Psychological impact of COVID-19 pandemic on healthcare workers in a highly burdened area of north-east Italy

The type of research applied in this study is explanatory in nature. This study represents the baseline evaluation of a larger longitudinal project jointly launched by the Verona University Hospital Trust and the Section of Psychiatry at the University of Verona. The longitudinal study will assess the psychological impact of the COVID-19 pandemic on hospital workers during the lockdown period, after two months, and at one year. Baseline data collection was carried out between 21 April and 6 May 2020 using a web-based questionnaire hosted on the survey platform ‘SurveyMonkey’. The online survey required about 15–20 min to be completed. At the baseline evaluation, each participant was asked to generate a password that should have been used also at the subsequent follow-ups; this was required for the research team to longitudinally link the questionnaires completed by a given hospital worker at each evaluation point. The study description and the invitation to participate as well as the link to the online questionnaire was published in the hospital's newsletter and sent via e-mail to all hospital workers by the trust administration. A reminder for completing the questionnaire was sent around after one week. The survey was anonymous, and confidentiality of information was granted. The survey was approved by the Ethics Committee of the Provinces of Verona and Rovigo (approval No. 22002; 17 April 2020).

The Verona University Hospital is the second-largest hospital trust in Italy in terms of bed quantity and the fifth largest in terms of admissions. The hospital staff is composed of 5942 personnel (including nearly 1200 residents of the medical specialty schools at the University of Verona). Beginning 17 March 2020, the Veneto regional government converted part of the hospital into a ‘COVID-19 hospital’. Thus, dedicated pathways for both suspected and confirmed COVID-19 cases were established within the hospital, as well as in other hospital units located in clearly restricted areas specifically devoted to the treatment of COVID-19 patients. All staff working in the Verona University Hospital during the lockdown phase of the pandemic were asked to participate in the study.

The psychological distress was assessed using the Impact of Event Scale-Revised (IES-R) (Weiss and Marmar, 1997; Craparo et al., 2013), a 22-item self-report that measures subjective distress caused by traumatic events on a 5-point scale from 0 (not at all) to 4 (extremely) during the previous seven days. The IES-R was completed only by those who had experienced stressful/traumatic events at work related to COVID-19. Participants were first asked to identify a stressful life event that they may have encountered at work and to specify which kind of stressful event it was. Respondents were then asked to rate how much they were distressed or bothered during the past seven days by each item listed in the IES-R. The maximum score is 88 (worst post-traumatic stress state). We used a cut-off score of 24 for detecting symptoms of post-traumatic distress that deserve clinical attention (Creamer et al., 2003).

Symptoms of anxiety were assessed by the Self-rating Anxiety Scale (SAS) (Zung, 1971, 1980) that contains 20 items, each rated on a 5-point scale from 1 (a little of the time) to 4 (most of the time). The maximum score of 80 indicates an extremely high anxiety level and the cut-off score for clinically significant anxiety symptoms is 36 (Dunstan and Scott, 2020).

Symptoms of depression were assessed by the Patient Health Questionnaire (PHQ-9) (Kroenke et al., 2001), a self-rated 9-item scale that asks if the subject has experienced symptoms of depression in the previous two weeks. Subjects are asked to rate how often each symptom occurred: 0 (not at all), 1 (several days), 2 (more than half the days), or 3 (nearly every day). Total PHQ-9 scores range from 0 (absence of depressive symptoms) to 27 (most severe depressive symptoms). We used a cut-off score of 10 to indicate a condition that potentially deserves clinical attention (Kroenke and Spitzer, 2002).

Two ad hoc instruments exploring perceived job stress and perception of risk, modified, respectively, from Imai et al. (2005) and Maunder et al. (2006), were administered.

Personal socio-demographic information and job characteristics were also collected, including gender, age, living condition, having psychological problems developed before the COVID-19 outbreak requiring specialised help, occupation, length of working experience, place of work (hospital unit). For the purpose of analysis, the various hospital units were grouped according to the degree of clinical engagement with COVID-19 patients, from most engaged to least engaged: Intensive Care Units (that during the lockdown phase were entirely dedicated to critically ill COVID-19 patients), sub-intensive COVID wards (i.e. infectious disease, pulmonary medicine and internal medicine wards specifically dedicated to COVID-19), frontline services dealing with COVID patients (i.e. radiology and emergency department), non-COVID wards, laboratory diagnostic services (i.e. laboratory medicine, transfusion medicine, immunology, pathology, microbiology) and administration.

Statistical analyses were carried out using SPSS 22 and Stata 15. Descriptive statistics were reported in frequencies and percentages. Comparisons between categorical variables were performed by Chi-square or Fisher's exact tests where appropriate.

The precision of the proportion estimate for each adverse outcome was determined by calculating the margin of error for the two-sided 95% confidence interval (95%CI).

The association between each adverse outcome (post-traumatic distress, anxiety and depression) and each potential risk factor selected a priori, on clinical or empirical grounds and derived from the relevant literature (Kisely et al., 2020; Pappa et al., 2020) was explored by estimating unadjusted odds ratios (ORs) and 95%CIs using univariable logistic regression models. Subsequently, multivariable logistic regression models for the same outcomes gave adjusted ORs and 95%CIs. Goodness-of-fit measures were estimated for these models.

In a secondary analysis, missing data on outcomes were estimated using a multiple imputation approach based on logistic regression (‘mi impute logit’ Stata command).

A sensitivity analysis using an alternative modelling strategy, that is by considering the outcomes in terms of continuous scores and estimating linear regression models, was performed.

The alpha level was set to 0.05 for all effects.

|Overall, 2195 workers, corresponding to 36.9% of the eligible population, completed the on-line survey.

The representativeness of participants was assessed by comparing two key characteristics found to be associated with response/non-response pattern for all the three outcome domains and for which official statistics from the Verona University Hospital were available (i.e. occupation and exposure to COVID-19 patients). For details, see on-line supplementary part 1. Overall, the study sample overlapped with the Verona University Hospital staff both in terms of occupational profile and percentage of healthcare workers employed in units directly engaged with COVID-19 patients, thus indicating that the sample addressed here is representative of the eligible population.

Table 1 shows personal and job characteristics of participants.

Table 2 (upper part) shows percentages of healthcare workers reporting some kind of job-related stress experienced during the COVID-19 pandemic. The vast majority of participants reported more stress at work; moreover, most participants reported more conflict among colleagues, increased workload and additional tasks that they were not responsible for pre-COVID.

Table 2 (lower part) reports the perception of risk. The vast majority of healthcare workers were afraid of falling ill with COVID-19; however, they considered the risk of infection as part of their job. When stratifying levels of job stress and perception of risk by place of work, a significantly higher percentage of staff reporting more conflict among colleagues, increased workload, unusual additional tasks and greater fear of infection was found among those working in ICUs and in sub-intensive care wards for COVID-19 than in other hospital units; moreover, when stratifying by occupational profile, a higher percentage of nurses reported greater job stress and perception of risk than other hospital workers (see the on-line supplementary part 2).

The IES-R, the SAS and the PHQ-9 were completed, respectively, by 91.3%, 90.7% and 90.2% of eligible participants. When comparing percentages of completers and non-completers with respect to personal information across the three outcome measures, no staff characteristic was significantly associated with the pattern of response/no response, suggesting that the sample addressed here was not biased in terms of completion of instruments. Differences between completers and non-completers are detailed in the on-line supplementary part 3.

Overall, 63.2% reported having experienced some traumatic events related to COVID-19. The percentages of those who reported COVID-related traumatic experiences significantly differed by occupational profile (nurses 74%; other healthcare workers 63%; physicians 59%; residents 56%; administrative staff 41%; p < 0·001) and place of work (ICUs 89%; sub-intensive COVID units 90%; other frontline services 77%; no COVID wards 59%; laboratory services 49%; administration 45%; p < 0·001), with nurses and healthcare staff working in both ICUs and sub-intensive COVID units reporting more traumatic events. The frequency distribution of specific traumatic events is given in the on-line supplementary part 4. In brief, the most frequent traumatic themes were related to the fear of infection, demanding work conditions and dealing with the death and dying, reported, respectively, by 28.7%, 27.4% and 15.4% of respondents.

Among those who reported a COVID-related traumatic experience, 53.8% (95% CI 51·0%–56.6%; margin of error 2.8%) showed clinically relevant symptoms of post-traumatic distress. Moreover, in the overall sample, 50.1% (95% CI 47·9%–52.3%; margin of error 2.2%) reported symptoms of clinically significant anxiety and 26.6% (95% CI 24.7%–28.5%; margin of error 1.9%) symptoms of at least moderate depression.

As shown in Table 3, stratifying by personal and job characteristics, the proportion of participants scoring above the cut-off points in the three outcome scales widely differed.

Overall, women, nurses and staff working in ICUs or sub-intensive COVID units reported higher percentages of post-traumatic distress, anxiety and depression. Specifically, 57% of women reported a severe level of post-traumatic distress, 56% displayed severe anxiety and 30% severe depression. In addition, 65% of nurses reported severe post-traumatic symptoms, 63% showed severe anxiety and 33% severe depression. Healthcare workers in ICUs reported symptoms of post-traumatic distress in 66% of cases, 64% severe anxiety and 42% severe depression; similarly, staff working in sub-intensive COVID units reported percentages of post-traumatic distress, anxiety and depression, respectively, in 65%, 61.4% and 36% of cases.

When computing the percentage of participants scoring above the cut-off point in all the three outcome scales by occupational profile and place of work, the highest proportions were found, respectively, among nurses (35%) (Fig. 1a) and healthcare staff working in ICUs (39%) (Fig. 1b).

Table 4 shows the unadjusted odds ratios (ORs) estimating the association between each adverse outcome (post-traumatic distress, anxiety and depression) and each potential risk factor.

Table 5 reports multivariable analyses. Adjusted ORs showed that living alone, having a work experience longer than 20 years, being a nurse, having pre-existing psychological problems, having experienced a fear of getting infected with COVID-19 are associated with increased risk of developing severe symptoms of post-traumatic distress; whereas, working in hospital wards/units/offices non-directly engaged in the diagnosis and treatment of COVID patients appeared to be a protective factor concerning those who work in ICUs or sub-intensive COVID units.

With respect to anxiety, being a woman, a nurse or other healthcare staff, having experienced a traumatic event related to COVID-19, having pre-existing psychological problems, and having perceived fear of getting infected are associated with an increased risk of developing symptoms, whereas staff working in non-COVID units display a reduced risk with respect to those working in ICUs.

An increased risk of developing depression was associated with being female, living alone, being a nurse or other healthcare worker, having experienced a traumatic event related to COVID-19, having pre-existing psychological problems, and having perceived fear of getting infected, whereas, working in hospital wards/units/office non-directly engaged in the diagnosis and treatment of COVID-19 patients appeared to be a protective factor compared with those working in ICUs or sub-intensive COVID units.

Overall, being a nurse increased the risk of developing symptoms of post-traumatic distress and anxiety by at least two times with respect to physicians. Having experienced a COVID-related traumatic event increased the risk of developing anxiety and depression by at least 2.5 times. Moreover, having pre-existing psychological problems that required specialised treatment increased the risk of developing post-traumatic distress, anxiety and depression by more than 2.5 times. Finally, the fear of getting infected by COVID-19 increased the risk of developing symptoms of post-traumatic distress by at least two times and the risk of developing anxiety by nearly three times.

Multiple imputation (MI) analysis for missing data was performed for all regression models. Variables included in the MI were occupation, exposure to COVID-19 patients and having pre-existing psychological problems, all significantly associated with response/no response pattern for the outcome measures. All the three variables included in the MI had no missing value (n = 2195). Overall, the regression estimates from the complete case and MI analyses are overlapping for all models (details in the on-line supplementary part 5).

Regression models were re-estimated by using an alternative modelling strategy, i.e. treating outcomes as continuous scores. Sensitivity analysis showed that the results of the regressions did not significantly change using an alternative modelling strategy (see the on-line supplementary part 6).

The relevance of our findings should be viewed in the light of the fact that: (a) Italy was the first major western country to be affected by the COVID-19 pandemic (and since then one of the most affected countries in the world) and the first in the western world that had to face the related healthcare emergency WorldMeter, 2020); (b) Veneto, within Italy, was the region (together with Lombardy) where the first COVID-19 outbreak was registered and since then has been one of the most affected Italian regions (Ministero della Salute, 2020; Regione Veneto, 2020); (c) the province of Verona was the most burdened area within Veneto during the lockdown period (11 March−4 May 2020), both in terms of deaths and infected cases (Regione Veneto, 2020). Other strengths of this study include the multidimensional examination of psychological outcomes, the large hospital population working in a critical geographical area, the wide array of occupational profiles considered, as well as the ability to collect data, given extreme time constrains and the difficulties associated with conducting research amidst a pandemic. Among the possible limitations, it should be mentioned that the relatively low response rate was 37%. However, web-based surveys have generally lower response rates than face-to-face or telephone interviews or mail surveys; a meta-analysis reported a mean response rate similar to our study (39.6%) (Cook et al., 2000). In addition, surveys involving physicians have even lower overall response rates (35%) (Cunningham et al., 2015). Another limitation to be considered is that highly distressed healthcare workers, who are generally less likely to engage in surveys or other forms of psychological interventions, might be underrepresented in the study (Kang et al., 2020).

It is necessary that healthcare systems around the world carefully address the psychological well-being of healthcare workers by, e.g. active monitoring reactions and performance, altering assignments and schedules, modifying expectations, assessing occupational risks and offering − where necessary − psychosocial support (Pfefferbaum and North, 2020). A rapid response team in crisis situations including mental healthcare professionals should be established within each COVID-19 frontline hospital (Kang et al., 2020). The type of intervention to provide depends on the stage of the pandemic and the specific mental healthcare needs, ranging from peer support to professional aid. Among the first kind of interventions, a modified version of the ‘Anticipate, Plan, and Deter’ responder risk and resilience model (Schreiber et al., 2019) seems to be promising (Albott et al., 2020). Among professional interventions, novel therapy approaches such mindfulness, relaxation therapies, EMDR protocols provided through telemental health platforms seem to have a promising role (Bassan et al., 2020; Xiao et al., 2020; Yang et al., 2020). In order to prevent the development of psychological disorders in the long term, active monitoring and provision of psychological support should be delivered also once the crisis begins to recede (Greenberg et al., 2020).

Future longitudinal research is needed to evaluate the psychological impact of the pandemic on healthcare workers over the medium- and long-term period and to establish patterns and co-occurrence of risk factors for adverse mental health outcomes. Intervention studies in real-world settings should be additionally conducted and implemented to investigate under which interventions and specific circumstances resilience may be best fostered and the mental health of frontline healthcare professionals supported during and after a disease outbreak.

Data will be available upon reasonable request.