Direct and Indirect Effects of Five Factor Personality and Gender on Depressive Symptoms Mediated by Perceived Stress

This study was designed to investigate associations among five factor personality traits, perceived stress, and depressive symptoms and to examine the roles of personality and perceived stress in the relationship between gender and depressive symptoms. The participants (N = 3,950) were part of a cohort study for health screening and examination at the Kangbuk Samsung Hospital. Personality was measured with the Revised NEO Personality Inventory (NEO-PI-R). Depressive symptoms were assessed using the Center for Epidemiologic Studies Depression Scale (CES-D). Perceived stress level was evaluated with a self-reported stress questionnaire developed for the Korea National Health and Nutrition Examination Survey. A higher degree of neuroticism and lower degrees of extraversion, agreeableness, and conscientiousness were significantly associated with greater perceived stress and depressive symptoms. Neuroticism and extraversion had significant direct and indirect effects (via stress as a mediator) on depressive symptoms in both genders. Agreeableness and conscientiousness had indirect effects on depression symptoms in both genders. Multiple mediation models were used to examine the mediational roles of each personality factor and perceived stress in the link between gender and depressive symptoms. Four of the personality factors (except openness) were significant mediators, along with stress, on the relationship between gender and depressive symptoms. Our findings suggest that the links between personality factors and depressive symptoms are mediated by perceived stress. As such, personality is an important factor to consider when examining the link between gender and depression.


Introduction
We hypothesized that personality traits would affect depressive symptoms by acting as vulnerability factors of perceived stress after experiencing negative events and that the effects of personality on perceived stress might mediate the association between gender and depressive symptoms. We tested two specific aims in the present study. First, we examined associations of each of the five personality factors and depressive symptoms through perceived stress. Secondly, we examined the mediating roles of each personality factor and perceived stress in the relationship between gender and depressive symptoms.

Materials and Methods Participants
The participants in this study came from a cohort that underwent health screening and examination at the Kangbuk Samsung Hospital in Seoul, South Korea. All subjects provided information about socio-demographic characteristics, dietary and health-related behaviors, and family and medical history of a wide range of diseases. Among the 4225 participants aged 19-69, individuals with a history of psychiatric problems (N = 183) or cancer (N = 97) were excluded in the current study. Participants completing all measures of personality, stress, and depression were included, for a total of 3,950 participants (men, N = 1,134; women N = 2,816) were finally included. The cohort's characteristics were previously described in detail [27].

Statement of Ethics
This study was approved by the Institutional Review Board of Ewha Womans University Mokdong Hospital. We obtained written informed consent from all participants. All applicable institutional and governmental regulations concerning the ethical use of human volunteers were followed during the research.

Measures
Depressive symptoms were assessed using the Center for Epidemiologic Studies Depression Scale (CES-D), which consists of self-reported responses to a 20-item questionnaire. The CES-D measures current depressive symptoms during the past week with a 4-point Likert scale. The total depressive score is the summation of all scores and ranges from 0 to 60. Stress level was evaluated by a self-reported stress questionnaire developed for the Korea National Health and Nutrition Examination Survey [28]. The questionnaire for perceived stress consisted of one item inquiring about the categories of stressful life events that had been experienced during the past month and nine items inquiring about stress levels due to those events. Stress levels were rated on a 5-point Likert scale (0: never; 4: very often). Total stress scores ranged from 0 to 36, which was the summed score of all items. The Cronbach's alpha coefficients for CES-D and stress questionnaires were 0.996 and 0.980, respectively. Missing responses for CES-D (3.91%) and stress questionnaires (3.46%) were imputed with multiple imputation methods using "mice" library in R [29]. We used the predictor variables including demographic factors such as age, marriage, education, and employment status as candidate risk factors of depression according to the previous literature [30]. We also added neuroticism as a predictor variable to impute for CES-D and stress score because neuroticism has strong associations with these variables.
The Revised NEO Personality Inventory (NEO-PI-R) was used to measure each of the FFM personality domains (neuroticism, extraversion, openness to experience, agreeableness, and conscientiousness). The Korean version of the Revised NEO Personality Inventory showed good reliability and validity [31]. A robust factor structure has been replicated with this instrument in various cultures including Korea [32]. The instrument is comprised of items about personal characteristics rated on a 5-point Likert scale. Each of the personality factors consists of the six facets that define each factor. The neuroticism factor includes anxiety, anger, depression, self-consciousness, impulsiveness, and vulnerability. The extraversion factor includes warmth, gregariousness, assertiveness, activity, excitement-seeking, and positive emotions. Openness to experience includes fantasy, aesthetic feelings, actions, ideas, and values. Agreeableness includes trust, straightforwardness, altruism, compliance, modesty, and tendermindedness. Conscientiousness includes competence, order, dutifulness, achievement-striving, self-discipline, and deliberation. The Cronbach's alpha coefficients for neuroticism, extraversion, openness, agreeableness, and conscientiousness were 0.908, 0.905, 0.809, 0.820, and 0.906, respectively. Raw scores were converted to T-scores (mean = 50, standard deviation = 10) using Korean combined-sex norms (N = 7418).

Statistical Analysis
Pearson's chi-square tests and independent sample t-tests were conducted to examine gender differences in demographic characteristics, stress, depression, and personality scores. Cohen's d was calculated as an effect size for t-tests. Pearson correlation analysis was performed to determine the association among the study variables. SAS 9.3 was used for all statistical analyses.
The mediation effect is referred to as the indirect effect of the independent variable (IV) on the dependent variable (DV) through the intervening variables. The mediation effect occurs when the correlation between the IV and DV is eliminated (complete mediation) or reduced (partial mediation) when the mediator is controlled for. To examine the effect of personality on depressive symptoms mediated by stress (mediation model 1), mediation analysis was conducted using an SAS macro provide by Preacher and Hayes (2008) [33]. Using mediation analysis, the direct effects of the IV and the indirect effect through mediator (M) on the DV were estimated. The indirect effect of the IV on DV via M was quantified as a×b. Path c represents the total effect of each personality factor on depressive symptoms, and path c' is the direct effect of personality factor on depressive symptoms. The total effect (c) of IV on DV equals the sum of the direct and indirect effects (c = c'+ab) [33].
We also tested multiple serial mediation effects of personality and stress on the association between gender and depression [34] (mediation model 2). For the multiple mediation analysis, three indirect effects of the IV on DV through M1 and M2 (IV!M1!Y, IV!M2!Y, IV!M1!M2!DV) can be estimated (Fig 1). Path a1 represents the effect of gender on the personality, and path a2 represents the effect of gender on the stress. Path a3 is the effect of personality on stress. Path b1 and b2 represent the effect of personality and stress on depression, respectively. The total effect (c path) of gender on depressive symptoms was quantified as sum of the direct and indirect effects of gender on depressive symptoms. Path c' is the direct effect of gender on depressive symptoms. The indirect effect of the IV on DV via M1 was quantified as a1×b1, and the indirect effect of IV on DV via M2 was quantified as a2×b2. The indirect effect via M1 and M2 equals subtraction of direct effect (c') and indirect effects via M1 (a1×b1) and M2 (a2×b2) from total effect (c) (c-[c'+ (a1×b1) + (a2×b2)]).
To test the mediation effect, 5000 bootstrapped resamples and a 95% confidence interval were applied to construct the indirect path. Bias-corrected confidence intervals that did not include 0 were considered significant for the indirect effect of the IV on DV via a mediator. Effect size in mediation analysis was assessed with completely standardized indirect effect size (Effect size cs ) indicating that DV is expected to decrease by the magnitude of effect size standard deviations for every one standard deviation increase in IV indirectly via mediators. Effect size cs is interpreted as small (0.01), medium (0.09), and large (0.25) [35,36]. Socio-demographic factors, which are known to be correlated with depression [37,38], such as age, education (less than high school, college), working status (employed, not employed), marital status (single, married/cohabiting) were included as covariates in all mediation models to control for their potential confounding effects on depressive symptoms. The mediation analyses without covariates were also conducted in order to validate our proposed mediation models.

Descriptive Analysis and Gender Differences among Study Variables
Descriptive statistics of demographic and psychological variables by gender are presented in Table 1. The mean stress level was significantly higher in women compared to men (t = -9.04, df = 2480.16, p<0.001). Mean depression scores measured by the CES-D were higher in

Correlation Analysis
The bivariate correlation analysis results among personality, stress, and depressive symptoms in all subjects are shown in Table 2

Direct and Indirect Effects of each Personality Factor on Depression via Stress
The first mediation model (model 1) tested how the personality factors were related with depressive symptoms via their effects on perceived stress ( Table 3). The mediation analysis was performed separately for each gender because of significant gender differences in personality traits, stress, and depressive symptoms.  significant mediation model were 0.03-0.23 for men and 0.09-0.028 for women, which are suggested to be a small to medium effect size for men, and moderate to large effect size for women (S1 Table). The mediation models without covariates showed similar findings (S2 Table).

Direct and Indirect Effects of Gender on Depression via each Personality Factor and Stress
The next analysis was designed to examine the direct effect of gender (IV) on depression (DV), and the indirect effects when mediated by personality (M1) and stress (M2) were significant (Fig 1).  The indirect effect of gender on depression through the pathway from extraversion and stress was significant (M1&M2, coefficient: 0.108, 95% CI: 0.040, 0.182), but effect size was very small (Effect size cs <0.01). The indirect effects of gender on depression were significant in each pathway via extraversion (a1×b1, coefficient: 0.054, 95% CI: 0.020, 0.102) and via stress (a2×b2, coefficient: 0.877, 95% CI: 0.585, 1.178).
Thus, the pathway from gender and depressive symptoms was mediated by neuroticism and extraversion with or without a path through perceived stress. Agreeableness and conscientiousness mediated the association between gender and depression only by an indirect path through perceived stress.

Supplementary Analysis
Because five factor personality traits were correlated among each other, it was difficult to determine unique effects of each personality trait when each personality factor was separately examined. We re-ran analyses to test unique effects of each personality trait with controlling for other traits by including all five traits simultaneously in the mediation models. The mediation model 1 showed consistent findings in neuroticism and extraversion for men and women (S5 Table). The mediation model 2 confirmed that neuroticism was a significant mediator between gender and depressive symptoms (S6 Table).

Discussion
We found that neuroticism and extraversion were directly and indirectly associated with depressive symptoms via perceived stress, while agreeableness and conscientiousness were Note. M1, mediator 1 (personality); M2, mediator 2 (stress); CI, confidence interval # a1×b1, the indirect effect of gender on depressive symptoms via each personality factor $ a2×b2, the indirect effect of gender on depressive symptoms via stress associated with depressive symptoms only through perceived stress. Moreover, female gender was associated with higher neuroticism and agreeableness and lower extraversion and conscientiousness, which were related to higher level of stress, and this could in turn lead to greater depressive symptoms. These findings may contribute to the understanding of how personality factors and perceived stress are related with depressive symptoms and how these factors contribute to the relationship between gender and depression. To our knowledge, this is the first study to examine the mediational roles of the five factor personality traits in relation to perceived stress on gender differences in depressive symptoms. Personality factors are known to be involved in the regulation of emotion and cognitive vulnerability that contribute to predisposition to depression [39]. Our findings show that high neuroticism and low extraversion had direct effects on depressive symptoms and also affected the stress level, which was correlated with increased depressive symptoms. Conversely, agreeableness and conscientiousness had indirect effects on depressive symptoms, but only when mediated by stress. Individuals with high neuroticism tended to promote maladaptive cognitive processes such as ruminative thinking and hopelessness, which increased the risk of depressive symptoms [40]. Higher extraversion was associated with positive emotionality and sociability, reducing stress-related responses and depressive symptoms [5]. Agreeableness and conscientiousness have been correlated with avoiding interpersonal problems and having positive emotional coping in response to stressful situations [1,4]. Our study supports the premise that certain types of personality factors (e.g., higher neuroticism and lower extraversion) may promote negative emotional regulation or maladaptive reactivity to stress, which increases the predisposition to depressive symptoms.
The current results indicate that female gender was associated with higher neuroticism and lower extraversion, which contributed to greater depressive symptoms directly or indirectly when mediated by stress. This result is consistent with previous findings that women scored consistently higher in neuroticism than did men across varying age ranges and world regions and that a high degree of neuroticism in adolescent girls was associated with later negative life experiences [25]. In addition, higher extraversion was associated with a reduced risk of anxiety and depressive disorders [5], while extraversion was found to have a protective effect against the risk of suicide in men [41]. Because gender differences in extraversion were not consistent with previous findings [42], it is possible that extraversion might not always explain the lower risk of depression in men.
Agreeableness and conscientiousness were previously found to be negatively associated with depressive symptoms [43], and our results demonstrated that these associations were mediated by stress in both genders. Agreeableness and conscientiousness did not solely explain the relationship between gender and depression after controlling for stress. These findings support the hypothesis that conscientiousness might play a role in stress-related vulnerability [44], which in turn might result in differential effects of gender on depression rather than as an single mediator.
It was previously reported that agreeableness has positive effects on coping with stress related with interpersonal aspects [45], but the current findings show that female gender was associated with higher agreeableness as well as greater stress and depressive symptoms. It was reported that the association between agreeableness and coping with stress was dependent on stressor types, and higher agreeableness might be less effective to cope with certain types of interpersonal stress [11]. In the present study, agreeableness showed significant but weak negative correlations with perceived stress and depressive symptoms, suggesting that higher agreeableness in women might have small effects on reducing general stress levels and depressive symptoms.
Despite the findings of significant gender differences in openness, this personality trait did not have a direct effect on depression or an indirect effect when mediated by stress. A previous Five Factor Personality, Gender, Stress and Depressive Symptoms study showed that higher scores in the subsets of openness to experiences was related to depression [46], while another reported that depressive individuals showed lower levels of openness to experiences [47]. In our study, openness to experience was not significantly associated with depressive symptoms in either women or men. Furthermore, gender differences in openness to experience were not associated with either higher perceived stress or greater depressive symptoms in women.
Although personality traits have been known to be important risk factors for depressive symptoms, the current study showed that the correlations between personality traits and stress or depressive symptoms had small to moderate effect sizes (r<0.4). Specifically, the correlation coefficients of agreeableness and conscientiousness with depressive symptoms or stress did not reach the minimum effect size of practical significance suggested by Ferguson [48]. One possible reason for the small effect size might be the multiple influences on depressive symptoms such as biological and environmental factors [49]. It was previously reported that several mediators account for the relationship between gender and depression [26]. However, perceived stress was only considered as a mediator with the personality factors in our analysis. So, the reasons that certain personality factors showed weak correlations with depressive symptoms might be due to various pathways including psychosocial and environmental factors by which certain personality factors could indirectly affect depression symptoms.
There are several strengths in the present study, including study sample size and methodological approaches. Although gender differences in depressive symptoms were previously reported across different countries [50], our study is first performed in the Korean population to examine the mediating roles of five factor personality traits on the link between gender and depressive symptoms. The large sample size derived from a population cohort was useful for studying the risk factors and etiology that may contribute to the development of depressive disorders in Korean adults. Despite these strengths, there are also several limitations. It was not possible to find a causal relationships among personality, perceived stress level, and depressive symptoms because one of the assumptions of mediation analysis is temporal precedence between variables [33]. Although this was a cross-sectional analysis and can therefore only be correlative, several studies support the hypothesis that personality traits as heritable and relatively stable individual characteristics are longitudinally associated with psychological distress and depressive symptoms.
In conclusion, our findings suggest that personality traits may be related with depressive symptoms through direct or indirect paths via their influence on perceived stress. Our results also showed that personality and stress play mediating roles in the link between gender and depression. Our approaches may be useful for understanding the effects of personality factors on vulnerability to stress and gender differences in depression susceptibility.
Supporting Information S1 Data. The data underlying the findings in this study. (XLSX) S1 Table. Completely standardized effect sizes for the mediation effect of stress in the associations between personality and depression.  Table. Completely standardized effect sizes for multiple mediation effects through personality and stress in the association between gender and depression. (DOCX) S4 Table. Multiple mediation effects through personality and stress in the association between gender and depression without covariates. (DOCX) S5 Table. Mediation effects of stress in the associations between personality and depression with controlling for other personality factors. (DOCX) S6 Table. Multiple mediation effects through personality and stress in the association between gender and depression with controlling for other personality factors.