2.3.4. Occupational Stress

Occupational stress was assessed using the Occupational Stress Indicator (OSI; [58,59]). Two scales were taken into account: Sources of Stress and Effects on Health. The former is composed of 61 items distributed into six subscales: Job Factor (JF; 9 items; e.g., "*Having too much work to do*"), Managerial Factor (MF; 11 items; e.g., "*Having personal beliefs in contrast with those of the company*"), Relationships with Others Factor (RF; 10 items; e.g., "*Little encouragement from supervisors*"), Career Factor (CF; 9 items; i.e., "*Holding a position under your ability*"), Home–Work Interface Factor (IF; 11 items; e.g., "*Inability to stop working when you are at home*"), and Organizational Structure Factor (OF; 11 items; e.g., "*Luck of information and involvement in decisions*"). The latter is composed of two subscales, examining the Effects on Health from two perspectives: Psychological (PSY; 18 items; e.g., "*During a working day, do you feel irritated or agitated, though a clear reason does not always seem to be?*") and Physical (PHY; 12 items; e.g., "*Inability to fall asleep or sleep without interruption*") Effects. Internal reliability was excellent for each subscale, with Cronbach's alpha ranging from 0.81 to 0.92.

#### *2.4. Statistical Analyses*

Descriptive statistical analyses were used to analyze demographic data. Prior to conducting the main analyses, MANOVAs were performed to evaluate whether any significant statistical differences were estimated on the study variables according to gender differences. Mediation analyses were applied to verify whether social support functions as a buffer in the relationship between EI and occupational stress during COVID-19. The process involved examining path a, the association between EI (IV) and social support (M); path b, the impact of social support (M) on occupational stress (DV); and path c and c', the total and direct effect of EI (IV) on occupational stress (DV). The three sources of social support (family, friends, and significant others) were considered and included in the model as three distinct mediators. Before testing the mediating model, the multivariate normality distribution of data was first examined through the Mahalanobis distance computation. Since the Mardia's coefficient (192.47) exceeded the critical value associated with twelvedegrees-of-freedom (168), the assumption of multivariate normality was not met. Therefore, we chose to apply the bootstrapping (percentiles) method, a non-parametric resampling procedure recognized as a robust and accurate method for mediation analysis [60] and the best-suited technique to perform when the multivariate normality is violated. IBM SPSS (version 20) and Jamovi (version 1.6.23) were used for the analyses.
