“Resilience,” which is “the capacity to recover from adversity” [1], is receiving increasing attention from researchers and mental health professionals. A resilient individual has an optimistic perspective toward adversity and can manage stress, respond to change, and deal with adverse conditions [2-4]. Recent studies have revealed that resilience significantly affects the quality of life of people with terminal diseases or post-traumatic stress disorders [5,6]. Initially, resilience was considered a fixed personality attribute [7]. In contrast, animal and human research has shown that coping capacity is influenced by environmental and life threats. Therefore, resilience is no longer considered a “fixed attribute” but rather a “dynamic process” that evolves and improves in response to environmental stimuli [8]. Since the concept of resilience has evolved, its association with mental diseases in indivi-duals exposed to specific stresses has been examined. This is aimed at preventing mental illness by providing training or cultivating an atmosphere that promotes resilience [8].

Diverse strategies for measuring resilience involve understanding and defining it. A systematic review concerning strategies for evaluating resilience indicated that the majority of resilience scales, such as the Connor Davidson Resilience Scale (CD-RISC) [9] and the Resiliency Scales [10], evaluate the availability of protective variables that promote endurance to psychopathology [11]. Both measures were designed to evaluate human attributes, such as self-efficacy or optimism, rather than resilience. Another renowned resilience scale, the Ego Resiliency Scale, is much longer (102 items) [12]. Applying questionnaires with a relatively large number of items may be time-consuming and restrictive because of low response rates and data loss [11].

The Brief Resilience Scale (BRS) was designed by Smith et al. [1] in 2008 to evaluate resilience. The six-item BRS was developed to evaluate the ability to recover from stress rather than to assess protective variables that promote resilience as an outcome. In addition to its distinctive conceptual definition of resilience, in terms of quality based on several sources of validity evidence, reliability, responsiveness, and interpretability, the BRS was among the top three of the 15 resilience measures assessed by Windle et al. [11].

Several studies have analyzed the psychometric properties of the BRS. In recent years, validation studies of the BRS have been conducted in young adults in Brazil (Cronbach’s α = 0.76) [13], general adult populations in Germany (Cronbach’s α = 0.85) [14,15], Spanish adults with several medical conditions (Cronbach’s α = 0.83) [16], and Mexican and Chilean university students (Cronbach’s α = 0.77) [17]. Tansey et al. [18] proposed a two-factor model for the construct structure of the BRS after comparing BRS scores with a sample of American clients of vocational rehabilitation services. The authors used two confirmatory factor analysis (CFA) techniques: one assessing the one-factor model and the other evaluating the two- factor model. The results of the CFA analyses on these two models suggest that the one-factor model fit less well than the two-factor model. The internal consistency (Cronbach’s alpha) of the factor scores for negative and positive phrased items was 0.83 and 0.79, respectively. This two-factor model was also supported by Chinese [19], Polish [20], and Spanish samples [16].

Suicidality is a major public health concern [21]. In 2021, suicide was the leading cause of death among Koreans aged 20−30 years and the second leading cause of mortality among Koreans aged 40−50 years [22]. Suicide during peak economic activity may enormously affect worker productivity and welfare [23]. Previous research has linked increased suicide risk with high work- related stress [24], such as a tedious job, significant responsibility, high job pressure, and workplace bullying [25,26]. Moreover, high workloads and stress at work were found in Korea [27]. Previous research suggested that a low resilience score is associated with an increased chance of suicide attempts [28]. Lifetime analysis revealed that low resilience predicted increased suicidality across all age groups [29]. A review of data on the modifiers of suicidality offered the “buffering hypothesis” to explain how resilience decreases the probability of suicide [30]. Hence, the predictive validity of the BRS in identifying individuals experiencing severe stress and suicidality seems important in screening workers. However, to our knowledge, the predictive validity of the BRS in evaluating suicidality among employees has been insufficiently researched.

Based on these results, we tested the hypothesis that the Korean version of the BRS has strong psychometric qualities that indicate a two-factor structure, independent of cultural factors. Through correlation analysis between the BRS and CD-RISC, as well as between the BRS and mental psychiatric assessments of stress and depression, we examined the reliability and accuracy of the scale. To investigate the factor structure of the BRS, we examined a two- factor model. In addition, we assessed the sensitivity of the BRS to identify high-risk suicidal groups. We anticipated that groups with a high suicide risk would have lower resilience levels and hence, projected that workers without suicidality have higher resilience levels.

Participants

Participants (19 to 75 years) willingly attended a mental health examination program at the Workplace Mental Health Institute, Kangbuk Samsung Hospital in Seoul, Korea [31]. The Kangbuk Samsung Workplace Mental Health Institute is undertaking research to prevent suicide and promote employee mental health, and the current analysis is one of their research efforts. Employees from 54 enterprises and local government entities participated in the study. Participants were classified according to the Standard Industrial Classification System in the United States. All participants were assessed using the same evaluation tool, and the contents of the tool are described in the Instruments section. Sample 1 participants comprise a non-overlapping sample of mental health examination program recipients. Sample 2 participants visited the hospital twice monthly for analysis using the BRS for the test-retest test. Participants in Sample 1 did not overlap with those in Sample 2.

Sample 1

Sample 1 was collected between June 2015 and October 2019. We excluded 882 employees with incomplete questionnaires or missing sociodemographic information from the initial 15,360 respondents. Sample 1 included 14,522 adults (5,952 females, 41.0%) who were 16−72 years old (mean [M] = 38.95, standard deviation [SD] = 9.23). Table 1 presents the sociodemographic characteristics of Sample 1. A total of 41.0% of Sample 1 was single (never married, divorced, separated, or widowed). Most participants (69.3%) had a high educational level (uni-versity degree or higher).

Sample 2

Sample 2 was collected from those who attended a workplace mental health checkup program twice in a row between January 2017 and December 2018. We excluded 233 employees with incomplete questionnaires or missing sociodemographic information from the initial sample of 447 respondents. Sample 2 comprised 239 adults aged 24−57 (M = 33.97, SD = 7.06). Most participants were male (61.5%). A total of 52.3% of the participants were single (never married = 123; divorced, separated, or widowed = 2). Most participants (61.5%, n = 147) had a high educational level (university degree or higher).

Ethics Statement

This study was approved by the Institutional Review Board of Kangbuk Samsung Hospital (IRB no. KBSMC 2022-03-046). The board waived the requirement for informed consent, as we only used de-identified data routinely collected in workplace mental health screening checkups.

Instruments

The BRS consists of six 5-point Likert scale items (1 = strongly disagree to 5 = strongly agree) that measure the ability of an individual to bounce back from stress and difficulty [1]. Items 1, 3, and 5 of the BRS are positively worded, while items 2, 4, and 6 are negatively worded. However, before the analysis, we reverse-coded the negatively worded items. As BRS scores indicate a respond-ent’s resilience level, individuals with higher scores are more resilient. With a Cronbach’s α ranging from 0.80 to 0.91 across four samples, the original English version of the BRS displayed good internal consistency.

Three board-certified psychiatrists and two board-certified psychologists fluent in Korean and English translated the BRS into Korean and back-translated it into English. The process was repeated until the physicians decided that the Korean version was equivalent to the English version and was appropriate for Korean patients. Based on their comments, investigators revised the sections that were confusing or may have been misunderstood by Korean patients. After evaluation by a professional translator and Korean literature experts, all investigators agreed with the final version. Based on the original English form, the scale was translated according to the acknowledged criteria created for intercultural research (WHOQOL Trans-lation Methodology) [32].

The Korean version of the CD-RISC has been used to assess resilience [33]. This measure comprises 25 items rated on a 5-point Likert scale (ranging from 0 = never to 5 = almost always). The higher the CD-RISC score, the higher the resilience. In a previous study, the scores of the Korean version presented a high degree of internal consistency reliability (Cronbach’s α = 0.99) [33]. They also demonstrated good reliability in Sample 1 (Cronbach’s α = 0.95).

The Center for Epidemiologic Studies Depression Rating Scale (CES-D) was used to measure the severity of depressive symptoms [34,35]. The CES-D consists of 20 items rated on a 4-point Likert scale ranging from 0 to 3 points (total score, 0−60) related to the symptoms of depression, with higher scores indicates more significant depression levels. Scores of 21 or above indicate a depressed state. Cho and Kim [34] demonstrated good internal reliability of the scale, with Cronbach’s α ranging between 0.84 and 0.91. The Cronbach’s α for the CES-D in this study was 0.86, indicating a good level of internal consistency within the context of our sample.

The PSS is a 10-item survey on a 5-point Likert scale ranging from 0 (never) to 4 (always) (very often). Items 1, 2, 3, 6, 9, and 10 are scored positively (0 = never, 1 = rarely, 2 = occasionally, 3 = often, 4 = very frequently), and items 4, 5, 7, and 8 are scored negatively. The higher the PSS score, the greater the perceived stress. The Korean translation scores showed acceptable reliability (Cronbach’s α = 0.82), contemporaneous indications of validity, and sensitivity [36]. In addition, they showed good internal reliability in our sample (Cronbach’s α = 0.75).

We assessed suicidality using questions from a component of the Korean National Health and Nutrition Examin-ation Survey. This survey is conducted yearly by the Korean government to research public health status and generate statistics for establishing and evaluating public health policies [37]. Suicidality was organized as suicidal ideation (SI), suicidal plan (SP), and suicidal action (SA) by self-reported dichotomous questions. Responses to the questions were “Yes” or “No.” Regarding SI, SP, and SA, participants were asked if they had contemplated suicide, if they had created a precise plan for suicide, and if they had attempted suicide within the preceding year. The content of the specific questions has been presented in a previous study [38].

Participants’ responses were grouped according to their suicide status as follows: 1) those who had never contemplated suicide (no suicidality), 2) those who had only experienced suicidal ideation (without SP or SA, only SI), and 3) those who had planned or attempted suicide (SP or SA).

Statistical Analyses

The item endorsements for each response category and the accompanying skewness and kurtosis values were computed. Next, the validity, reliability, and factor structure of the BRS were investigated. We used confirmatory factor analysis with maximum likelihood implemented in AMOS version 24 (IBM Co.) to validate the factor structure of the BRS with probable non-normality in the data distribution. Given that the factor structure of prior research indicates that the parameters of a model with two factors provide the best fit, we evaluated a two-factor model with one factor for positively worded items and one factor for negatively worded items [16,39]. Model fit was assessed using the Chi-squared statistic, standardized root mean squared residual (SRMR), root mean square error of approximation (RMSEA), goodness-of-fit index (GFI), and comparative fit index (CFI). Although there is no consensus on the acceptable levels of fit indices [40], values of 0.08 or less for the RMSEA, 0.08 or less for the SRMR, and 0.95 or more for the CFI and GFI, are considered to indicate acceptable fit [41]. As suggested in previous research [42], we evaluated reliability using both Cronbach’s α and composite reliability (McDonald’s ω) to assess internal consistency. We examined the test-retest reliability using Pearson’s correlation and the intraclass correlation coefficient for absolute agreement. To determine the congruent validity of the BRS and divergent validity between the BRS and depression and perceived stress, we analyzed correlations among the BRS, resilience measured with the CD-RISC [33], depression measured with the CES-D [35], and perceived stress measured with the PSS [36]. In addition, two approaches were used to evaluate the sensitivity of BRS. First, we examined the influence of demographic factors (age, sex, and education level) on BRS scores to determine whether the effect of these variables was in the same direction as that observed in prior research. To do this, we conducted analyses of variance (ANOVAs) with age, sex, and degree of education as independent factors and the BRS score as the dependent variable. We also assessed the scale’s capacity to identify groups with varying levels of suicidality. To do this, we employed ANOVA with the total BRS score as the dependent variable and groups classified by suicidality as independent variables: 1) those who had never contemplated suicide (no suicidality group), 2) those who had only experienced SI (with-out SP or SA, only SI group), and 3) those who had planned or attempted suicide (SP or SA group). Statistical significance was defined as a pvalue < 0.05 (two-tailed). Except for CFA, which was performed using AMOS version 28, all analyses were performed using SPSS version 25 (IBM Co.).

Psychological Characteristics of Each Sample

Table 2 presents the psychological characteristics of the participants. The skewness and kurtosis values were within the acceptable range to consider the scores of psychological assessments in all samples normally distributed [43].

Distribution of BRS Scores

Table 3 displays each BRS question’s response distribution, means, standard deviations, skewness, and kurtosis data in Sample 1. The findings revealed that item distribution encompassed the whole range of possible re-sponses. In addition, the skewness and kurtosis values were within the acceptable range for considering the BRS item answers to be normally distributed. The mean total BRS score was between 6 and 30, the skewness value was −0.22, and the kurtosis value was 0.36. In the Sample 2 test and retest, no one responded 1 to BRS 1, but the skewness and kurtosis values were within the acceptable range. The total BRS scores ranged between 5 and 30, the skewness value was −0.46, and the kurtosis value was 0.36.

Factor Structure

The standardized model of the two-factor BRS in Sample 1, as well as the unstandardized estimates and standard errors for Sample 1, are shown in Figure 1 and Table 4, respectively. All the estimated loadings were significant (p < 0.001). The factor analysis results yielded KMO values of 0.92 for the BRS in Sample 1. Regarding the fit statistics in Sample 1, the Chi-square statistic (χ² = 391.92, χ²/df = 48.99) was significant, probably due to the size of the sample [44]. However, the SRMR (0.01), RMSEA (0.06), GFI (0.99), and CFI (0.99) were well within the limits that indicate model acceptability (0.08 or less for the RMSEA, 0.08 or less for the SRMR, and 0.95 or more for the CFI and GFI). The findings demonstrated that the model with two first-order components for questions that were positively and negatively worded offered a good fit to the BRS scores.

Reliability Analyses

For the 6-item BRS in Sample 1, Table 5 shows the corrected item-total correlations and alpha if the item was eliminated. Individual item-adjusted item-total correlations were greater than 0.30, indicating their eligibility for scale construction [45,46]. The BRS values of all the samples ranged from 0.67 to 0.80. Reliability using Cronbach’s α and McDonald’s ω showed good reliability in Sample 1, with α = 0.91 and ω = 0.91. There were no items whose Cronbach’s α exceeded the total Cronbach’s α, after excluding the item. Pearson’s test-retest correlation was 0.64 (p < 0.001).

Convergent and Concurrent Validity

As anticipated, the BRS had a substantial positive correlation with CD-RISC-measured resilience (r = 0.90, p < 0.001). Moreover, the correlation between the BRS and depression was negative and statistically significant (r = −0.22, p < 0.001). Table 6 showed that the BRS and perceived stress were also significantly correlated (r = −0.48, p < 0.001).

Sensitivity of the Scale to Sociodemographic Variables

Regarding sex differences, male had a significantly higher level of resilience (M = 18.56, SD = 3.23) than female (M = 17.33; SD = 3.21) in Sample 1, as shown by the ttest (t = 22.53; p < 0.001). Similarly, ANOVA showed significant differences between age groups (F = 177.19; p < 0.001). Mean differences (M_diff) were significant between the age group 20−30 (M = 17.50, SD = 3.22) and the age groups of 41−50 (M = 18.55, SD = 3.25, M_diff = −1.04, p < 0.001) and 51−60 (M = 19.18, SD = 3.20, M_diff = −1.68, p < 0.001), except for age group 31−40 (M = 17.58, SD = 3.21, M_diff = −0.08, p = 1.00). The effect of education level was significant (F = 110.06; p < 0.001). Post-hoc tests after Bonferroni correction showed that BRS scores were significantly higher in the master’s degree or higher group (M = 19.43, SD = 3.11) than in the university degree (M = 18.04, SD = 3.23, M_diff = 1.39, p < 0.001), college degree (M = 17.59, SD = 3.26, M_diff = 1.84, p < 0.001), and high school or below (M = 17.71, SD = 3.37, M_diff = 1.73, p < 0.001) groups. BRS scores in the university degree group were also significantly higher than those in the college degree group (M_diff = 0.45, p < 0.001) and the high school or below group (M_diff = 0.34, p < 0.001). There was no significant difference between the college degree and high school and below groups.

Sensitivity of the Scale to Detect High-risk Populations

The ANOVA showed that differences in the level of resilience were significant among the groups (F = 568.89; p < 0.001). The Bonferroni post-hoc test showed that the BRS scores were higher in the no suicidality group (M = 18.49, SD = 3.14) than in the only SI group (M = 16.56, SD = 3.05, M_diff = 1.93, p < 0.001) and SI or SP group (M = 15.75, SD = 3.59, M_diff = 2.74, p < 0.001).

Our study indicated that the Korean version of the BRS is a reliable and viable research instrument. Based on the current study, the value of the Korean version of the BRS is identical to that of the original version, and it has solid psychometric qualities. Under Korean circumstances, the BRS accurately measures resilience as the capacity to resist in the sense of “resiliently bouncing back,” adapting to obstacles, or recovering from stress. It may also provide important information about an individual’s perceived stress, depression, and suicidality. Our results also corroborate the psychometric features described in the literature [1], as well as the findings of multiple international validation studies [13,16,19,20].

Based on CFA, a two-factor model was validated, indicating the homogeneity of the BRS, which is compar-able to previous studies conducted in other languages [16,19,47]. Although it was initially suggested that the BRS is unidimensional with negatively worded items, the scale was created solely using exploratory factor analysis data from four samples [1]. However, several later BRS studies have utilized CFA to confirm the underlying component structure of the scale, revealing that it consists of two latent factors: positively worded items and negatively worded items [13,16,18,19]. This tendency has also been supported in BRS literature on employees or populations in underdeveloped nations [17,48].

Our research results showed that the Korean version of the BRS showed good internal consistency based on Cronbach’s α (0.91) and McDonald’s ω (0.91), with values similar to those obtained in the English (Cronbach’s α ranging from 0.80 to 0.91) [49], Spanish (Cronbach’s α = 0.83) [16], and Polish versions (Cronbach’s α = 0.83) [20]. Regarding test–retest reliability, the Korean version of the BRS exhibited strong test-retest reliability, with values comparable to those observed in the original and Spanish versions [1,16].

Moreover, our analyses, designed to examine convergent and divergent validity, revealed that BRS scores were substantially correlated with a resilience question-naire. As predicted, a positive connection was observed between the BRS and CD-RISC resilience indices. Regard-ing convergent validity, we discovered negative associations between the BRS score and symptoms of depression and perceived stress. Several previous studies have reported negative correlations between resilience and symp-toms of mental dysfunction and perceived stress [20,39, 50,51]. These correlations of resilience with factors related to individual mental health show the necessity of including resilience in health-screening systems for em-ployees.

Similar to previous research [1,14,16], we discovered that males tend to have higher BRS scores. Regarding age, our results contradict earlier findings [14,49], indicating that BRS scores increase with age. The age range of the research sample may be a possible explanation. Our participants’ average age was between the two clinical samples (M = 62.80 and 47.70) and the two student samples (M = 20.04 and 19.80) from the original paper [1]. There was no significant difference between those aged 21 to 30 and those aged 31 to 40 in our data, in which BRS scores were much higher than 40, which may indicate that the change in resilience with age is not linear. Regarding educational differences, there was no difference between the college degree group and the high school diploma or less group. Nevertheless, the higher education group had higher BRS scores, indicating the sensitivity of the BRS scores. This argument supports educational policies that promote higher levels of education since they are associated with a greater degree of resilience [52,53].

Furthermore, our results provide evidence of the scale’s sensitivity in detecting groups with high suicidality. Ac-cording to a previous study [54], we hypothesized that groups with suicidality (SI, SP, or SA) would score lower in resilience. Our results showed that the group without suicidality had higher BRS scores than those with suicidality who had experienced SI, SP, or SA. Our research has provided evidence for the predictive validity of the BRS, demonstrating that resilience scores can detect the risk of suicide.

Our study benefited from its population-based characteristics, large sample sizes, and large age ranges. Our study has also added to resilience studies by demonstrating that subjective resilience assessments may predict the probability of suicide among employees. However, these investigations had limitations. The first potential constraint relates to the sample’s composition: the research group comprised only employees. In future research, it would be worthwhile to increase the diversity of the study groups to make them more representative. In addition, although we obtained health examination data, we could not acquire information on each participant’s health status. Therefore, based on several previous studies on the BRS concerning the ability to recover from a health issue, we suggest using subsamples for future research. This will enable better comparisons between groups and the development of normative studies that provide data specific to each type of population for the use of the scale in clinical settings.

The Korean BRS is a reliable method of measuring re-silience. Owing to the brevity of the scale, it may be used in epidemiological research on mental health and psychological resilience in diverse groups and in investiga-tions requiring quick and economic measurements. The current findings imply that the Korean version of the BRS has excellent convergent and divergent validity with well- established measures of CD-RISC, depression, and perceived stress. Furthermore, the quality of the psychometric features of its scores based on a large sample makes this version of the BRS preferable to other resilience measures that are presently accessible in Korean.

This research was financially supported by the Ministry of Trade, Industry, and Energy (MOTIE), Korea, under the “Fostering Program for The Value-Chain through Industrial Intelligence” (reference number P0017821) supervised by the Korea Institute for Advancement of Technology (KIAT). This funding source had no role in the design of this study and will not have any role during its execution, analyses, interpretation of the data, or decision to submit the results.

No potential conflict of interest relevant to this article was reported.

Conceptualization: Sang-Won Jeon, Changsu Han. Data acquisition: Sang-Won Jeon, Changsu Han. Formal analysis: Junhyung Kim. Funding: Changsu Han. Supervi-sion: Sang-Won Jeon, Hyun-Ghang Jeong, Moon‑Soo Lee, Seung-Hoon Lee. Writing- original draft: Junhyung Kim. Writing- review & editing: Sang-Won Jeon, Changsu Han, Hyun-Ghang Jeong, Moon‑Soo Lee, Seung-Hoon Lee.