Association between probable posttraumatic stress disorder and quality of life in emergency service workers: a mediation analysis using a third potential model with both in parallel and in series effects; a cross-sectional study

Background The present study used a third potential model with both in parallel and in series effects to identify mediators in�uencing the association between probable posttraumatic stress disorder (PTSD) and quality of life (QOL). In addition, we sought to clarify how stress-coping behaviors in�uence components of QOL. Methods Participants were 220 emergency service workers. PTSD was assessed using the Impact of Event Scale-Revised (IES-R). QOL was assessed using the MOS 36-Item Short-Form Health Survey (SF-36). Mediation analysis and moderated mediation analysis of interaction effects with a third potential model were used to examine the effects of stress-coping (Brief COPE scores), social support (Multidimensional Scale of Perceived Social Support [MSPSS] scores), and resilience (Connor-Davidson Resilience Scale [CD-RISC] scores) on the association between probable PTSD and QOL.


Background
Emergency service workers, such as re ghters, emergency medical technicians (EMT), and military personnel, tend to have highly stressful occupations because they are commonly faced with fatalities in accidents or disasters [1][2][3] and dangerous environments (e.g., extremely high temperatures, poisonous dust, re, and collapsing buildings) [4][5][6][7].Previous studies have reported that emergency service workers often experience stress symptoms and negative mental health outcomes [8][9][10][11].There is evidence that the stressors experienced by emergency service workers can cause mental disorders such as depression, anxiety, and posttraumatic stress disorder (PTSD) [2,12,13].Factors that have been proposed to affect PTSD symptoms include age, sex, education, length of employment in emergency service work, lack of continuous psychological care, exposure to human remains, stress-coping, social support, and resilience [14][15][16][17][18][19][20][21].The above factors may in uence the association between quality of life (QOL) and PTSD among rescue workers.A previous study investigating the harmful effects of PTSD on re ghters indicated that reduced QOL was associated with PTSD [22].QOL is an important concept for understanding the mental and physical health of workers.QOL is de ned by the World Health Organization as "individuals' perceptions of their position in life in the context of the culture and value systems in which they live and in relation to their goals, expectations, standards and concerns" [23].Most previous studies of QOL have been based on this de nition [24].There has been a substantial amount of research on health-related QOL, and several studies on QOL among emergency service workers have indicated the importance of assessing QOL.Schnurr et al. reported that war veterans with mental disorders exhibited lower QOL [25].Another study reported that non-emergency service worker retirees exhibited better mental health scores on a QOL measure compared with emergency service worker retirees [26].A similar study by Bartlett et al. reported that emergency service workers experienced sleep disturbance caused by their work [27], and that QOL was related to sleep disturbance [22].In addition, some research suggests that stress-coping, social support, and resilience are associated with QOL.Nachar et al. reported that stress-coping and social support predicted QOL in a clinical sample with PTSD [28].Connor et al. argued that resilience is an important aspect of mental health state [29], and some studies have reported that resilience affects QOL in participants with mental disorders [30][31][32].
Owing to job-related stressors, emergency service workers may be particularly likely to experience mental disorders affecting their daily lives.To the best of our knowledge, few previous studies have simultaneously examined the relationships between stress-coping, social support, and resilience related to QOL in emergency service workers who rescue victims from res, tra c accidents, and earthquakes.There are likely to be complex associations between QOL and other factors.Therefore, further research is needed to understand how various factors affect QOL in emergency service workers.
Previous studies have indicated that stress-coping, social support, and resilience mediate QOL in re ghters, clinical patients, and disaster victims with PTSD, respectively [28,33,34].The factors of stress-coping, social support, and resilience are also related, and there is evidence that resilience is associated with stress-coping and social support [35,36].Based on ndings from a previous study, we hypothesized the existence of several associations between factors of probable PTSD and QOL, as shown in Fig. 1 [28, [33][34][35][36].We hypothesized that probable PTSD would be mediated in parallel by stress-coping and social support, and, in turn, these factors would be mediated in series by resilience to the effects on QOL among emergency service workers.
In Japan, because there is little mental health support for emergency service workers, most workers are required to individually manage any mental distress that arises from their work [37].Therefore, it is necessary to identify methods to mitigate the lower QOL associated with PTSD among emergency service workers in Japan.The current study aimed to identify the mediating effects of stress-coping, social support, and resilience (and other confounding factors) on QOL scores using models with in parallel or in series effects.Speci cally, we sought to examine each subscale in stress-coping to elucidate whether different types of stress-coping behaviors exert different in uences on QOL among rescue workers.

Aim, design, and setting of the study
The present study aimed to identify mediators in uencing the association between probable PTSD and QOL and to determine how stress-coping behaviors in uence components of QOL in a mediation model.
A multi-center cross-sectional survey was conducted at re stations in Japan.Participants were between 20 and 65 years old.All participants were re ghters or emergency medical technicians.A questionnaire was sent to 373 participants, 220 of whom returned completed scales (response rate = 58.9%).This research was carried out with the permission of the re departments involved.Recruitment started in April 2015 and was completed in August 2015.Ethical approval was granted by the Clinical Research Ethics Committee of Chiba University, Japan (approval number 1731).All participants provided informed consent by post.

Predictor variables:
The Impact of Event Scale-Revised (IES-R) scale is the most commonly used measure for screening emergency service workers for mental health problems [38][39][40][41][42].The IES-R has a three-factor structure (eight items assessing intrusion, eight items assessing avoidance, and six items assessing hyperarousal).The scale was developed by Weiss to assess probable PTSD among individuals suffering from severe stressors [41].A previous study indicated that the intrusion and hyperarousal subscales exhibited a good t with the Diagnostic and Statistical Manual of Mental Disorders Fourth Edition criteria developed by the American Psychiatric Association, whereas the avoidance subscale did not [43].Importantly, the measure can distinguish between individuals with or without PTSD [43].

Mediating variables:
The Brief COPE inventory measures how often individuals use different coping strategies to handle di cult situations [44].The Brief COPE inventory comprises 14 factors with two items each (Self-distraction, Active coping, Denial, Substance use, Using emotional support, Using instrumental support, Behavioral disengagement, Venting, Positive reframing, Planning, Humor, Acceptance, Religion, and Self-blame).Responses are made using a four-point rating scale (1 = "I haven't been doing this at all", 2 = "I've been doing this a little bit", 3 = "I've been doing this a moderate amount", 4 = "I've been doing this a lot").Higher scores indicate the use of many coping methods.We included each subscale of the scale in the analyses.

Outcome variables:
The MOS 36-Item Short-Form Health Survey (SF-36) [46] is frequently used to screen participants for health-related QOL issues.The SF-36 contains eight factors (physical functioning, role limitations due to physical health, role limitations due to emotional problems, vitality, mental health, social functioning, bodily pain, and general health) and two composite scores (a physical component score and a mental component score).The SF-36 measures general health status, and high scores indicate better QOL.A previous study analyzed the validity of the Japanese version of the SF-36, revealing three components: the Mental Component Summary (MCS), Physical Component Summary (PCS), and Role/social Component Summary (RCS) [47].To summarize the items for the three components, we used an online scoring service for the SF-36.

Analysis
We conducted three steps of analysis.First, descriptive data for participant characteristics and measures were compared using the Wilcoxon rank-sum test or chi-square test.Correlations between scores on each measure were analyzed using Spearman's correlation analysis.In this analysis method, values of 0.90-  [47].Second, to examine the univariate associations [48,49], we analyzed the associations between scores on the IES-R, the Brief COPE, the MSPSS, the CD-RISC, and the SF-36, using a linear regression model.
Finally, we conducted a moderated mediation analysis using a multiple mediator model to examine the mediation pathways as a path model for the above measures.In the moderated mediation analysis, factors such as age and length of employment are potential moderator variables if the interaction effect (independent variable × moderator variable) is signi cant [50].In the current study, the moderators were not signi cant.Therefore, the confounding factors controlled in the mediation analysis were: age, length of employment, and education.Partial mediation was assumed if the mediation path model produced a signi cant direct effect [51].Complete mediation was assumed if the mediation path model did not produce a signi cant direct effect [51].The total effect is the sum of the direct effect and the total indirect effect.That is, the effect of PTSD (IES-R scores) on QOL (SF-36 scores) was mediated by coping (Brief COPE scores), support (MSPSS scores), and resilience (CD-RISC scores).A Bayesian inference method was used for the mediation analysis.The Bayesian method of estimation is suitable for use with small samples and non-normal distributions [52], and the results can be analyzed using new computational algorithms called Markov chain Monte Carlo algorithms [53].Model t indices were evaluated using the posterior predictive p-value (PPP), the comparative t index (CFI), the Tucker-Lewis index (TLI), and the root mean square error of approximation (RMSEA).A low PPP value is considered to indicate a poor model t [53,54].Following Cain et al., a cut-off value of PPP < .10 was used to determine whether the model should be rejected [55].The model t indices of CFI and TLI values of > 0.95 indicated an acceptable t to the data, and an RMSEA value < 0.08 indicated an acceptable t [56,57].
The normality of variables was assessed using probability-probability plots and the Shapiro-Wilk test, and the homogeneity of variance was assessed using Levene's test.As the normality and homogeneity of variance assumptions were violated, non-parametric statistical tests were used.
Missing values were handled using the full information maximum likelihood (FIML) method.This approach presupposes that data were missing completely at random or missing at random.Thus, if data were missing in a non-random way, the results might be misinterpreted.To address this issue, we conducted a sensitivity analysis using re-examined data with a list-wise deletion method [58].We conducted the linear regression, moderation, and mediation analysis using Mplus, version 8.4 (Muthén & Muthén, Los Angeles, CA, USA).Other analyses were conducted using Stata, version 13 (Stata Corp LP, College Station, TX, USA).

Participants' characteristics
This study used a cross-sectional design.Questionnaires were sent to 373 participants, 220 of whom returned responses.All participants were male.Less than 2% of participants had missing data on the SF-36 and the CD-RISC, and less than 10% had missing data on any of the confounding factors.Table 1 shows participant demographic characteristics by IES-R scores.The IES-R scores indicated that a quarter of participants had probable PTSD (IES-R score ≥ 25).There were signi cant differences in probable PTSD indicated by CD-RISC scores, subscales of the Brief COPE scale (Self-distraction, Denial, Emotional support, Behavioral disengagement, Religion, and Self-blame), and MCS in SF-36 scores.Half of the participants were between 30 and 39 years of age.Most participants had worked at the re station for longer than 10 years.

Linear regression models of participant characteristics adjusted for confounding factors
Table 3 shows the results of linear regression analyses between IES-R scores, MCS scores in SF-36, subscale scores in the Brief COPE, and MSPSS scores.The  CD-RISC scores showed a signi cant positive association with MCS scores in SF-36 (β = 0.167, 95% CI = 0.100 to 0.234, p < .001).

Moderated mediation
In the mediation analysis without moderators and covariates, four models passed through all measures as the scores of IES-R, Brief COPE, MSPSS, CD-RISC, and SF-36, but directly, not through the pathway from IES-R to CD-RISC.These ndings indicated acceptable model ts and that all direct paths were signi cant.The model t indices of the four models were as follows: Active coping: PPP = .276,RMSEA = 0.076, CFI = 0.963, TLI = 0.938; Behavioral disengagement: PPP = .346,RMSEA = 0.059, CFI = 0.982, TLI = 0.943; Venting: PPP = .311,RMSEA = 0.067, CFI = 0.968, TLI = 0.897; Positive reframing: PPP = .240,RMSEA = 0.085, CFI = 0.961, TLI = 0.877 (not shown in Table ).Moderated mediation analysis based on the above models yielded non-signi cant results, indicating that there were no signi cant differences in the moderators, including age and length of employment.Therefore, age and length of employment were confounding factors controlled in the mediation analysis.
Mediation analysis with a multiple mediator model    Supplementary Table 2 shows sensitivity analysis using data with a list-wise deletion method.Model ts for the three subscale scores in the Brief COPE comparing the FIML method and the list-wise deletion method had similar values, and sensitivity analysis revealed acceptable model t (Active coping: PPP = .209,RMSEA = 0.086, CFI = 0.982, TLI = 0.919; Behavioral disengagement: PPP = .280,RMSEA = 0.073, CFI = 0.987, TLI = 0.945; Venting; PPP = .236,RMSEA = 0.083, CFI = 0.982, TLI = 0.923).Therefore, the assumption of not missing at random (NMAR) was rejected.However, the model t for Positive reframing on the subscale scores in the Brief COPE indicated that a good model t was not retained (PPP = .141,RMSEA = 0.100, CFI = 0.976, TLI = 0.894).For the above reason, the missing data in Positive reframing might re ect NMAR.For β coe cient values, the three subscale scores in the Brief COPE (Active coping, Behavioral disengagement, and Positive reframing) using the list-wise deletion method exhibited similar values to the results of the FIML method.However, the 95% CIs of the β coe cient values between IES-R scores and Venting on scores in the Brief COPE included zero, and were not signi cant.It is possible that the assumption of NMAR for Venting score data was valid for the present analysis.

Discussion
The present study aimed to identify mediators in a third potential model to determine whether stress-coping, social support, and resilience mediate the association between probable PTSD and QOL among re ghters and emergency medical technicians working in emergency services.In addition, to analyze the associations between measures mentioned above, we sought to determine how behavior in stress-coping in uences components of QOL.The results supported our initial hypothesis, revealing that probable PTSD directly and indirectly predicted mental health state and QOL in models through three mediators.

Effects of independent variables
A linear regression analysis of the univariate showed associations between probable PTSD, QOL, stress-coping, social support, and resilience.Greater probable PTSD directly predicted lower mental health and role/social functioning in QOL.A previous study indicated that war veterans with PTSD exhibited impaired social functioning in QOL [59], potentially developing impairments in their social lives.Such cases typically involve a cost to society, associated with healthcare utilization [60].This contributes to the importance of identifying effective methods for protecting individuals from developing PTSD and improving QOL.
The current results revealed that the association between PTSD and QOL involved many confounding factors and independent variables, such as stresscoping and social support.In the present study, lower social support was predicted by greater PTSD.Therefore, social support is likely to protect against worsening PTSD [33].Previous studies indicated that, among war veterans with PTSD, lower family support was associated with PTSD [68], and poorer social functioning was exhibited compared with war veterans without PTSD [33].Importantly, social support might promote improvements in QOL among rescue workers.
Greater resilience in the present study was predicted to occur when more adaptive coping strategies were employed, including Active coping, Venting, and Positive reframing, and when avoidance strategies were employed less, including Behavioral disengagement and in stress-coping, in accord with previous studies [35,69].In addition, greater resilience exhibited an association with better mental health state and QOL but not probable PTSD.In the current study, we did not nd the association between PTSD and resilience reported in a previous study of patients in the Emergency Department, which found that lower resilience predicted the impact of PTSD [35].Based on the discrepancy between this previous nding and the current results, we propose that probable PTSD may indirectly predict mental health on QOL-mediated stress-coping and social functioning via resilience.
The main ndings of the multiple mediation analysis The results of the mediation analysis without covariates indicated that four stress-coping strategies (Active coping, Behavioral disengagement, Venting, and Positive reframing), social support, and resilience partially mediated the association between probable PTSD and mental health in QOL.The p-values for the association between probable PTSD and resilience in the linear regression model and mediator model analysis were not signi cant.These ndings suggest that resilience was not directly associated with probable PTSD.In contrast, the association between probable PTSD and QOL in the multiple mediation model, as a partial model, was signi cant.
In a mediator model with covariates, all of the mediator variables predicted QOL, except for stress-coping.Mediator models using two coping strategies (Active coping and Positive reframing) supported our research hypothesis.The ndings of the present multiple mediation analysis in the two mediator models revealed that probable PTSD directly predicted stress-coping and social support, which were indirectly associated with QOL when mediated by resilience.In the present study, resilience, as a third mediator, was indicated to have an association with stress-coping and social support.Previous studies indicated a positive association between stress-coping and resilience [35], revealing that re ghters reported that receiving social support from a boss in uenced workers' resilience [70].In the current study, we found a signi cant positive association between resilience and QOL among emergency service workers.Therefore, both stress-coping and social support through resilience appear to help prevent PTSD and maintain good QOL.Thus, Active coping and social support strengthen resilience and improve mental health and QOL among emergency service workers with PTSD.
However, it is possible that Positive reframing in coping strategies skewed the results of the FIML method.Speci cally, Positive reframing might lead to type I errors.Therefore, the main ndings of the current study suggested that only the mediation model using Active coping exerted protective effects against PTSD and improved mental health in QOL.Active coping to confront di culties was found to be a particularly important behavior, revealing that this behavior requires practice with social support for the behavior to maintain mental health and QOL among emergency service workers.One method of Active coping is talking about the feelings associated with anxiety, and to receive acceptance from another person as a form of social support.
In addition, mental health screening in emergency service workers at risk from PTSD should include an examination of resilience.If, as indicated in the current study, stress-coping is sometimes negatively associated with QOL, failure to take resilience into account could produce an inaccurate assessment of mental health, as resilience might improve QOL.Although the present ndings are suggestive, because this was a cross-sectional study, the causal relationships between PTSD and QOL (mediated by stress-coping, social support, and resilience) could not be con rmed.
This study involved several limitations that should be considered.First, the study sample was composed of only male participants.It is possible that the observed association between PTSD symptoms and QOL would have differed if sex had been included as a confounding factor.Previous studies have also failed to examine the effect of sex; for example, a study by Schnurr et al. on PTSD and QOL included only male veterans [25].Second, because the current study used a cross-sectional design, we could not con rm a causal relationship between QOL and probable PTSD, or the direction of such a relationship.Third, the use of retrospective data about events that occurred in the past may have led to recall bias, distorting the results.Fourth, statistical analysis using FIML can involve bias when data are not missing at random, because it is based on the assumption of missing at random or missing completely at random [71].
Comparing values between the FIML method and the list-wise deletion method revealed that model t indices for Positive reframing indicated not to retain good model t, and the β coe cient on Venting in Brief COPE scores included zero in 95% CIs.The use of these two coping strategies might have confounded the results.
Despite these limitations, the current study had several strengths.To the best of our knowledge, this is the rst study to assess the association between PTSD and QOL mediated by stress-coping, social support and resilience among emergency service workers.The mediation analysis indicated associations between stress-coping, social support, and resilience.The current results suggested that stress-coping and social support may strengthen resilience, and resilience may prevent a decline in QOL.The next stage of our research will be to conduct an in-depth longitudinal study to determine whether stress-coping, social support, and resilience in uence mental health.

Conclusions
Figures

Figure 1 Mediation
Figure 1 Mediation model as a path model for the effect of PTSD (IES-R) on QOL (SF-36) mediated by coping (Brief COPE), support (MSPSS), and resilience (CD-RISC).Abbreviations: PTSD, posttraumatic stress disorder; QOL, quality of life; IES-R, the Impact of Event Scale-Revised; MSPSS, Multidimensional Scale of Perceived Social Support; CD-RISC, Connor-Davidson Resilience Scale; SF-36, the MOS 36-Item Short-Form Health Survey; MCS, Mental Component Summary; PCS, Physical Component Summary; RCS, Role/social Component Summary.The Brief COPE used 14 subscales: Self-distraction, Active coping, Denial, Substance use, Emotional support, Instrumental support, Behavioral disengagement, Venting, Positive reframing, Planning, Humor, Acceptance, Religion, and Self-blame.The SF-36 used three subscales: MCS, PCS, and RCS.
1.00 indicate a very high correlation, values of 0.70-0.90indicate a high correlation, values of 0.50-0.70indicate a moderate correlation, values of 0.30-0.50indicate a low correlation, and values of 0.00-0.30indicate a very low correlation or no correlation

Table 1
Demographic characteristics of Japanese emergency service workers

Table 3
Results of linear regression showing associations between IES-R and subscale scores in Brief COPE and SF-36 scores Abbreviations: IES-R, the Impact of Event Scale-Revised; SF-36, the MOS 36-Item Short-Form Health Survey; MCS, Mental Component Summary; PCS, Physical Component Summary; RCS, Role/social Component Summary; MSPSS, Multidimensional Scale of Perceived Social Support; CD-RISC, Connor-Davidson Resilience Scale; 95% CI, 95% con dence interval.Covariates adjusted for age, length of employment, and education.βsarenot standardized.*p<.05;**p<.01;***p<.001.Supplementary Table1presents the results of the linear regression model showing associations between subscale scores of the Brief COPE, CD-RISC scores, and SF-36 scores.There were signi cant positive associations between CD-RISC scores, subscale scores on the Brief COPE and MSPSS scores.CD-RISC scores were signi cantly positively associated with three subscale scores on Brief COPE (Active coping: β = 3.

Table 4 and
Table5show the results of four mediation models using a multiple mediator adjusted for confounding factors, including all scores on the IES-R, four subscale scores in the Brief COPE, MSPSS scores, and MCS scores in the SF-36.All of the model t indices in the mediation model indicated an acceptable model t.All of the direct effects were signi cant, but the speci c indirect effects of Behavioral disengagement and Venting were not signi cant.The direct effects between IES-R scores and MCS scores in SF-36 scores ranged from − 0.175 to − 0.174, indicating partial mediation.Direct effects in the mediation analysis revealed the following associations between measures: IES-R scores were signi cantly positively associated with four subscale scores in Brief COPE (Active coping: β = 0.020, p < .01;Behavioral disengagement: β = 0.038, p < .001;Venting: β = 0.015, p < .05;Positive reframing: β = 0.021, p < .01)and signi cantly negatively associated with MSPSS scores (βs = − 0.187, p < .01 to .001);CD-RISC scores were signi cantly positively associated with Active coping (β = 3.206, p < .001),Venting (β = 2.244, p < .01),and Positive reframing (β = 4.783, p < .001) on subscale scores in Brief COPE scores.However, Abbreviations: PTSD, posttraumatic stress disorder; QOL, quality of life; IES-R, the Impact of Event Scale-Revised; MSPSS, Multidimensional Scale of Perceive 36, the MOS 36-Item Short-Form Health Survey; MCS, Mental Component Summary; PPP, the posterior predictive p-value; RMSEA, the root mean square error o Lewis index; 95% CI, 95% con dence interval; SD, Standard Deviation.*p< .05;**p< .01;***p< .001. a Subscales in Brief COPE: Active coping, Behavioral disengagement, Venting, and Positive reframing.βsare not standardized.Covariates adjusted for age, length of employment, and education.

Table 5
Speci c indirect effects, total indirect effect, and total effect in mediation analysis PTSD, posttraumatic stress disorder; QOL, quality of life; IES-R, the Impact of Event Scale-Revised; MSPSS, Multidimensional Scale of Perceive SF-36, the MOS 36-Item Short-Form Health Survey; MCS, Mental Component Summary; PPP, the posterior predictive p-value; RMSEA, the root mean square err Tucker-Lewis index; 95% CI 95% con dence interval; SD, Standard Deviation..
The identi ed factors were consistent with the ndings of previous studies of re ghters and war veterans with PTSD, which indicated that greater stress-coping is predicted by PTSD [61, 62], and Active coping, Positive reframing, and Venting in the Brief COPE can effectively indicate improvement of PTSD symptoms [61, 63, 64].Most of the stress-coping strategies in the present study directly predicted mental health state in QOL, and more maladaptive coping strategies, including Self-distraction, Self-blame, and Behavioral disengagement, indicated worse mental health and QOL.Holubova et al.reported that participants with mental disorders tend to engage in negative rather than positive coping (adaptive) strategies, and that negative coping (maladaptive) strategies are likely to be associated with lower QOL[65].Maladaptive coping, which negatively affects QOL, involves Self-distraction, Substance use, Behavioral disengagement, Denial, Self-blame, and Venting.In contrast, adaptive coping, which improves QOL, involves Acceptance, Active coping, Planning, Positive reframing, Humor, Religion, the Use of emotional support, and the Use of instrumental support [65-67].It is possible that Japanese emergency service workers tend to engage in maladaptive coping, and stress-coping might worsen QOL in rescue workers with PTSD.