Study design and participants

This cross-sectional, survey-based study recruited HCWs from four hospitals in Turin city, Piedmont, Northern Italy. The hospitals involved were all facing the pandemic and were active in the management of patients with COVID-19, covering a population area of about 870,000 inhabitants. All procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. The study received ethical approval by the local institutional review board (A.S.L. Città di Torino). The study was performed according to the Strengthening the Reporting of Observational Studies in Epidemiology Statement (Supplementary material available at https://doi.org/10.1192/bjo.2020.161).

We chose an online survey to collect the data. The online survey allowed for social distancing to be ensured and the elimination of interpersonal interaction, as required by the Italian Government. Before beginning, on the first page of the questionnaire, participants were informed on the scope of the study and their voluntary participation in the survey, which was executed in accordance with current legislation regarding ethical standards and privacy protection. Participants were asked to confirm that they agreed with the information provided, were aged ≥18 years and consented to participation online by selecting the term ‘I agree’. The questionnaire was anonymous and electronic consent was obtained from all participants before taking part. The survey was sent by email simultaneously to all physicians and nurses of the referring hospitals on 27 April 2020, explaining that enrolment would end automatically upon achievement of the target sample size. To create the survey, we used the free platform Google Forms 2020 (Google Inc., USA; see https://www.google.it/intl/it/forms/about/). A minimum sample size of 377 participants was estimated, using the formula N = Z ² × p(1−p)/α ², considering a margin of error of 5% (α = 0.05), a 95% confidence interval (Z = 1.96) and a population proportion of 50% (P = 0.50). To favour subgroup analysis, the number of participants was at least doubled. After reaching the target number, recruitment was interrupted at midnight of that day (1 May 2020). These 5 days over which the survey was open were during the more demanding period of the COVID-19 pandemic in Italy (10 March to 3 May 2020). At survey closure, a total of 797 respondents had completed the questionnaire.

Survey characteristics

The survey investigated the demographic, family and work characteristics of the recruited HCWs, and participants were asked to complete a set of validated rating scales to assess state and trait anxiety, distress and occupational burnout.

Sociodemographic data included age range, gender, job position and the result of a COVID-19 swab test. Ad hoc questions on the family situation explored the following areas: having children, current living situation (with partner, without partner) and family division owing to the pandemic (i.e. living separated from own family for safety reasons, fear of endangering relatives). Work queries included contact frequency with patients with COVID-19 patients (‘frequent’ if exposure to infected patients was continuous or almost continuous, ‘rare’ if it was only occasional or absent), occurrence of job changes related to the emergency (i.e. relevant modifications in usual job duties and activities as a result of the pandemic) and variations in workload (increased or not increased, based on the subjective opinion of the HCW).

Tools

Italian validations of the following three rating scales were used. The Impact of Event Scale – Revised (IES-R)^{Reference Weiss, Marmar, Wilson and Keane18,Reference Pietrantonio, De Gennaro, Di Paolo and Solano19} is a self-report scale widely used to measure current subjective distress related to a specific stressful life event, including outbreaks.^{Reference Vanaken, Scheveneels, Belmans and Hermans20} It consists of 22 items on a 5-point Likert scale (from 0, ‘not at all’, to 4, ‘extremely’), with respect to how distressing each item has been during the past week. The IES-R assesses symptoms of avoidance, intrusion and hyperarousal. The total score ranges from 0 to 88. The scores identify four levels of distress symptoms: normal, indicating the absence of post-traumatic symptoms (0–23); mild, indicating clinically detectable post-traumatic symptoms (24–32); moderate, indicating the possible presence of post-traumatic stress disorder (33–36); and severe, indicating severe post-traumatic symptoms (≥37).^{Reference Weiss, Marmar, Wilson and Keane18}

The State-Trait Anxiety Inventory – Form Y (STAI-Y)^{Reference Spielberger, Pedrabissi and Santinello21,Reference Spielberger, Gorsuch, Lushene, Vagg and Jacobs22} is one of the most widely used self-reported measures of anxiety.^{Reference Bergua, Meillon, Potvin, Bouisson, Le Goff and Rouaud23} It is a self-report inventory aimed at assessing both state (STAI-Y1) and trait (STAI-Y2) anxiety, consisting of 40 items on a 4-point Likert scale (from 1, ‘almost never’, to 4, ‘almost always’).^{Reference Spielberger, Gorsuch, Lushene, Vagg and Jacobs22} State anxiety refers to ‘subjective feelings of tension, apprehension, nervousness, and worry, and activation or arousal of the autonomic nervous system, which exist at a given moment in time and at a particular level of intensity’ (page 4).^{Reference Spielberger, Gorsuch, Lushene, Vagg and Jacobs22} Trait anxiety refers to ‘relatively stable individual differences in anxiety-proneness, that is, to differences between people in the tendency to perceive stressful situation as dangerous or threatening and to respond to such situations with elevations in the intensity of their state anxiety reactions’ (page 5).^{Reference Spielberger, Gorsuch, Lushene, Vagg and Jacobs22} Twenty items are dedicated to each anxiety type. Scores for both subscales range between 20 and 80. For both subscales, scores of 40–50 indicate mild anxiety, 50–60 indicate moderate anxiety and >60 indicate severe anxiety.^{Reference Spielberger, Gorsuch, Lushene, Vagg and Jacobs22}

The Maslach Burnout Inventory (MBI)^{Reference Sirigatti, Stefanile and Menoni24,Reference Maslach and Jackson25} is the most common instrument used to measure occupational burnout.^{Reference Poghosyan, Aiken and Sloane26} The questionnaire is composed by three scales: emotional exhaustion (nine items), examining the feeling of being emotionally bitter and exhausted from work; depersonalisation (five items), which measures a cold and impersonal response; and personal accomplishment (eight items), assessing the feeling of one's own competence and desire to succeed in working with others. High scores on the personal accomplishment scale, unlike previous emotional exhaustion and depersonalisation scales, indicates greater personal achievement and, consequently, a lower level of burnout. All of the MBI items are scored with a 7-point Likert scale (from 1, ‘never’, to 7, ‘daily’). Consistent with the literature, we considered the level of burnout to be high if emotional exhaustion scores were ≥24, personal accomplishment scores were ≤29 and depersonalisation scores were ≥9; moderate if emotional exhaustion scores were 15–23, personal accomplishment scores were 30–36 and depersonalisation scores were 4–8; and low if emotional exhaustion scores were ≤14, personal accomplishment scores were ≥37 and depersonalisation scores were ≤3.^{Reference Sirigatti, Stefanile, Maslach and Jackson27}

Statistical analysis

The normal distribution of variables was tested with Kolmogorov–Smirnov and residual analysis tests. Continuous variables were expressed as mean ± s.d. or median and interquartile range, as appropriate. Qualitative variables were expressed as absolute values of frequency and percentage values. Differences between independent groups were evaluated by a t-test for continuous variables with normal distribution, and the Mann–Whitney U or Kruskal–Wallis test for continuous variables with non-normal distribution. Categorical variables were compared with the chi-square test or Fisher's exact test, as appropriate. A multivariable logistic regression was applied, using the presence of moderate or severe symptoms (for each psychological scale) as a dependent variable and possible risk factors as independent variables. The associations between risk factors and psychological symptoms are presented as odds ratios and 95% confidence intervals. Statistical significance was considered for P-values <0.05 in all analyses. Statistical analysis was performed with software package SPSS (IBM SPSS Statistics for Windows, version 22; IBM Corp., New York, USA).