INTRODUCTION
The sinking of the ferry MV Sewol, which took place on April 16, 2014, left 304 passengers dead or missing out of 407 total passengers. This disaster is considered to be the most psychologically shocking and traumatic event in Korean history, comparable to the Korean War [1-3]. Studies conducted outside of Korea have reported that 50% of adults experience the unexpected death of someone who holds a special meaning to them at least once in their lifetime, and for one-third of these adults, such loss is the most traumatic event in their lives [4,5]. The death and the loss of a family member owing to a sudden event can be the most painful experience in life. As a result, bereaved families exhibit noticeable changes in their thinking and experience negative emotions such as anger, helplessness, and depression [6-9], while facing serious crises in interpersonal relationships and social life [10,11]. Trauma at a level beyond control causes individuals to feel helplessness, embitterment, and prolonged grief at a level similar to post-traumatic stress disorder (PTSD) and is also closely associated with various mental health issues, including alcohol addiction [12-16]. Follow-up studies on bereaved families of Sewol Ferry Disaster victims and those of survivors found that these individuals continue struggling with moderate-to-severe depression even 7 years after the disaster, while their mental health issues, such as suicidal ideation, have persisted for a long time [2,17].
Thus far, most studies on the bereaved families of Sewol Ferry Disaster victims comprised status surveys or qualitative investigations of post-traumatic stress (PTS) or post-traumatic embitterment (PTE). Depression is a very common condition in individuals after the loss of a family member. This disorder causes cognitive and physical changes that affect an individual’s ability to function, which can manifest differently depending on the person. The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) classifies depression into major depressive disorder and persistent depressive disorder or anxious, melancholic, and atypical subtypes according to the severity of symptoms for better understanding of the clinical features of the individuals [18]. Such classifications indicate the existence of different perspectives on depression and that there is a need for a common consensus. To classify and understand the subtypes of depression, various attempts with a variable-centered approach were made. However, such an approach is limited in explaining the diverse depressive characteristics in individuals; furthermore, it cannot be used as evidence for screening those who require immediate intervention. In particular, determining the existence and severity of depression based on total scores measured using a depression screening tool is not accurate enough to understand their characteristics.
Accordingly, this study used latent profile analysis (LPA), a type of person-centered approach, to overcome these limitations, identify the types that should be prioritized for intervention, and facilitate actual understanding. LPA assumes that the entire population is composed of heterogeneous subpopulations when the measured variables are continuous and classifies individuals into types based on their behavioral characteristics for the person-centered understanding of individual behaviors [19,20]. In addition, because these characteristics are identified more closely to reality, it is useful to identify high-risk populations and understand the characteristics of target populations for prioritized intervention, while it also enables a customized approach for modifying such characteristics [21].
While previous studies have established the general impact of the Sewol Ferry Disaster on the mental health of bereaved families [2,17], less is known about the specific contribution of interpersonal stress as key factors influencing the different profiles of depressive symptoms in this population. Existing literature has documented general interpersonal difficulties and the role of social support in disaster recovery [10,22-24]. However, this study aims to extend this understanding by examining how distinct dimensions of interpersonal stress—specifically, family and non-family—relationship with social support and subjective health status to predict different depressive symptom profiles among bereaved families.
In this study, latent types of depression were classified based on the sub-variables of the Patient Health Questionnaire-9 (PHQ-9) and the characteristics and influencing factors of each type were identified. Moreover, the study also investigated whether there are differences in PTS, PTE, and post-traumatic growth (PTG), which are considered to be significant variables in a traumatic event, by each latent subtype. Furthermore, this study aimed to use the findings as evidence for developing programs and formulating policies for bereaved families of Sewol Ferry Disaster victims.
METHODS
Study population
The study population consisted of 302 individuals aged ≥15 years who voluntarily participated in a survey for “2021 health and social welfare study on the victims of the Sewol Ferry Disaster” conducted by Ansan Mental Health Trauma Center among bereaved families of Sewol Ferry Disaster (including the families of unrecovered victims).
Study instruments
Depression
Depressive symptoms were measured using PHQ-9, which was originally developed by Spitzer et al. [25] to assess the mental health of patients in primary health care centers and subsequently adapted to the Korean population by Han et al. [26]. The reliability and validity of this instrument have been verified. This instrument measures how often the respondent experienced each item within a recent depressive situation, with each item rated on a 4-point scale (0=not at all; 1=several days; 2=more than a week; and 3=almost every day). The total score ranges between 0 and 27 points, with scores of ≤4, 5-9, 10-19, and ≥20 points indicating normal, mild, moderate, and severe levels, respectively. In this study, the cut-off score was set to 10 points. The reliability of this instrument was Cronbach’s α=0.92.
Drinking frequency
Drinking frequency was evaluated using one item measuring how often the respondent drank. The item was rated on a 5-point scale (1=not at all; 2=less than once a month; 3=2-4 times a month; 4=2-3 times a week; and 5=4 or more times a week). Higher scores indicated high drinking frequency.
Subjective health status
Subjective health status was evaluated using one item measuring how the respondent thought about his or her current health status. The item was rated on a 5-point scale (1=very healthy; 2=somewhat healthy; 3=average; 4=somewhat unhealthy; and 5=very unhealthy). Higher scores indicated poorer health status.
Interpersonal stress
Interpersonal stress was analyzed by dividing it into family relationship stress and non-family relationship stress. Family relationship stress was measured using one item evaluating whether there was any stress between the respondent and his or her family in the past year. The item was rated on a 5-point scale (1=almost never; 2=some; 3=moderate; 4=considerable; and 5=extreme). Higher scores indicated higher levels of family relationship stress. Non-family relationship stress was measured using one item evaluating whether there was any stress between the respondent and his or her family in the past year. The item was rated on a 5-point scale (1=almost never; 2=some; 3=moderate; 4=considerable; and 5=extreme). Higher scores indicated higher levels of social relationship stress.
Social support
Social support was measured using the Korean version of the Duke-UNC Functional Social Support Questionnaire [27], which is a self-reporting social support scale consisting of 14 items, with each item rated 1-5 points for a total score ranging between 14 and 70 points. Higher scores indicated sufficient social support. In this study, the reliability of this instrument was Cronbach α=0.96.
PTS
PTSD symptoms were measured using the PTSD Checklist for DSM-5 (PCL-5), which was revised from the PCL originally developed by Weathers et al. [28] to reflect the diagnostic criteria of DSM-5 and subsequently validated by Lee et al. [29]. This instrument consists of a total of 21 items on how often the respondent experienced the symptoms, with each item rated on a 5-point scale (0=not at all; 1=a little bit; 2=moderately; 3=quite a bit; and 4=extremely). The subcategories consist of intrusion (5 items), avoidance (2 items), negative alterations in cognitions and mood (7 items), and increased arousal (6 items). The total score ranges between 0 and 80 points and the cut-off score is 37 points. Higher scores indicate more severe PTSD symptoms. In this study, the reliability of this instrument was Cronbach α=0.97.
PTE
PTE was measured using the standardized Korean version of the PTED Self-Rating Scale (PTED Scale) [30] originally developed by Linden et al. [13]. The PTED Scale consists of 19 items, which are largely divided into items designed to assess psychological status, social functioning, emotional response to the event, and thoughts of revenge. Each item was rated on a 5-point scale (0=not true at all; 1=hardly true; 2=partially true; 3=very much true; and 4=extremely true). The mean total score ranges between 0 and 5 points with <1.6 points indicating regular state; 1.6 to <2.5 points indicating suffering from prolonged embitterment; and ≥2.5 points indicating serious impairment due to embitterment. The cut-off score was 2.5 points. Higher scores indicate more severe embitterment symptoms. In this study, the reliability of this instrument was Cronbach α=0.97.
PTG
PTG was measured using the PTG Inventory-short form (PTGI-SF), which was validated and condensed by Cann et al. [31] from the original PTGI developed by Tedeschi and Calhoun [32]. PTGI-SF consists of 10 items in four subdimensions of changes in self-awareness, increased depth of interpersonal relationships, discovery of new possibilities, and increased spiritual/religious interest. This inventory measures how much change the respondent experienced after a crisis event with respect to each item, with each item rated on a 6-point scale (0=no change; 1=change to a very small degree; 2=change to a small degree; 3=change to a moderate degree; 4=change to a great degree; and 5=change to a very great degree). This inventory has a total score ranging between 0 and 50 points with higher scores indicating greater PTG. In this study, the reliability of this instrument was Cronbach α=0.91.
Data collection and ethical consideration
This study was approved by the Institutional Review Board of Kangwon National University (KWNUIRB-2021-08-013-002). Prior to data collection, bereaved family groups were contacted to provide information about this study and the questionnaire survey and request their cooperation. Information about the purpose and procedures of the study was provided in writing to obtain voluntary consent. Data were collected through an online questionnaire survey platform between October 5 and December 13, 2021.
Statistical analysis
All statistical analyses in this study were performed using SPSS 21.0 (IBM Corp.) and Mplus 8.0 (Muthen & Muthen).
First, SPSS 21.0 was used to analyze the descriptive statistics of the variables and the reliability of the instruments used in this study.
Second, LPA was used to differentiate subgroups according to the 9 items regarding depressive symptoms of bereaved families of Sewol Ferry Disaster victims. LPA is a method that groups individuals who share a similar response pattern into a single class based on the collected data and identifies the profile of the latent class in actual data [20,33]. LPA shares the same purpose as cluster analysis in that they both classify subgroups according to the inter-relatedness of the participants, but LPA has few methodological advantages over traditional cluster analysis. First, LPA offers higher classification accuracy based on probabilistic estimation for class classification, and measurement errors can be controlled by using latent variables. Moreover, because goodness-of-fit (GoF) indices are used to determine the optimal latent profile model, errors due to the subjective judgment of the researcher can also be minimized [19].
Generally, there are no absolute criteria for selecting the optimal number of profiles. This study used estimation of models based on classes with one to five latent profiles to find the optimal number of latent profiles, while comprehensively considering various GoF indices and the ease of model comparative testing and interpretation [34]. Statistical indices were used to verify the information ratio, model comparative testing, and quality of classification. Specifically, Akaike information criterion (AIC), Bayesian information criterion (BIC), and sample-size adjusted BIC (SABIC) were used for the information ratio, where lower values of all three indicate the fitness of the model [35]. For model comparative testing, Lo-Mendell-Rubin adjusted likelihood ratio test (LMR-LRT) and parametric bootstrapped likelihood ratio test (BLRT) were used [36]. Meanwhile, the quality of classification was checked using entropy and posterior probability of belonging. Entropy has a value between 0 and 1, with higher values indicating clear class classification, and a value of approximately 0.8 or higher is considered to indicate good classification [37]. For the posterior probability of belonging, a diagonal matrix of ≥0.7 is generally determined to show relatively accurate class classification [38]. After compilating these statistical indices, class interpretability was considered to determine the optimal number of latent classes.
After the classification of latent profiles, a three-step approach of LPA was used to investigate the influence of influencing factors and differences in PTS, PTE, and PTG by each latent profile [39]. The three-step estimation procedure is used to control the influence on class classification when independent and outcome variables are included together with latent classes in the model being analyzed. The three-step estimation procedure has the advantage of preventing a class from being altered by independent and outcome variables [40].
Study model
The study model used in this study is shown in Figure 1. First, latent profiles were classified based on the responses given by bereaved families of Sewol Ferry Disaster victims with respect to depressive symptoms. Next, the significance of influencing factors that were assumed to have an influence on the latent profiles of depressive symptoms was assessed. Last, the differences in the levels of PTS, PTE, and PTG by each latent profile were tested.
RESULTS
Descriptive statistical analysis
The descriptive statistics of the variables used in this study and the relationship with the deceased are provided in Table 1. Among a total of 302 participants, there were 112 males (37.1%) and 190 females (62.9%), with a mean age of 47.03 years. The average household income was 2,420,000 won. The average monthly drinking frequency was 1.48 times. The mean subjective physical health status of bereaved families was rated to be 3.07 points, while the mean scores of family relationship stress and non-family relation ship stress were 2.08 and 1.99 points, respectively. Moreover, the sum of depression and PTS were 9.93 and 28.35, respectively. Mean PTE and PTG scores were 1.63 and 1.62 points, respectively.
The descriptive statistics for the severity of depressive symptoms are presented in Table 2. A score of ≤4 is considered normal, 5 to 9 is considered mild, 10 to 19 is considered moderate, and ≥20 is considered severe. The moderate level showed the highest rate at 37.1% (n=112), followed by the normal level at 26.2% (n=79), the mild level at 25.5% (n=77), and the severe level at 11.3% (n=34).
Determination of number of latent profiles
Table 3 shows the results from the analysis of classes with two to four latent profiles of depressive symptoms in bereaved families. The model with five latent profiles included a class that comprised 5% of samples and the model was not appropriately estimated. Accordingly, it was discovered that the use of five classes was not appropriate for the data evaluated in this study. The analysis results showed that all information ratios (AIC, BIC, and SABIC) tended to decrease with an increasing number of latent profiles. LMR-LRT analysis results showed that when the number of latent profiles was two and four, the null hypothesis was rejected at significance levels of 1% and 0.1%, respectively, but when the number of latent profiles was three, the null hypothesis could not be rejected at significance level of 5%. However, with BLRT, which is traditionally accepted as a more reliable result, the null hypothesis was rejected at a significance level of 0.1% in all models. Additionally, entropy values were larger than 0.8, which is considered an empirically acceptable level, in all models. Considering the actual meaning of the classes and ease of interpretation for the final estimated model to be described later, the model with three classes was selected as the optimal model. With respect to the classification rate, the model with three latent profiles showed no problems with the classification rate.
Characteristics by latent profile
The mean values of latent profiles of depressive symptoms in bereaved families are presented in Table 4, while the distribution is shown in Figure 2. In the comparison of three groups, the first group, accounting for 37.9% of all respondents, showed low scores in all PHQ items. Accordingly, this group was named the overall low-level (LOW) group. The second group, accounting for 41.9% of all respondents, tended to show higher scores for feeling tired or having little energy (1.90 points), trouble falling or staying asleep, or sleeping too much (1.79 points), no interest or pleasure in doing things (1.59 points), feeling down, depressed, or hopeless (1.38 points), and poor appetite or overeating (1.35 points) than other items. Accordingly, this group was named the lethargy and physical symptoms (LPS) group. The third group, accounting for 20.2% of all respondents, tended to show higher scores for items related to depressive symptoms than the other two groups. Accordingly, this group was named the overall high-level (HIGH) group.
Verification of influencing factors of latent profiles of depressive symptoms
The influence of the factors that were assumed to influence the final model of three latent profiles was verified. Table 5 shows the results from verifying which variables have a statistically significant influence on latent profiles among gender, age, household income, drinking frequency, physical health, family relationship stress, non-family relationship stress, and social support.
The participants with higher levels of family relationship stress and lower levels of social support were more likely to belong to the HIGH group than LPS group. Moreover, the participants with poorer physical health and higher levels of non-family relationship stress were more likely to belong to the LPS group than the LOW group. Furthermore, the participants with poorer physical health, higher levels of family relationship stress, higher levels of non-family relationship stress, and lower levels of social support were more likely to belong to the HIGH group than LOW group.
Differences in PTS, PTE, and PTG by each latent profile of depressive symptoms
Table 6 shows the results from testing significant differences in PTS, PTE, and PTG by each latent profile classified according to depressive symptoms.
With respect to PTS and PTE, differences between all groups were found to be significant. The results showed a highest-to-lowest order of HIGH, LPS, and LOW groups, with these three groups showing significant differences in PTS (χ2=337.04, p<0.001) and PTE (χ2=258.94, p<0.001). In other words, the HIGH group had higher levels of PTS and PTE. In contrast, no significant differences in PTG were observed between the groups.
DISCUSSION
In this study, latent profiles of depressive symptoms in bereaved families of Sewol Ferry Disaster victims were identified using a person-centered approach, and the relationships between latent profiles; in addition, antecedents and outcome variables were verified. The major findings of this study were as follows.
First, latent profiles were classified as LOW (37.9%), LPS (41.9%), and HIGH (20.2%) groups according to the depressive symptoms in bereaved families. Of these three groups, the LOW group showed low scores for all items regarding depressive symptoms. The LPS group tended to show relatively high scores for items related to feeling tired or having little energy, no interest or pleasure in doing things, sleep problems, and poor appetite or overeating. The HIGH group showed higher scores for depression-related items than the LOW and LPS groups, which confirmed that this was the target group that required prioritized intervention. The group also showed a pattern of reporting sleep problems, feeling hopeless, and feeling bad about oneself or being a failure at a high rate.
Considering the lack of studies on bereaved families of Sewol Ferry Disaster victims using quantitative methodologies with depression as the outcome variable, using LPA to identify the types of latent classes in the actual data set this study apart from previous ones. The total score of depressive symptoms of the subjects of this study was 9.93 points, and 37.1% of the subjects had a total score of 10-19 points (moderate level) and 11.3% had a score of 20 points or higher (severe level). Although the proportion of clinical groups such as depression and PTSD in bereaved families has been gradually decreasing over time, this result is similar to the results of a previous studies [2,17] that reported that the level is still high compared to a control group with similar conditions who did not experience the Sewol Ferry Disaster.
Second, an examination of the influencing factors of depressive symptom profiles showed that poorer subjective health status was associated with a higher likelihood of belonging to the HIGH or LPS group than the LOW group. Such results can be interpreted as bereaved families with poorer physical health have overall higher level of depression symptoms and suffer from lethargy and somatic symptoms, which support the results from previous studies reporting that bereaved families experience dental disease, pancreatitis, loss of concentration, and insomnia owing to extreme stress even long after the disaster [10,22] and that somatic symptoms of disaster victims can be useful predictors for differentiating between normal and high-risk mental health groups [41,42]. Moreover, some members of bereaved families exhibited symptoms suggesting alcohol dependence, such as frequent alcohol consumption to relieve insomnia [23,43], while it has been reported that some victims, including bereaved families, showed increased alcohol consumption after the disaster in 2014. Therefore, it is necessary to provide psychological and medical support, including continued support for treatment costs, to bereaved families to prevent the deterioration of their mental health, including depressive symptoms. Moreover, a higher level of family relationship stress was associated with a higher likelihood of belonging to the HIGH group than the LOW or LPS group. This can be interpreted as bereaved families with higher levels of interpersonal relationship stress may have an overall high level of depressive symptoms. Such findings are consistent with previous studies reporting that bereaved families who experience a traumatic event often experience problems in interpersonal relationships, including family relationships, which can pose psychological stress [10,22,44,45]. Therefore, interventions for improving family relationships and social relationship stress coping skills are needed. Also, beyond solving individual problems, family therapy may also be useful for the health of the family, including changes in family roles and family systems.
Social support was also a factor that influenced the profiles, and specifically, the participants with lower levels of social support were more likely to belong to the HIGH group than the LOW or LPS group. This indicated that bereaved families with lower social support may have an overall high level of depressive symptoms, which aligns with previous studies suggesting that social conflict and isolation experienced by bereaved families delay psychological recovery, whereas social support can act as a restorative factor for extreme grief reactions [10,24,46]. Therefore, there is a need to continue to provide interventions to increase social support, such as establishing healing communities and self-help groups within the community. In summary, individuals with poorer physical health, higher family relationship stress, higher non-family relationship stress, and lower social support were more likely to belong to the HIGH group. Because trauma caused by social disasters does not have a set time for healing, interventions customized for high-risk groups with such characteristics should consider the causal relationship with physical disorders attributable to mental stress and continue providing psychological support while extending the duration of medical support. In addition, family interventions, such as family education, family counseling, marriage counseling, and case management, are needed to strengthen family relationships, while various psychology-, physical-, and activity-based specialized programs should be provided to cope with non-family relationship stress. To strengthen the sense of social support, it is also necessary to establish a social support system that can build psychological stability and a sense of bonding through self-help groups and peer support activities.
Third, examination of differences in PTS, PTE, and PTG by each latent profile of bereaved families showed that PTS and PTE were significantly higher in the HIGH group than LOW and LPS groups. After experiencing a major disaster, negative emotions such as grief should dissipate over time, but recovery reactions have not often taken place among bereaved families of Sewol Ferry Disaster victims or families of survivors even after a long period. Considering this, it is important to provide professional healing programs that focus on resilience and empowerment in relation to PTS and PTE. The findings showed no statistically significant difference in PTG by latent profiles but considering that PTG was lower in the HIGH group, it is necessary to provide psychological acceptance-based interventions that can promote changes in self-perception, changes in the meaning of interpersonal relationships, and new possibilities.
Limitations of this study and recommendations for future studies are as follows. First, this study could not confirm that demographic variables, such as gender, age, household income, and drinking frequency were predictors of each subgroup. Such findings were in contrast to the results reported in previous studies [23], and thus, replication studies are needed to generalize the findings. Second, this study investigated results at a specific time point based on self-reported responses regarding depression. Therefore, caution should be taken when interpreting the results. Another limitation is that because the data used in this study were cross-sectional data, long-term characteristics of individuals could not be identified as with panel study data. Depressive symptoms may change owing to an external environment or an event. Therefore, data collected over a longer period can be used to perform latent transition analysis or growth mixture modeling for longitudinal examination of how depressive symptoms change and which factors influence such change. This study did not identify under-age characteristics in participants aged 15 years or older. Further study is needed that considers the characteristics of trauma in under-age individuals. Despite these limitations, the significance of this study is that it differentiated depression in bereaved families of Sewol Ferry Disaster victims by latent profiles and presented the characteristics of each type, unlike methodologies used in existing studies, and applied person-centered latent profile models to present the direction for interventions for depression among bereaved families.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Print
