Depression and burnout among Chinese nurses during COVID-19 pandemic: a mediation and moderation analysis model among frontline nurses and nonfrontline nurses caring for COVID-19 patients

A large sample online survey was conducted. The researchers selected the hospitals to be studied based on the following criteria: (1) The selected hospitals were public hospitals, and the scale of the hospital was a tertiary hospital or above (the number of hospital beds was 500 or above); community hospitals, private hospitals, and clinics that provide primary medical, preventive, rehabilitative, and health care services were not be included in this study; (2) Comprehensive hospitals were selected to ensure the representativeness of the study; specialty hospitals (excluding specialty COVID-19 hospitals) were not be included in this study; (3) In the case of specialty COVID-19 hospitals, the hospital scale was 100 hospital beds or above; and (4) Hospitals for which the manager could be successfully contacted and who agreed for the research to be conducted in their hospital. Ultimately, a total of ten hospitals in Shanghai, Chongqing, Sichuan, Gansu, and Yunnan (including five specialized COVID-19 hospitals) met the above requirements, and the managers were willing to participate in this study. Departments of specialized COVID-19 hospitals, such as fever outpatient clinics, isolation wards, and observation wards, were enrolled in this study. Five researchers connected the managers of all of the departments online by using the software WeChat or Tenchent conference within the selected hospitals. Inclusion criteria were as follows: (1) more than 18 years old; (2) no cognitive impairment or psychiatric disorders; and (3) registered nurses. In the demographic information section of the questionnaire, the study subjects were asked to provide personal information related to the inclusion exclusion criteria, and the researcher screened the study subjects based on the information they filled in, with the subjects who did not meet the inclusion criteria being removed.

Prior to formal data collection, an online questionnaire was created by the researchers utilizing the WJX software (www.​wjx.​com), which contained three components: a note about the purpose of the study, the methodology, and considerations for participation; an informed consent form; and the questionnaire. After obtaining consent from the manager of each department, the online questionnaires were sent to them. After the online questionnaires were sent to the managers of each department, the department manager organized a half-hour online meeting at a time when most of the nurses were off (e.g., Friday evening), and the researcher explained the basic content of the survey to the nurses. After the explanation, the questionnaires were sent to each nurse by the managers of the departments, and each nurse was free to choose the time to complete the questionnaires within one week according to their own situation. A small number of nurses on duty conducted additional meetings based on departmental scheduling. They were also asked to complete the questionnaires within a week.

Nurses had to read the note for at least 30 s, then the informed consent form for at least 30 s and select the option of willing to participate in this study. Then, they were able to complete and submit the questionnaires. A total of 6103 questionnaires were sent from March 2022 to September 2022, and 5500 nurses chose to participate in this study. Of these, 5350 nurses met the inclusion criteria.

Strict quality control measures were utilized to ensure the authenticity of the questionnaire data in this study. For example, a person could only fill out a questionnaire after reading and signing the informed consent. Each questionnaire was filtered using automatic filtering rules and manually inspected by the researcher after submission. The filter criteria were as follows: (1) Questionnaires with similar questions with opposite answers were excluded; (2) Questionnaires for which the same response was recorded for each item were excluded; and (3) Questionnaires that took less than one minute to complete were also be excluded. Questionnaires could not be submitted unless they were complete. After extreme value testing and data cleaning, a total of 4517 study subjects were included. The details of the sampling process are demonstrated in Fig. 2.

A demographic characteristics questionnaire was developed to collect the nurses’ general information, including gender, age, level of hospitals, educational background, monthly income, job title, actual positions, marital status, and fertility status.

The depression subscale of the Depression-Anxiety-Stress Scale-21 (DASS-21) was utilized. The DASS was developed by Lovibond et al. in 1995, with the initial objective of distinguishing and defining common emotional disorders, such as depression, anxiety, and stress, through a consistent measurement system for administration and scoring [36]. The scale can be applied as a psychometric aid for clinical diagnosis and as a rapid and effective subject screening tool for relevant studies [36, 37]. The DASS-21 is a revised and streamlined version of the DASS that enhances the efficiency of identifying and assessing symptoms of corresponding emotional disorders, while retaining the same stable factor structure and equally favorable reliability and validity as the full version of the DASS [36]. The depression subscale of DASS-21 has seven items measured on a four-point Likert scale, with higher scores representing more severe depression. The Cronbach's α coefficient of this study is 0.854. The total Kaiser–Meyer–Olkin measure of this study was 0.898. Results of confirmatory factor analysis indicated the χ² /DF of the subscale is 2.392, GFI = 0.999, RMSEA = 0.016, indicated that this scale has good construct validity. The total score was obtained by multiplying the sum of all subscale scores by 2. Depression was classified as normal for a score of 0 to 9, mild for a score of 10 to 13, moderate for a score of 14 to 20, and severe for a score of 21 or higher [36].

The Maslach Burnout Inventory-General Survey (MBI-GS) developed by Schaufeli et al., which contains the three dimensions emotional exhaustion, depersonalization, and decreased personal accomplishment and 16 items, was utilized [38]. The Chinese version of the MBI-GS was revised by Chaoping Li et al. and contains 15 items. The Chinese version of the MBI-GS is measured on a 7-point Likert scale, with 0 representing "never" and 6 representing "very often" [39]. The Cronbach's α coefficients for emotional exhaustion, depersonalization, and decreased personal accomplishment were 0.918, 0.883, and 0.890, respectively. The total Cronbach's α coefficient of the MBI-GS was 0.942. The Kaiser–Meyer–Olkin measures for emotional exhaustion, depersonalization, and decreased personal accomplishment were 0.877, 0.788, and 0.901, respectively. The total Kaiser–Meyer–Olkin measure of the MBI-GS was 0.952. Results of confirmatory factor analysis indicated the χ²/DF of the subscale is 3.483, GFI = 0.965, RMSEA = 0.023, indicated that this scale has good construct validity.

The Comprehensive Interpersonal Relationship Diagnostic Scale developed by Richang Deng was utilized. It is used to measure individuals' psychological distress in interpersonal relationships, with 28 items in 4 dimensions: conversational behavior, communicative distress, treating others with respect, and heterosexual interaction. Each item is answered yes (1 point) or no (0 points). The higher the total score is, the more serious the interpersonal distress [40]. The Cronbach's α coefficient of this study was 0.878. The Kaiser–Meyer–Olkin measure of the Comprehensive Interpersonal Relationship Diagnostic Scale was 0.930. Poor interpersonal relationships were classified as mild for a score of 0–9, moderate for a score of 10–14, and severe for a score of 15 or above [40]. Results of confirmatory factor analysis indicated the χ² /DF of the subscale is 5.490, GFI = 0.930, RMSEA = 0.064.

The Simplified Coping Style Questionnaire (SCSQ) developed by Yaning Xie was utilized in this study [41]. The questionnaire has 20 items, with questions 1 to 12 measuring positive coping styles and questions 13 to 20 measuring negative coping styles. Only the positive coping style subscale was used in this study. SCSQ was based on a 4-point Likert scale, A score of 0 indicates no adoption and a score of 3 indicates frequent adoption. The Cronbach's α coefficient of this study is 0.913. The Kaiser–Meyer–Olkin measure of the Simplified Coping Style Questionnaire (SCSQ) is 0.942. Results of confirmatory factor analysis indicated the χ²/DF of the subscale is 4.586, GFI = 0.977, RMSEA = 0.233, indicated that this scale has good construct validity.

Data were analyzed by the software SPSS 26.0 and the SPSS PROCESS 3.3 macro. Frequency and percentage statistics were adopted to describe the demographic characteristics of nurses. Quantitative data was described by mean ± SD. The normality of the data was tested and a nonnormal distribution was detected for all variables. Therefore, the Spearman correlation analysis and nonparametric tests were conducted to investigate the potential relation between variables. Before testing the moderated mediation model, the mediation role of positive coping styles and interpersonal relationship were tested by model 4 in PROCESS 3.3 macro. Then, the moderated model was tested by model 5. A bootstrap method with 10,000 samples was conducted to reduce bias, and 95% confidence intervals were tested. If 0 was not included in the 95% confidence intervals, then the path coefficient was regarded as significant. Since the moderating variables are dichotomous, they are directly enrolled in the model processing.

A total of 4517 subjects were enrolled in this study. The majority of the participants were women (4377, 96.9%). More than half of the participants (2942, 65.1%) had obtained a bachelor’s degree. The majority of nurses in this study were salaried at less than ¥10,000: 1762 (39.0%) of the nurses earned less than ¥5000 per month and 2606 (57.7%) earned ¥5000–10, 000 per month. The majority of the participants (3218, 71.2%) were married, and nearly half of the participants (2040, 45.2%) had a child. A total of 2334 (51.7%) were frontline nurses caring for COVID-19 patients, and 3671 (81.3%) were afraid of being infected with COVID-19. Details of the participants’ demographic features are demonstrated in Table 1.

Significant connections between nurses’ gender, age, educational background, marital status, professional titles and the depression levels were detected by nonparametric tests. Fear of COVID-19 infections and being frontline nurses caring for COVID-19 patients were likely to be related with nurses’ depression. Details are shown in Table 1.

The results of Spearman correlation analysis indicated that burnout among nurses was closely related with nurses’ depression (r = 0.681, p < 0.001). Positive coping styles were negatively related with nurses’ depression (r = -0.360, p < 0.001). Poor interpersonal relationships were positive associated with nurses’ depression (r = 0.551, p < 0.001). Furthermore, burnout, positive coping styles, and interpersonal relationships were related with each other (Table 2).

Mediating effects analysis was conducted after controlling for demographic variables. The results of mediation analysis of positive coping styles and interpersonal relationships are demonstrated in Table 3. The total effect of burnout on depression was estimated as β = 0.66 (95% CI: 0.64 to 0.68). Direct effect of burnout on depression was estimated as β = 0.51 (95% CI: 0.48 to 0.53). The total indirect effect of burnout on depression was also estimated (β = 0.15, 95% CI: 0.14 to 0.17). Furthermore, the mediation analysis indicated that positive coping styles mediate the relationship between burnout and depression (β = 0.03, 95% CI: 0.03 to 0.04). Interpersonal relationships also mediated the relationship between burnout and depression (β = 0.12, 95% CI: 0.11 to 0.13). Details of the mediation analysis are shown in Table 3.

After parallel mediation analysis was performed, a moderated mediation analysis was conducted. On the basis of the consideration of conditioning factors, the total indirect effect of burnout on depression was estimated as β = 0.15 (95% CI: 0.14 to 0.17). Positive coping styles mediated the relationship between burnout and depression (β = 0.04, 95%CI: 0.03 to 0.04). Interpersonal relationships also mediated the relationship between burnout and depression (β = 0.12, 95% CI: 0.10 to 0.13).

Moderated analysis identified that being frontline nurses for COVID-19 patients moderated the direct effect of burnout on depression (β = 0.12, 95% CI: 0.08 to 0.16). Details shown in Table 4. Regarding the direct impact of burnout on depression, frontline nurses (β = 0.54, 95% CI: 0.51 to 0.57) and nonfrontline nurses (β = 0.42, 95% CI: 0.38 to 0.45) were both significant (Fig. 3) (Table 5). The final moderated mediation model was demonstrated in Fig. 4.

Concordant with hypothesis 1, positive coping styles could mediate the association between burnout and depression, which was similar to the views pointed out by previous studies [42‐44]. For instance, Wang et al. indicated that coping styles could mediate the relationship between perceived pressures and mental distress (anxiety, depression) among physicians in China [42]. Zhou et al. also indicated that a positive coping style can serve as a protective factor to mediate the relationship between burnout and anxiety. Physicians with lower levels of burnout are more likely to address issues through positive coping styles, which in turn lead to a reduction in psychological distress [44]. According to conservation of resource theory (COR), both the loss of existing resources and the failure to acquire new resources can trigger a stress response in individuals [45]. When confronted with a desperate situation of resource depletion, the individual's self-protective defense mechanism will be triggered, and the individual will redeploy coping styles [45]. When the individual's burnout is severe and psychological resources are depleted, the individual may adopt less positive coping styles and more negative coping to protect themselves due to defense mechanisms [46]. Individuals are likely to experience depression when their positive coping style is reduced and negative coping is enhanced.

Regarding hypothesis 2, this study verified that burnout among nurses could impair individuals' interpersonal relationships and lead to depression. Research on medical students has demonstrated that burnout among medical students is closely related to interpersonal problems [27]. According to COR theory, individuals’ social competence tends to be impaired when they suffer from emotional exhaustion [46]. Thus, these individuals may experience more challenges in interacting in social situations and then engage in less-than-appropriate social behavior [13]. Furthermore, interpersonal relationships have been proven to be an important predictor of depression [47, 48]. The lack of social skills and poor speech can cause problems in interpersonal relationships, and subsequently, this interpersonal stress can lead to depression [49]. Regardless of the reasons, interpersonal relationship distress predicts individuals’ depression levels [50]. Therefore, the relationship between burnout and depression could be partially mediated by interpersonal relationships.

The results of this study indicated that the mediating effect of interpersonal relationships was significantly larger than that of positive coping styles, and this difference occurred mainly in the second half of the mediation model. Charles indicated that an individual's living environment is a complete ecosystem, including family, friends, work, etc. [51]. Human beings are the subject of continuous interaction with various ecosystems [51]. Therefore, we infer that when an individual's interpersonal relationships are impaired, this may lead to an imbalance in the system that causes psychological problems. As individuals adopt less positive coping styles, they may be more inclined to adopt avoidance, substance abuse, and other coping styles due to psychological defense mechanisms, which can somewhat buffer the onset of negative emotions [52‐54]. However, whether the difference between the two mediating effects stems from the above mechanism needs to be further explored.

The results of moderated analysis confirmed hypothesis 3, as burnout could affect depression among nurses directly, and the direct impact of burnout could be magnified among frontline nurses caring for COVID-19 patients. Compared to those who were nonfrontline nurses for COVID-19 patients, the association between burnout and depression increased significantly. Similar to the results of the moderated analysis of this study, a study focused on 987 subjects also stated that perceived health risk related to COVID-19 significantly affected the relationship between burnout and job performance [32]. A safe workplace serves as a work resource that can assist in reducing employee burnout and improving employee performance [55]. During the COVID-19 pandemic, a safe workplace could contribute to reducing employees' psychological problems by reducing their fear of external risks and increasing their sense of psychological security [55]. When nurses are exposed to high-risk environments for prolonged periods of time, such as being confronted with chronic uncertainty and fear regarding COVID-19, psychological security may be diminished, resulting in nurses being more vulnerable to mental health problems. Additionally, COVID-19 is a severe social concern that affects multiple aspects of society and is more than a medical issue [56]. In contrast to the common stressors that are confronted in daily life, COVID-19 is unpredictable, causing great uncertainty and fear among employees [56]. When employees are exposed to high levels of pressure, their psychological resilience may be lost and have a negative impact on their mental status [32]. Thus, frontline nurses for COVID-19 patients who suffer from burnout are more vulnerable to depression.