Self-e cacy and fatigue among health care workers during COVID-19 outbreak: A moderated mediation model of posttraumatic stress disorder symptoms and negative coping

Tianya Hou (  liumi9512@126.com ) Centro Universitario de la Defensa en la Escuela Naval Militar de Marin https://orcid.org/0000-00017361-1935 Wei Dong Naval medical university Ruike Zhang Naval Medical University Xiangrui Song Naval Medical University Fan Zhang Naval Medical University Wenpeng Cai Naval Medical University Ying Liu Naval Medical University Guanghui Deng Naval Medical University


Background
On the last day of 2019, Coronavirus Disease 2019 (COVID- 19) with unknown etiology was rst reported in Wuhan, China [1]. On 30 January, 2020, the COVID-19 outbreak was declared a Public Health Emergency of International Concern [2]. The outbreak as a global health threat rapidly spread [3] and more than 1 million con rmed cases have been reported from almost every country. A near exponential growth in the number of new con rmed cases has been witnessed over the past few weeks [1]. The COVID-19 epidemic is straining health care systems with the escalating demand on health care workers (HCWs) and health facilities. The availability of local professional HCWs would largely determine whether the pandemic could be defeated [4] and it is of great importance to maintain and enhance the e ciency, quality and safety in the health sector amid COVID-19 outbreak. Work-related fatigue as a longstanding problem in health care settings was associated with reduced vigilance and poor work performance [5], which would enhance the incidence of medical errors and jeopardize work e ciency and quality. Owing to the shortage of HCWs and overwhelming number of con rmed cases during the COVID-19 pandemic, HCWs need to work overload and face numerous stressors, which makes them more vulnerable to experience fatigue [6]. According to the literature regarding the 2003 severe acute respiratory syndrome (SARS) pandemic, the prevalence of fatigue was 22.1-70.3% [7,8]. Hence, there is an urgent demand on investigating the in uential factors and underlying mechanisms of fatigue in order to design targeted interventions against fatigue.
Fatigue as a multidimensional state could be caused by numerous factors, which makes the identi cation of underlying mechanisms challenging [9]. A two-stage approach to manage fatigue has been proposed. The rst stage is to deal with treatable factors while the second stage is to address residual fatigue. Fatigue has been widely studied based on clinical samples and self-e cacy has been identi ed as one of the in uential nonpharmacological factors [10,11]. Self-e cacy is de ned as the belief of one's capacity to successfully accomplish speci c goals [12]. Numerous studies have suggested the enhancement of self-e cacy could be particularly effective in ameliorating fatigue [13] since it could protect against the adverse in uence of stressors [14]. However, there is a lack of knowledge regarding the association between self-e cacy and fatigue among HCWs during the pandemic. In addition, the mechanisms behind the association are not well-understood.
Apart from fatigue, HCWs are also vulnerable to develop post-traumatic stress disorder (PTSD), a psychiatry disorder caused by the witness or experience of traumatic events, due to the exposure to the threats of life, the witness of the death of patients and colleagues and the fear of being infected during the COVID-19 pandemic [15]. The most common PTSD symptoms are recurrent memory regarding traumatic events, avoidance and heightened arousal [16]. An extensive body of literature showed general self-e cacy was a signi cant predictor of PTSD symptom [17,18]. Previous literature employed an induction task with false feedback technique, through which participants were guided to believe they had low or high self-e cacy. After the induction, participants from high self-e cacy group presented better performance in problem solving [19]. More importantly, high self-e cacy participants showed less distress after the trauma lm paradigm in comparison to those from low self-e cacy group [20]. In addition, a recent study conducted by Titcombe-Parekh et al. suggested the increase in self-e cacy could impact neural circuits with respect to executive function and regulation of emotion, which further contribute to the decrease of PTSD symptoms [21]. Furthermore, another recent research based on a sample of civilian war victims reported hyper-arousal and active avoidance symptoms of PTSD mediated the relation between exposure to trauma and somatic symptom such as fatigue since the PTSD symptoms might lead to enhancing muscle tension, increasing alertness of pain and negative appraisals towards experience [22,23]. Thus, it is possible that PTSD symptoms mediated the association between self-e cacy and fatigue of HCWs during the outbreak.
Coping style is de ned as the thoughts or behaviors individual adopts to handle the adversity and stress [24], which is considered to be consistent over time [25]. An emerging body of studies provided evidence regarding the interaction effect of self-e cacy and coping style [26], which indicated that the effect of self-e cacy might be in uenced by coping style. Besides, according to the integrative framework of coping process, self-e cacy and coping style are interrelated to determine health outcome [27]. As proposed by Levin et al. [28], the association between self-e cacy and health outcome might be different depending on the coping strategies adopted. Nonetheless, it remains unexplored whether negative coping style plays a moderation in the effect of self-e cacy on PTSS and fatigue among HCWs during the COVID-19 epidemic. Also, the results found that avoidant coping moderated the effect of self-e cacy on health outcomes, whereas positive coping failed to moderate the association [28]. Moreover, previous literature presented consistent results regarding the relation between negative coping and health outcome and the inconsistent results concerning the effect of positive coping on health outcome since the effectiveness of positive coping is a more crucial determinant in the positive outcome [29,30]. Therefore, the current study would only focus on the moderating effect of negative coping as there was no method to measure the effectiveness of positive coping during the pandemic.
In sum, the current study aimed to explore the prevalence of fatigue among HCWs during the outbreak of COVID-19, investigate the mediating role of PTSD symptoms and the moderating role of negative coping in the association between self-e cacy and fatigue. Thus, we proposed a moderated mediation model (see Fig. 1) to address the hypotheses that PTSD symptoms might mediate the effect of self-e cacy on fatigue and negative coping might moderated the direct and/or indirect (self-e cacy -PTSD symptoms path) effect of self-e cacy on fatigue among HCWs during the COVID-19 pandemic.

Participants
This cross-sectional survey was performed in Anqing City, Anhui Province, China. The city borders Hubei province, the epicenter of the COVID-19 outbreak. All data were collected between March 13 -20, 2020, more than 2 months after the outbreak of COVID-19. Cluster sampling procedure was adopted to recruit a total of 528 HCWs through local Health Commission. The inclusion criteria were a) having participated in the ght against COVID-19, b) age > 18. Finally, 527 subjects were included in the analysis (effective response rate 99.8%). The study was approved by the research ethics committee of Naval Medical University. Before lling out the online questionnaires, informed written consent was obtained from each participating HCW. In order to protect HCWs privacy and encourage honest reporting, the questionnaires were nished anonymously. In addition, participants were told the participation was voluntary and they could withdraw at any time.

Measures
Self-e cacy The Chinese version of the General Self-E cacy Scale (GSES) developed by Zhang and Schwarzer [31] was used to measure self-e cacy. The scale consists of 10 items with only one dimension and each item is scored on a 4-point Likert scale from 1 (not true at all) to 4 (exactly true). The range of the total scores is 10-40, with higher scores indicating higher level of self-e cacy. The scale has been demonstrated with good construct validity, impressive test-retest reliability and excellent internal consistency in the Chinese samples [31,32]. In the present study, the Cronbach's Alpha for GSES was 0.900. PTSD symptoms PTSD symptoms were measured by PTSD Checklist-Civilian Version (PCL-C) [33]. The scale consists of 17 items with three subscales (re-experiencing, avoidance and hyperarousal). Each item is rated on a 5point Likert scale ranging from 1 (not at all) to 5 (extremely). The 17 items were summed to create a total score representing the severity of PTSD symptoms, with higher scores denoting more severe PTSD symptoms. The Chinese versions of the scale has presented high internal consistency and adequate convergent validity [34]. The Cronbach's Alpha in the present study for PCL-C was 0.963.

Negative coping
Negative was assessed by the negative coping subscale of Simpli ed Coping Style Questionnaire (CSCQ), which consists of 8 items [35]. Participants rated each item on a 4-point Likert scale ranging from 0 = never used to 3 = often used. The average score of the 8 items indicated the tendency to use negative coping. Higher scores of negative coping represents that the participants are more likely to use negative coping. The negative coping subscale has shown good internal consistency and test-retest reliability [35]. In the current study, the Cronbach's Alpha for CSCQ was 0.804.

Fatigue
The 14-item Fatigue Scale (FS-14) was employed to evaluate fatigue [36]. The scale includes physical and mental fatigue subscales with 8 and 6 items respectively. Each item describes a symptom which is relevant to fatigue. Participants rated each item with two responses: 0 (no symptom) and 1 (having symptoms). The total score is 0-14 points. According to the previous literature based on Chinses samples [37], a cut-off ≥7 indicated the caseness of fatigue. The Chinese version of the scale has been widely used in health care settings with good validity and reliability [6,38]. In the study, the Cronbach's Alpha for FS-14 was 0.907.

Covariates
In the current study, the covariates included age, gender, marital status, educational level, years of working and technical title. Age was grouped into 20-29 years, 30-39 years, 40-49 years and 50-59 years.
Marital status was divided into unmarried (single, divorced and widowed) and married. Educational level was categorized into two groups: high school or under and university or above. Years of working was divided into 10 years or less and more than 10 years. Technical title was classi ed into three groups: junior, intermediate and senior.

Statistical analysis
Firstly, we used descriptive analyses to describe demographic and working characteristics. Independent ttest and one-way analysis of variance (ANOVA) followed by LSD post hoc test were used to compare group differences in fatigue. Secondly, bivariate correlations between all the study variables (self-e cacy, PTSD symptoms, negative coping and fatigue) were calculated by Pearson's correlation analyses. Thirdly, the mediation effect was examined according to Mackinnon's four-step procedure [39]. Four conditions need to be met: (1) a signi cant association between self-e cacy and fatigue; (2) a signi cant relationship between self-e cacy and PTSD symptoms; (3) a signi cant relationship between PTSD symptoms and fatigue while controlling for self-e cacy; (4) a signi cant coe cient for the indirect association between self-e cacy and fatigue via PTSD symptoms. The last condition was examined by the bias-corrected percentile bootstrap method [40], which produced 95% bias-corrected con dence interval (CI) with 5000 replacements. The effect would be determined if 95% CI does not include 0. Hayes PROCESS macro (Model 4) [40] was employed to estimate parameters for the mediation effect.
Finally, the moderated mediation effect was examined by Model 8 [40]. As mentioned above, the effects were established if 95% the bias-corrected bootstrap CIs of the interaction excluded 0. Then, Johnson-Neyman technique [41] was employed to plot the conditional effects and con dence bands at different values of negative coping. In addition, z-scores for each variable were calculated before the analysis. Furthermore, all models were controlled for age, gender, marital status, education, years of working and technical title. All statistical analyses were performed by SPSS 25.0 and two-tailed P-values less than 0.05 were regarded as statistical signi cance.

Demographic and working characteristics and fatigue
The characteristics of the sample and the group comparisons on fatigue are presented in Table 1 The prevalence of fatigue among HCWs was 56.7% (FS-14 ≥ 7). There was signi cant difference in fatigue among different age groups (F = 3.176, P =0.024). LSD post hoc test indicated HCWs aged 30-39 years presented signi cant higher fatigue than those aged 20-29 years and those aged 50-59 (all P < 0.05). No signi cant differences were found in fatigue by gender, marital status, education, years of working and technical title (all P > 0.05).

<Table 2 was inserted here>
Mediating effect of PTSD symptoms The study assumed PTSD symptoms would mediate the relationship between self-e cacy and fatigue. We followed Mackinnon's four-step procedure to examine the mediation effect (see Table 3). Firstly, selfe cacy was signi cantly associated with fatigue (β = 0.40, P 0.001) (see Model 1 in Table 3). Secondly, self-e cacy was signi cantly related to PTSD symptoms (β = 0.30, P 0.001) (see Model 2 in Table 3). Thirdly, PTSD symptoms were signi cantly correlated with fatigue when we controlled for self-e cacy (β = 0.50, P 0.001) (see Model 3 in Table 3). Finally, the indirect effect of self-e cacy on fatigue via PTSD symptoms was signi cant (ab = -0.15, SE = 0.03, 95% CI = [-0.21, -0.10]). The mediation effect accounted for 37.7% of the total effect. In sum, all four criteria for mediation effect have been met and PTSS symptoms mediated the effect of self-e cacy on fatigue of HCWs during the COVID-19 pandemic.

Moderated Mediation effect analysis
The study anticipated negative coping might play as a moderator in the direct and indirect (the rst stage of the mediation pathway: self-e cacy -PTSD symptoms) effects of self-e cacy on fatigue. As presented in Table 4, the results of moderated mediation analysis showed the interaction of self-e cacy and negative coping had a signi cant effect on PTSD symptoms (β = -0.158, P 0.001), which indicated that the relation between self-e cacy and PTSD symptoms was moderated by negative coping. The moderated mediation effect was established since the indirect pathway was moderated by negative coping [40]. Additionally, negative coping also moderated the direct effect of self-e cacy on fatigue (β = 0.075, P 0.05). Table 4 also showed the conditional direct and indirect effects of self-e cacy on fatigue at different values of negative coping (1 SD below the mean, the mean, and 1SD above the mean). The direct effect of self-e cacy on fatigue was stronger at 1 SD below the mean of negative coping (β= -0.306, 95%CI: -0.391, -0.221) than 1SD above the mean (β= -0.157, 95%CI: -0.256, -0.058). As shown by Johnson-Neyman technique [41], negative coping would moderate the direct effect of self-e cacy on fatigue when the standard scores of negative coping were lower than 1.494, in which 95% CI did not contain zero (see Figure 2). <Figure 2 was inserted here> Nonetheless, the indirect effect of self-e cacy on fatigue was attenuated at 1 SD below the mean of negative coping (β= -0.090, 95%CI: -0.141, -0.051) in comparison to 1SD above the mean (β= -0.256, 95%CI: -0.332, -0.188). Johnson-Neyman technique presented that negative coping would moderate the association between self-e cacy and PTSD symptoms when the standard scores of negative coping were more than -1.401 as 95% CI did not include zero (see Figure 3).

Discussion
The research based on a sample of HCWs during the COVID-19 epidemic investigated the prevalence of fatigue and explored the potential mechanisms underlying the association between self-e cacy and fatigue with PTSD symptoms and negative coping as the mediator and moderator.
The prevalence of fatigue among HCWs was 56.7%, which is higher than the study during the outbreak of SARS with an overall incidence rate of fatigue among paramedics of 44% in Toronto [8]. This could be attributed to higher infectivity and rapider transmission of COVID-19 than SARS [42] with more people being infected and heavier workload for HCWs during the COVID-19 outbreak. Interestingly, when it comes to the comparison with the prevalence of fatigue among HCWs in the non-epidemic period, the results were inconsistent. Considerable studies found the lower prevalence of fatigue ranging from 21.6 to 45.5% [43][44][45][46], whereas Da Silva et al. [47] reported the overall incidence rate of fatigue among nursing workers in Brazil was 52%, which is in line with our ndings. Moreover, several research observed higher incidence rates of fatigue ranging from 83.7 to 91.9% [38,48,49]. The discrepancy might be explained by the different de nitions of fatigues, diverse assessment tools, inconsistent cut-off points and so forth. For instance, Cai et al. [50] employed a score of 4 on a 11-item fatigue scales as the cut-off point to de ne the occurrence of fatigue, while O'Donnell et al. [46] measured fatigue through only one question via selfassessment of average level of fatigue during the previous week. These differences might be attributed to the different prevalence of fatigue among HCWs. However, there is no doubt that fatigue is a commonly experienced symptom among HCWs during the COVID-19 outbreak and more attention should be paid to deal with this issue in order to maintain the work safety and e ciency in the health care settings.
Our results found there were signi cant differences in fatigue among different age groups. Speci cally, the 30-39 years group HCWs reported signi cant higher level of fatigue in comparison to 20-29 years and 50 − 49 years groups, which is congruent with the previous literature [45,50]. The older HCWs with richer working experience and stronger professional skills usually worked as group leaders to make decisions, whereas the 30-39 years group implemented the decisions with physical labor. In addition, the 30-39 years group HCWs took more responsibilities than the 20-29 years group since the younger HCWs might lack experience and su cient professional knowledge and could not complete the work alone. Therefore, those reasons might explain the differences. In the current study, there was no gender difference in fatigue, which is consistent with previous literature [38]. However, several previous studies claimed that women were more likely to suffer from fatigue [6,48]. The inconsistent results might be attributed to socio-economic status of HCWs. As Jenkins proposed [51], when controlling for socioeconomic backgrounds, the gender difference in case rates would disappear. In addition, one of the genders might be under-represented in some studies [52], which might partially explain the difference.
In consistent with our hypothesis, this study demonstrated a partially mediating role of PTSD symptoms in the association between self-e cacy and fatigue, which indicated the potential mechanisms regarding how self-e cacy would indirectly affect fatigue. HCWs with low self-e cacy could not only directly contribute to higher level of fatigue, but also indirectly aggravate fatigue via PTSD symptoms. This is consistent with the previous literature regarding the protective role of self-e cacy in PTSD symptoms and the positive association between PTSD symptoms and fatigue [17,22]. This study extended the previous literature by combining self-e cacy as a protective factor and PTSD symptoms as a risk factor to explore fatigue, which has profound implications for the prevention and mitigation of fatigue of HCWs. The self-e cacy-based program and intervention for PTSD could be designed to reduce the occurrence of fatigue during the COVID-19 pandemic, which might further decrease the medical errors and enhance the work quality.
More importantly, the moderated mediation analysis presented negative coping as a relatively stable trait could moderate the direct and indirect effects of self-e cacy on fatigue of HCWs. This is in line with the integrative framework of coping behaviors and the previous study [27,28], which suggested the moderating role of coping style in the link between self-e cacy and health outcomes. To the best of our knowledge, this is the rst study to explore such moderated mediation effects. As revealed by Johnson-Neyman technique, it is noteworthy that with increasing negative coping, the direct effect of self-e cacy on fatigue became weakened. When the standard score of negative coping enhanced to 1.49 and over, the direct association was not signi cant any more. In contrast, the indirect effect of self-e cacy on fatigue via PTSD symptoms strengthened as the level of negative coping increased. Likewise, Johnson-Neyman technique showed that the effect of self-e cacy on PTSD symptoms had no statistical signi cance when the standard score of negative coping was − 1.40 and lower. This adds to our understandings of fatigue with important practical implications. Interventions for PTSD should be prioritized for HCWs with higher levels of negative coping as self-e cacy would be more likely to in uence fatigue through PTSD symptoms.
Several limitations should be addressed. Firstly, this cross-sectional study failed to infer the causal relationship. The longitudinal or experimental studies should be conducted to further explore the relation. Secondly, the data were obtained through the self-report questionnaires, which might cause self-reported biases. Further study could collect data from diverse informants. Thirdly, the participants of our study were only from Anqing City, which might limit the generalization of the results to other areas. Further study would recruit subjects from diverse regions. Finally, as mentioned above, fatigue could be in uenced by many factors [9]. Our model could just explain part of the variance. A more integrative model is suggested for future study.

Conclusions
The prevalence of fatigue among HCWs during the COVID-19 outbreak was 56.7% in the study. PTSD symptoms partially mediated the effect of self-e cacy on fatigue. In addition, both the direct effect of self-e cacy on fatigue and the mediating effect of PTSD symptoms were moderated by negative coping. Speci cally, the direct effect was weaker and the indirect effect was stronger for HCWs with higher level of negative coping. For HCWs who participated in ghting COVID-19, especially those with higher level of negative coping, it might be of vital importance to design program combining the improvement of selfe cacy and interventions for PTSD to reduce fatigue. Availability of data and materials

Abbreviations
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests Funding This research received no speci c grant from any funding agency in the public, commercial, or not-forpro t sectors.
Authors' contributions TH and WD participated in conception, design of the work, data interpretation and analysis, drafting the manuscript. RZ, XS and FZ participated in the acquisition and interpretation of data. WC and YL participated in the data analysis and all authors revised the draft. TH and GD made contribution to the concept and design of the study, acquisition of data, manuscript revision and supervision. All authors approved this nal version to be published.  Table 3 Mediation analysis (N=527). Note: All models are adjusted for age, gender, marital status, education, years of working and technical title. *** P 0.001 Table 4 Conditional process analysis (N=527).  The conditional direct effect of self-e cacy on fatigue at the values of negative coping.

Figure 3
The conditional effect of self-e cacy on PTSD symptoms at the values of negative coping.