*2.7. Statistical Analysis*

All variables were evaluated using descriptive statistics and values were expressed as means and standard deviations for continuous variables. The prevalence of fatigue and depression was determined as a percentage. Independent *t*-tests were performed to evaluate continuous variables by gender. Analysis of variance (ANOVA) with Bonferroni correction was used to check for differences between groups in continuous variables. Pearson Correlation was performed to determine the strength and direction of the relationship between variables. Linear regression models were built to investigate whether related variables were significant predictors of the independent variable. The evaluation of the linear regression hypotheses (linear relationship, independence, homoscedasticity and normality) was carried out by visual inspection of the variables, residual diagrams and quantile-quantile (QQ) plots. Statistical significance was set at *p* < 0.05 (two-tailed) and analyses were performed using IBM SPSS Statistics 23 (IBM SPSS Statistics for Windows, Version 23.0, IBM Corp, Armonk, NY, USA). Mediation and moderation analyses were conducted using the Hayes SPSS Process Macro. Average scores for the research parameters were compared versus the results obtained in our previous studies in the pre-COVID-19 era [28,35]. IBM SPSS AMOS 23 Graphics was utilized to construct Figure 1.

**Figure 1.** Mediation analysis of Copenhagen Burnout Inventory (CBI) on Sense of Coherence questionnaire (SOC)—Beck's Depression Inventory (BDI) relationship.

#### **3. Results**

A total of 101 men (15.3%) and 559 women (84.7%) nurses participated in the study; 375 nurses reported being married (56.8%), 211 (32.0%) unmarried, and 74 (11.2%) divorced. The sample evidenced no difference in marital status (ANOVA *p* > 0.05). Female nurses had higher burnout values (*t* test *p* < 0.01, 49.03 vs. 38.74) and depression (*t* test *p* < 0.01, 11.29 vs. 6.93), but also lower SOC values compared to men (*t* test *p* < 0.05, 59.45 vs. 65.13, Table 1). Gender differences were statistically significant in both the burnout subscales and the SOC (Table 1).

Regarding depression, 25.5% showed mild depression, 13.5% moderate depression and 7.6% severe depression. On the fatigue scale, 47.1% scored above cutoff. The mean depression score, compared to previous studies [35], was statistically higher in both women (11.29 vs. 8.5 sample *t*-test *p* < 0.01) and in men (6.93 vs. 4.6 sample *t*-test *p* < 0.01). The average SOC was statistically lower compared to previous studies [28] in Greek nurses (60.33 vs. 63.6 sample *t*-test *p* < 0.01).

**Table 1.** General characteristics of nursing staff and SOC/CBI scores with regards to gender.


Notes: \* independent *t*-test *p* < 0.05; \*\* independent *t*-test *p* < 0.01. Abbreviations: P, Participants; D.S., Descriptive Statistics; W.E., Work Experience; BDI, Beck's Depression Inventory; A, Comprehensibility; B, Manageability; C, Meaningfulness.

Significant negative correlations were evidenced among scores on the SOC scale (*p* < 0.01) with both CBI as well as with BDI scales. However, a positive correlation (*p* < 0.01) was indicated between CBI and BDI (Table 2).


**Table 2.** Correlations among age, work experience (in years), CBI, SOC, and BDI.

Notes: \* *p <* 0.05 or \*\* *p <* 0.01.

A stepwise multiple regression analysis was performed to identify the best predictors for BDI. We defined depression as a dependent variable and as independent variables: work experience, age, burnout and sense of coherence. We tested this model for the absence of multicollinearity. This regression showed that 43.7% of the variation in the BDI score can be explained by the CBI, while an additional 8.3% is explained by SOC; the other variables did not explain the variance in BDI (Table 3).


**Table 3.** Stepwise multiple regression (only statistically significant variables are included).

Notes: Beta = standardized regression coefficient; correlations are statistically significant at the \* *p <* 0.01 level.

Bootstrapping was performed with the Hayes SPSS Process Macro to examine whether burnout mediated the relationship between SOC and depressive symptoms: based on 5000 bootstrap samples, a significant indirect relationship between SOC and depressive symptoms was mediated by burnout (Table 4, Figure 1). The outcome variable for the analysis was BDI. The predictor variable for the analysis was SOC. The mediator variable for the analysis was CBI. The indirect effect of CBI on BDI was found to be statistically significant [B= −0.1597, 95% C.I. (−0.1899, −0.1308), *p* < 0.05]. The model explains 42.5% of the variance in the outcome variable. Standardized coefficients for the variables are depicted in Figure 1.

**Table 4.** Mediation analysis of Copenhagen Burnout Inventory (CBI) on Sense of Coherence (SOC)– Beck Depression Inventory (BDI) relationship.


\* Based on 5000 bootstrap samples.

Finally, the moderation role of SOC in the relationship between CBI and BDI was assessed. A simple moderation analysis was performed using the PROCESS method (with CBI as the predictor variable, BDI as the outcome variable and SOC as the moderator variable) (Table 5). The interaction between CBI and SOC was found to be statistically significant [B = −0.0051, 95% CI (−0.0065, −0.0036) *p* < 0.05]. The effect of CBI on BDI showed corresponding results.

**Table 5.** Moderation analysis: SOC as a negative moderator of the relationship between CBI and BDI.


At low moderation (SOC = 47.00) the conditional effect was 0.2564 [95% CI (0.2230, 0.2898), *p* < 0.05]. At middle moderation (SOC = 61.00) the conditional effect was 0.1855

[95% CI (0.1583, 0.2127), *p* < 0.05]. At high moderation (SOC = 74) the conditional effect was 0.1196 [95% CI (0.0864, 0.1529), *p* < 0.05] (Figure 2). These results identify SOC as a negative moderator of the relationship between CBI and BDI.

**Figure 2.** The moderation effect of SOC, between CBI and BDI relationship, at low (47) middle (61) and high (74) degree of SOC.
