had affected their relationships with family, friends and colleagues, and nearly one-third had a close friend or relative who had been infected with COVID-19 either in Australia or overseas (table 2). Approximately one-third (n=2389, 30.4%) reported having a pre-existing mental illness diagnosed before the pandemic (table 3). Many participants subjectively believed they had experienced mental illness during the pandemic including anxiety (n=4875, 62.1%), burnout (n=4575, 58.3%) and depression (n=2175, 27.7%). Mental illness symptoms measured by objective scales demonstrated a similar or worse trend, with 4694 (59.8%) participants experiencing mild to severe anxiety, 5458 (70.9%) moderate to severe burnout (EE) and 4495 (57.3%) mild to severe depression. Participants had a high score for resilience with a mean (SD) of 3.21 (0.66) out of 4. There was correlation between subjective reporting and objective evidence of moderate to severe mental illness symptoms for anxiety (r=0.346, p<0.001), depression (r=0.346, p<0.001) and EE (r=0.354, p<0.001).

Predictors of poor mental health

In the multivariable regression model, independent, personal predictors for worse mental health on all measured outcomes (anxiety, depression, burnout and PTSD) included female gender, younger age, experiencing worsening of personal relationships and low resilience scores (table 4). In addition, independent, personal predictors for anxiety and PTSD included having previous mental health conditions, having a family member or friend infected with COVID-19 and concerns about household income. Depression was also associated with having previous mental health conditions and concerns about household income, whereas EE was also associated with previous mental health conditions. Independent, workplace predictors for worse mental health outcomes on all measured scales (anxiety, depression, burnout and PTSD) included having a nursing, allied health or other non-medical role, frontline area, working with patients infected with COVID-19 and being worried about being blamed by colleagues on contracting COVID-19 infection (table 5). There were no significant associations between other demographic, work environment, relationship or financial covariates and each mental illness score.

Main findings

To our knowledge, this is the largest, national, cross-sectional study examining psychosocial distress during the COVID-19 pandemic in Australia that has included all frontline healthcare occupations and areas. Despite participants receiving high scores on the validated resilience instrument, the majority experienced anxiety or depressive symptoms, or EE (burnout). This indicates that the protective effects of resilience are not sufficient to prevent psychological harm during the pandemic. A significant proportion also experienced PTSD symptoms. Although less than half of the participants worked with patients with COVID-19 and very few had been infected with COVID-19 or quarantined, many experienced disruptions to family life, altered social relationships and financial worries. Our findings are consistent with those reported in international studies: high mental health burden on frontline workers during COVID-1910 11 15 and SARS pandemics.34 Fears of transmitting COVID-19 infection to family and of being blamed by colleagues for not taking adequate precautions if they did contract COVID-19 were extremely common. Personal, social and workplace predictors for mental illness symptoms have been identified.

Around the world, a growing number of largely country-specific, single timepoint, cross-sectional surveys have identified that mental health problems are common in HCWs during the COVID-19 pandemic. Prevalence estimates are as follows: 33% to 59% for anxiety, 30% to 62% for depression, 41% to 51% for burnout and approximately 57% for acute distress or PTSD.10 11 13–15 35 The upper limits of these prevalence estimates are strikingly similar to our own findings. However, moderate to severe burnout (EE) was much more prevalent in our study (70.9%), which may be explained by the later timing of our study, by which time Australian HCWs had endured many months of social and workplace disruptions, and lockdown restrictions.

By contrast, two separate, small (n=320 and n=668), single-site, single timepoint surveys of HCWs undertaken in Melbourne from April to May 2020 and from May to June 2020 both identified a lower prevalence of adverse mental health outcomes.36 37 Their findings may again be partly explained by the earlier timing of the studies in the first wave and the lack of power in those studies due to smaller size of the samples. Comparing our data to international data, the high prevalence of symptoms of poor mental health in our study is interesting given the comparatively low case load of COVID-19 in Australia. One explanation is that anticipation and fear of a catastrophic crisis leading to high death rates of patients and HCWs (as Australian HCWs saw occurring overseas) contributed to adverse psychological outcomes.14 This concept of psychological distress being related to anticipated, perceived risk is important and highlights the critical importance of crisis preparedness, good government and organisational leadership and consistent clear communication. In addition, the pervasive media coverage regarding COVID-19 along with the many restrictions enacted in local lockdowns may have contributed to poor mental health in Australian frontline workers.

Similar to our findings, studies from overseas have found that predictors of poor mental health in HCWs during the pandemic include female gender, less years of work experience (which in our study correlated with younger age), pre-existing psychological illnesses, working in a nursing role and working in certain frontline areas.10–13 15 16 35 38 39 Many of these groups are at heightened risk of psychosocial harm during non-pandemic times, and it is possible that crises such as COVID-19 exacerbate harm in pre-existing vulnerable groups.40 Importantly, unlike previous small local and international studies, the large sample size in our study enabled us to demonstrate that female gender and working in nursing or allied health roles are independent predictors of poor mental health. The relationship between nursing and poorer mental health may be explained by the heightened risk of COVID-19 exposure from prolonged and frequent contact with patients. Moreover, nursing and allied health professionals generally have less choice regarding their daily work environments.11–13 16 Reduced finances were not associated with a nursing role and therefore did not explain the association.

The relationship between gender and adverse mental illness outcomes is intriguing, given that this relationship was identified even during the SARS pandemic.34 One possible explanation is that men and women have different coping styles,41 with men having greater odds of reporting DP in this study. In addition, a British study identified that women have had to bear greater responsibilities (on average, an extra 11.2 hours of unpaid work per week) than men as primary carers for dependents during the pandemic.42 General population data from the Australian Bureau of Statistics report similar findings, with women three times more likely than men to perform the majority of caregiving tasks and twice as likely to undertake the majority of unpaid domestic work.43 In our study, having young or old dependents was not a predictor of poor mental health. However, we did not specifically enquire about the number of additional unpaid hours undertaken in the home for domestic or caregiving tasks during the pandemic. As there was no difference in resilience scores between men and women, this gender difference requires further exploration. The lack of a relationship between PPE training and poor mental health in our study may relate to the majority of frontline staff receiving training and the relatively low rates of COVID-19 infection in Australia compared with other countries.

Limitations

The large sample size in our study enabled detailed examination of independent predictors of poor mental health. Most participants in our study were women, which is consistent with data from both the Australian Institute of Health and Welfare and the Australian Health Practitioner Regulation Agency demonstrating that 75% of the Australian health workforce is female.44–46 Because of the very broad survey dissemination strategy, calculation of a response rate was not possible. Selection bias and response bias may have led to overestimation or underestimation of psychological distress and rates of pre-existing mental health illness. Similarly, in line with other international surveys exploring the psychosocial effects of the COVID-19 pandemic on healthcare workers, we were not able to confirm clinical diagnoses of mental illness with the symptoms measured by the validated psychological scales. Nevertheless, these scales are validated and the only feasible option for measuring mental health symptoms in a large-scale survey such as this.

Because of the spontaneous and unexpected nature of the COVID-19 pandemic, no baseline data regarding mental health symptoms in non-pandemic times had previously been collected from a large cohort of Australian HCWs. Therefore, it is not possible to demonstrate a change in the prevalence estimates of mental health symptoms in this study. Nevertheless, the prevalence estimates in this study are much higher than those reported in earlier studies in non-pandemic times.2 47–49 Notably, the case load of COVID-19 in Australia at the time of survey closure was low relative to international settings, with 27 484 cases recorded.50 The prevalence of mental health impacts arising in the Australian context is indicative of harm related to the prolonged stress of a pandemic, even with relatively few cases. Participant responses were measured at a single timepoint, not longitudinally, to avoid excessively burdening the frontline healthcare workers during the pandemic. However, given the ongoing nature of the pandemic, we believe that longitudinal research is urgently required to better understand any persisting psychosocial effects of the pandemic on HCWs and any ramifications for patient safety and workforce retention. Similar prospective studies sampling Italian HCWs during the first and second waves of COVID-19 have reported growing prevalence of mental health issues as the pandemic continues, and it is likely that similar trends exist in Australia.51 Furthermore, research is required to examine the acceptability, uptake and effectiveness of any new interventions introduced to support the well-being of HCWs.

Implications

Although many factors, including lockdown restrictions, social disconnection and media coverage, likely have contributed to the high prevalence of mental health symptoms in frontline healthcare workers in this study, occupational factors cannot be ignored. Indeed, occupational factors (related to workloads, training, PPE, organisational leadership, communication and policies) must be actively considered because they represent important opportunities to intervene and prevent mental health issues. Both better crisis preparedness and new psychological support services for HCWs are needed. Importantly, such services should not just be short-term ‘fixes’ to address the current pandemic-related issues, but instead should provide long-term support given the high prevalence of pre-existing mental health diagnoses. These supports must be accessible and acceptable to HCWs. Although resilience was identified as a protective factor in this study, the overall resilience level of HCWs was already high, and as such, approaches that aim to build resilience are likely to have limited efficacy in this cohort. Furthermore, it is vital that health leaders in the government, secondary care and the community recognise that certain groups of HCWs are more vulnerable to mental health problems and therefore require additional targeted support interventions. Crucially important are organisational policies and practices that address burnout (and contributing factors such as information overload), given its extremely high prevalence and the risk it poses to workforce retention.7

The health workforce is an indispensable asset. Yet crises such as the COVID-19 pandemic are associated with significant mental health symptoms in frontline HCWs, with potentially wide repercussions for individuals, patients and the workforce. Crisis preparedness, along with long-term, evidence-based policies and practices that focus on preventing and actively addressing psychological well-being, is needed to protect, maintain and ‘future-proof’ the health workforce.