All employees (n≈29,900) were invited to participate in the study. A total of 65,685 emails were sent during the recruitment period, including initial emails to all staff, follow-up emails to staff members who did not open the first email, and two emails to all managerial staff. Thirty five thousand six hundred and sixty five (54.3%) of the recruitment emails were opened, and the link to the consent form was clicked 9,088 times. Of the 9,088 who linked to the consent form, 8,494 individuals (93.5%) consented to participate, for an overall response rate of 28.4%.

REDCap data were exported to and analyzed with SPSS v. 26 (IBM; Armonk, NY 2019). Categorical comparisons were evaluated with a chi square test. Continuous data were evaluated for distribution and analyzed with one of the following, depending on number of groups and normality of distribution: Student’s t-test or Mann–Whitney U test for two groups, and analysis of variance or Kruskal–Wallis H test for >2 groups. Correlations were evaluated with a Spearman rank correlation coefficient. A forward, conditional logistic regression model was constructed to evaluate the strength of contribution of many of the variables with univariate differences on outcomes of at least moderate symptoms of anxiety, depression, insomnia, or PTSD. Odds ratios (OR) and 95% confidence intervals (CI) were calculated. All results yielding p < 0.05 were deemed statistically significant. Since the number of responses was not known initially, no a priori power analysis was performed.

Thirty-one percent of participants (n = 2,023) reported moderate to severe symptoms of anxiety (≥10 on the GAD-7), while 29% (n = 1,888) reported mild symptoms (GAD-7 score 5–9). Of note, 23% of the survey participants did not respond.

Eighty-three percent of participants (n = 5,285) endorsed moderate to severe depressive symptoms (≥10 on the PHQ-9), although 28% did not respond. Seventeen percent (n = 1,046) endorsed mild depression (PHQ-9 score 5–9). Regarding the PHQ-9 question regarding frequency of suicidal ideation, 4.5% (n = 279) of respondents answered “several days” or more.

About one eighth (n = 820; 13%) of valid surveys contained positive responses to the PC-PTSD scale (score of ≥4), although 25% did not respond. Half of participants (n = 2,749; 51%) reported moderate to severe insomnia (scored at least 15 on the ISI). However, 36% of the survey participants did not respond.

While only 54% of study participants responded to questions regarding alcohol use, of those who responded, 49% (n = 2,220) endorsed heavy drinking over the past year (scored ≥ 3F; ≥4M on the AUDIT-C).

Females reported experiencing more severe levels of anxiety (33 vs. 22% reported moderate to severe symptoms; p < 0.001), depression (85 vs. 74%; p < 0.001), PTSD (13 vs. 11%; p = 0.023), and insomnia (52 vs. 46%; p < 0.001) than men (Table 2). Individuals who identified as Hispanic reported greater anxiety (38 vs. 30% reported moderate to severe symptoms; p < 0.001), PTSD (16 vs. 13%; p = 0.01), and insomnia (54 vs. 51%; p = 0.026) than non-Hispanic individuals. African American individuals endorsed high levels of anxiety (36%) and insomnia (50%), while those who identified as White endorsed high levels of depression (84%; p < 0.001).

Individuals working in the ICU and on medical floors endorsed the highest severity of depression, anxiety, PTSD, and insomnia; ED, office workers, and those working in outpatient settings endorsed the lowest severity (Table 3). Nurses and clinical staff endorsed the greatest symptoms of depression and anxiety, while nurses also endorsed high levels of PTSD and insomnia.

Variables that were associated with our primary outcomes (e.g., depression, anxiety, PTSD, insomnia) via univariate analysis were entered into logistic regression models to evaluate the relative contribution of each factor to each outcome. Categorical variables were either dichotomous (age ≥ 60, gender, mental health professional, emergency medicine professional, presence of CDC-defined medical risk factors at time of survey, endorsement of psychiatric history, feeling supported at work, endorsing barriers to working, quarantine of at least a week, adequate access to personal protective equipment (PPE), and feeling the ability to say no to uncomfortable work demands) or multi-level and compared to a reference group. Significant contributors to each model are in Table 4.

History of mental illness was an important predictor for all four outcomes (all ORs 1.9–3.0), as was endorsement of real-world barriers to working (all ORs 1.4–1.7), such as concerns about childcare (20%), and feeling unable to say no to work demands (p < 0.001). Age also was important, with significant influence (p < 0.005) in all four outcomes; older people (i.e., those ≥60) had lower levels of all four outcomes compared with their younger (<60) counterparts. Gender played a role in all outcomes except for PTSD. Concerns about risk of becoming infected, or having a complicated illness course if infected, contributed to all models. Access to PPE had a significant role (p ≤ 0.001) in depression and anxiety.

When asked where HCWs turn to for emotional support during COVID-19, a majority of respondents reported that they rely on family (68%) and friends (51%), although many also depend on work leadership (22%) and work peers (42%). However, many fewer individuals reported turning to specialized work programs, such as EAP (employee assistance program) or support groups (3.4%), or private mental health providers (5.3%).

Females, non-Hispanic individuals, and individuals who identify as White endorsed the greatest amount of emotional support and were most likely to seek out mental health treatment. Individuals who identified as White also endorsed significantly greater use of specialized work programs (p = 0.001). Conversely, individuals who identified as Black (21%; p < 0.001) and/or Hispanic (11%; p = 0.032) were significantly more likely to turn to religious groups for support (compared to 8% of total sample that identified religious groups as a source of support). See Table 5 for additional information.

Different demographic factors were entered into a logistic regression to determine the relative contribution of each factor on receiving/seeking out support. Race had the biggest influence on whether an individual received (or sought) support, followed by gender and ethnicity. See Table 6.

Reports of moderate to severe anxiety symptoms (31%) was consistent with our national data (Young et al., 2021) and with a meta-analysis of surveys of the public across the world (31.9%) (Salari et al., 2020), although much lower than the 45% reported in Lai et al.’s study from China (Lai et al., 2020). Current rates of anxiety were much greater than the prevalence of generalized anxiety outside of the pandemic (3–18%) (Kessler et al., 2005; Harvard Medical School, 2007).

Eighty-three percent of participants in this study reported moderate to severe depressive symptoms, compared to 14% reported in our national survey (Young et al., 2021), 50% in China (Lai et al., 2020), and 33.7% in the world population (Salari et al., 2020). The rates of suicidal ideation (5%) endorsed in this study was consistent with our national data (Young et al., 2021). The degree of insomnia reported by our participants (51% reported moderate to severe symptoms) is significantly greater compared to the prevalence of insomnia outside of the pandemic (approximately 30% have some symptoms) (Bhaskar et al., 2016). Similarly, of our participants who responded to questions about alcohol use, 49% endorsed heavy drinking during this time. This is significantly higher than the 12.7% of United States population who meet criteria for alcohol use disorder prior to the pandemic (Grant et al., 2017). Lastly, the prevalence of PTSD symptoms reported in this study (13%) was similar to our national data (Young et al., 2021) and much lower than the prevalence of 27% in China (Lai et al., 2020), although much greater compared to the 3.5% prevalence of PTSD in the US outside of the pandemic (Kessler et al., 2005).

The differences in the prevalence of psychiatric symptoms between the studies are multifactorial. The pandemic hit China before the US and very little was known about COVID-19 at that time. As time passed, we came to understand virulence, modes of transmission, optimal PPE, etc. By the time of our survey, increased knowledge and guidance might have resulted in the HCWs feeling less anxious.

The higher level of depression in this study may be related to the timing of the survey as compared to prior research. The study in China (Lai et al., 2020) occurred a few months into the pandemic while case volumes were high, the disease poorly characterized, and risks uncertain. We surveyed staff approximately 3 months into the pandemic locally. The infection had spread globally by that time, but our system’s rates were declining. This supports our hypothesis that as a wave passes, HCWs move from action driven by anxiety and work ethic, to emotional processing of what they have endured; processing that allows time for the pain, death, and disillusionment of the just witnessed tragedy to set in. If this were the case, it would not be surprising for depression and PTSD to increase. It is also possible that the duration of the pandemic caused a more significant impact on the HCWs. Lastly, the higher rates of PTSD in China may reflect a re-experiencing/triggering effect for those who suffered through SARS1.

In our study, Hispanics reported greater anxiety, PTSD and insomnia than non-Hispanic individuals. As compared to those who identified as White, African-Americans endorsed greater levels of anxiety and insomnia. This may be related to the disproportionately increased risk of COVID-19-related complications among these populations, as well as disparities in access to medical or psychiatric care, access to resources and socio-economic factors. It is also likely that pre-existing inequities, including the sociopolitical, racial, and environmental stressors that these individuals are faced with, are intensified during the pandemic, and extra attention must be given to the unique needs of these populations, as it pertains to the mental health consequences of COVID-19.

It is important to note that by happenstance, this survey was conducted at the same time of an increased awareness of racial inequality, injustice and systemic oppression in our country. Moral outrage, fueled by the ease by which society was able to view and absorb the horrific murder of George Floyd, an unarmed black man killed while in police custody, had fueled a global push for justice and equality not seen in half a century. It is possible that these societal factors had differential effects on HCWs in the study. Higher rates of depression among white HCW could be related to guilt/appreciation of these issues at a time of decreased psychological coping related to the pandemic. Recent surveys of the general population indicated that African- Americans are 20% more likely to experience serious mental health problems than the general population (U S Department of Health and Human Services, 2020), and 40% screened positive for depression and/or anxiety shortly after Floyd’s death (U S Department of Health and Human Services, 2020). It may be that HCWs are uniquely sensitive to these issues.

In our healthcare system, ICU and medical floor staff endorsed significantly higher levels of depression, anxiety, PTSD, and insomnia, compared to office workers and those working in ED’s or outpatient settings. In addition, clinical staff endorsed the greatest symptoms of depression and anxiety, while nurses also endorsed high levels of PTSD and insomnia ICU and inpatient staff had more frequent and prolonged direct contact with COVID-19 patients, resulting in significant changes to protocols, care areas, staff assignments, and workflow. Working on COVID-19 units requires constant vigilance, incessant donning and doffing of PPE, directly witnessing high rates of suffering and death, and associated feelings of futility despite valiant efforts. Workers without first-hand exposure to COVID-19 patients are probably less aware of the tremendously negative impact of COVID-19, and less direct exposure would be expected to decrease rates of PTSD. It is not surprising that inpatient and ICU nurses, who are overwhelmingly female, and have more intense and prolonged exposure to COVID-19 patients, reported higher levels of PTSD and insomnia.

Interestingly, the risk factors for negatively impacted wellbeing were not entirely work-related. Older age (as a dichotomy) afforded some lessening of severity of outcomes. While very few have experienced the stresses caused by the pandemic, it is possible that older people have a greater tolerance for acceptance of difficult situations. Respondents with a prior mental health history fared worse for all four outcomes, suggesting an increased need for support and/or treatment for this group.

It is interesting that the prevalence of psychiatric symptoms among ED workers was similar to that of outpatient and non-clinical staff. Perhaps healthcare providers with an ability to compartmentalize feelings self-select to work in Emergency Medicine. The pace of work in an ED may prevent fully processing emotional responses to tragedy. While the changes to the ED work environment and processes associated with COVID-19 were impactful, these changes may not have been as extreme a departure from the standard environment and operations as that in other healthcare areas. Additionally, patients with suspected COVID-19 infection are quickly triaged to medical floors, which may minimize ED staff exposure. Lastly, it is also possible that the ED staff are were less likely to admit to these symptoms in this survey.

Our data suggest that as the pandemic wears on, HCWs have become more defeated than afraid. When the survey was administered, the first wave of the pandemic had begun to recede, and the healthcare system had well-established procedures and processes to protect the workforce. By this time, HCWs’ anxiety and uncertainty were likely mitigated by increased knowledge and experience of the virus. Nevertheless, chronic physical and emotional stress can cause depressive symptoms. During the chaos of a pandemic, a decrease in HCW engagement in health behaviors would further drive depression.

The pandemic and associated challenges have had a wearing effect on our healthcare workers, and we have not yet replenished spent resources. Our findings support those from Ettman et al. (2020), who found the prevalence of depression in a general population sample in early April to be three times higher than prior to the pandemic; lower income, less financial security, and number of stressors all increased depression (Ettman et al., 2020). Generally, greater access to resources, whether financial or social, is protective against depression.

Our data show that female gender and real-world barriers to working (i.e., child/elder/pet care, responsibilities at second job, risk of infecting self/others) increase the likelihood of mental health symptoms. This is consistent with national and international data suggesting that while men have worse outcomes and higher mortality from COVID-19, women are more likely to bear the brunt of the social and economic consequences of the pandemic. The women’s advocacy group LeanIn.org surveyed 3,000 American adults in May of 2020 about changes to their everyday lives during the pandemic. Full-time working women with children reported almost 20 more hours a week than men on domestic tasks. The disparity is especially acute for women of color and single women, and only 41% of employees surveyed said that their employer has changed policies to allow more flexibility. This, coupled with the estimation that women comprise 70% of healthcare and social sector jobs worldwide (Boniol et al., 2019), suggests that women HCWs face the most significant burdens physically, socially, and economically during the pandemic.