Association of Social Support With Complicated Grief and Post-Traumatic Stress Symptoms Among Student Survivors of the Sewol Ferry Disaster: An Observational Longitudinal Study Authors: Myong-Wuk Chon, Jin-Won Noh, Kyoung-Beom Kim, So Hee Lee, Eun Ji Kim Published: 2026 ABSTRACT AbstractBackgroundThis longitudinal study examined how social support has interacted with complicated grief and post-traumatic stress disorder (PTSD) symptoms among student survivors of the Sewol ferry disaster.MethodsSixty-six student survivors were recruited and evaluated biannually from 2016 to 2021. Duke-UNC Functional Social Support Questionnaire, PTSD Checklist for Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, and Inventory of Complicated Grief were used for data collection. Demographic data were treated as covariates. Group-based trajectory modeling delineated longitudinal trajectories of social support, and a series of generalized estimating equation models compared elevated PTSD symptoms and probable complicated grief among the identified trajectory groups.ResultsA four-group solution for trajectories of social support was the optimal model (group 1, low baseline social support; group 2, moderate baseline support increasing over time; group 3, high baseline support; group 4, highest baseline support). Compared to trajectory group 4, trajectory group 2 had a significantly higher risk for elevated PTSD symptoms, while trajectory groups 1 and 2 had a significantly higher risk for probable complicated grief.ConclusionOur study highlights the importance of social support and its changes relating to psychopathology including complicated grief and PTSD symptoms after disaster exposure. Individuals sensitive to changes in the level of social support who also have relatively low initial support may be particularly vulnerable.Keywords: Sewol Ferry Disaster, Surviving Students, Social Support, Complicated Grief, Post-Traumatic Stress Disorder INTRODUCTION The Sewol ferry disaster in April 2014 resulted in the death of 304 of the 476 passengers. Among the passengers were 325 Danwon High School students on a school trip. Seventy-five students survived while the rest died in the tragedy. The 75 students who survived were particularly severely affected in terms of mental health,1 and the disaster also affected the mental health of bereaved family members2 and the general public, including first responders. In total, 26.3% of the Sewol ferry disaster survivors exhibited clinical post-traumatic stress disorder (PTSD) symptoms,3 and 24.5% still suffered from complicated grief 20 months after the incident.4 The anxiety and post-traumatic symptoms experienced by the surviving students peaked around the first anniversary of the disaster and slowly decreased thereafter. However, their depression did not show a significant reduction in intensity until 27 months after the disaster. Long-term follow up studies suggest that PTSD may persist for up to 10 years after a major disaster,5 and complicated grief may last for up to 6 years in individuals bereaved by disaster.6 Considering the substantial casualties and the high prevalence of psychological distress among the survivors of the Sewol ferry disaster, investigating factors associated with their long-term mental health outcome7 may improve our understanding of the clinical trajectory of other mass disaster survivors. Among the factors associated with the development and course of post-traumatic symptoms after a major trauma, social support has been extensively researched for its potentially protective role. One meta-analysis8 examined 50 primary studies (N = 27,073) involving children and youths exposed to trauma, which revealed, although small in the effect size, the protective role of social support. This association may be longitudinal, as suggested by another meta-analysis.9 Another systematic review of individuals bereaved by sudden or violent deaths10 examined 15 original studies selected from 6,556 candidate studies involving a total of 2,553 participants, and concludes that social support might help reduce the severity of depressive as well as post-traumatic stress symptoms. Although social support is reported to affect complicated grief symptoms,11 research investigating the role of social support for complicated grief in this specific population are scant. While social support comprises multiple components and its interaction with psychopathology is also complex, studies suggest the existence of several moderating factors. The effect of social support was weaker in individuals affected by natural disasters compared to survivors of combat or interpersonal violence; the influence of social support was stronger on veterans than civilians; and the detrimental impact of negative social reactions was stronger than the protective effect of positive forms of support.12 The role of social support appears to have neurobiological basis, as suggested by its association with structural changes in the brain13 possibly resulting from its interaction with genetic predisposition.14 The results of previous cross-sectional studies that have investigated the Danwon High School survivors indicate that social support played a significant role in their subsequent mental health and quality of life.15 Recent longitudinal studies examining perceived social support and mental health16,17 highlight the temporal change in social support interacts with the clinical course of individuals exposed to disaster. This interaction can manifest as a bidirectional effect: mental health symptoms may influence the perception of support, and changes in support, in turn, may affect clinical outcomes.18 Therefore, examining the temporal relationship between social support and mental health in our participants may enhance the understanding of our participants. While our participants shared general characteristics of disaster survivors, the group has unique characteristics attributable to their composition and the nature of the disaster. Therefore, we aimed to investigate the long-term interaction between the changed perception in social support and the psychiatric outcomes of the Danwon High School survivors. Based on the established protective role of social support and the result from previous studies regarding our participants, we hypothesized that higher levels of perceived social support would be associated with less severe post-traumatic symptoms and complicated grief over time. We also anticipated a bidirectional relationship, where an increase in psychopathology would lead to a decrease in perceived social support, and vice versa. METHODS Study sample and design In the summer of 2016, approximately 2 years after the tragedy, the surviving students (who had since graduated from high school) were invited for a longitudinal evaluation as part of the Disaster Cohort Study. Psychiatrists and psychologists from the mental health center established in Danwon High School contacted the students. Of the 75 eligible students, 66 initially consented to participate. The enrolled students were evaluated twice per year (up to 12 times) and the last follow-up visit was in the winter of 2021 (Fig. 1). Sample attrition analysis in the current cohort study. Perceived social support was measured using the Duke-UNC Functional Social Support Questionnaire (FSSQ).19 The FSSQ consists of 14 items and total scores range from 14 to 70. Higher scores indicate higher levels of perceived social support. The Korean version of the FSSQ has been validated and has adequate internal consistency.20 PTSD symptoms were measured using the Post-traumatic Stress Disorder Checklist for the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (PCL-5).21 The PCL-5 consists of 20 items and total scores range from 0 to 80. The Korean translation of the PCL-5 has been validated and has good internal consistency and test-retest reliability.22 Elevated PTSD symptoms were defined as PCL-5 scores ≥ 33, consistent with prior validation studies.22 Supplementary analyses using alternative cutoffs (28 and 30) were performed to evaluate the stability of results. These thresholds were selected based on prior studies23,24 demonstrating their utility in identifying elevated PTSD symptoms in trauma-exposed populations. Complicated grief was measured using the Inventory of Complicated Grief (ICG).25 The ICG consists of 19 items and total scores range from 0 to 76. A score of 25 suggests a level of grief likely to cause significant social, mental, and physical impairments.26 The Korean translation of the ICG has been validated and has good internal consistency.27 The covariates in the present study were sex, birth year, mental health service use, type of health insurance, and co-residence with parents. Sex was dichotomized as male and female, and birth year as 1997 and 1998. Mental health service use was indicated by the use of at least one of the following services: counseling center, religious center, mental health clinic, and discussions with a neighbor or friend. Type of health insurance was dichotomized as national health insurance and others; this latter category included both medical program beneficiaries and respondents who refused to answer. The South Korean medical aid program is comparable to the Medicaid program in the United States28 and serves as a proxy measure for poverty. Co-residence with parents was dichotomized as yes or no. Statistical analysis Descriptive analysis of the participants’ baseline characteristics and clinical data were generated (mean and standard deviation for numerical data and frequencies and percentages for categorical data). Univariate linear regression and Fisher’s exact test were performed to analyze numerical and categorical data, respectively. To identify distinct trajectories of social support, we employed Nagin’s group-based trajectory modeling (GBTM) framework.29 The optimal number of trajectory groups was determined by comparing models with two to eight groups, and the model with the smallest Bayesian Information Criterion (BIC) index was preferred, as BIC is recognized for its ability to balance model fit and complexity and to reduce the risk of overfitting, particularly in studies with limited sample sizes.30,31 Final model selection was followed by Burnham and Anderson’s guideline,32 where models with ΔBIC < 2 were considered equivalent in predictive accuracy. Model selection prioritized clinical interpretability when BIC differences fell within this threshold. For each candidate model, we tested combinations of polynomial functions to determine the optimal functional form for each trajectory. A trajectory group was considered statistically significant if all polynomial coefficients for that group were significant at P < 0.05. Model adequacy was further evaluated using diagnostics: a) the average posterior probability for each group exceeded 0.7, indicating good classification reliability, b) and the difference between the estimated and assigned group membership proportions was less than 50%. We utilized all available assessment data in the trajectory estimation despite the nature of the unbalanced panel, since the GBTM can handle missing data by fitting the model using maximum likelihood estimation. A series of generalized estimating equation (GEE) models were fitted to investigate the relationship between social support over time and factors associated with significant PTSD symptoms and complicated grief. Given the same exposure but different enrollment timepoints, the longitudinal analysis was conducted based on external time, which is the time elapsed since the disaster occurred. The Huber–White sandwich estimator33 was used to derive the heteroskedasticity-robust standard error. Two-tailed P values < 0.05 were considered statistically significant. All statistical analyses were performed using Stata/MP software (ver. 16.1; Stata Corp., College Station, TX, USA). Ethics statement The study protocol was reviewed and approved by the Institutional Review Boards of The Catholic University of Korea Seoul St. Mary’s Hospital (approval No. KC15OIMI0261) and the National Medical Center (approval No. H-1505–054–002). Written consent was obtained from all participants, who were invited for this cohort two years after the tragedy after they grew up to be adults eligible to give consents. The participants were informed that they could withdraw from the study at any time. RESULTS Baseline characteristics and probable diagnosis classifications Table 1 summarizes the sociodemographic characteristics and probable diagnosis classifications of the participants at baseline that enrolled between 2016 (two years from disaster) and 2021 (seven years from disaster). Baseline characteristics for participants who were followed up in all waves provided in Supplementary Table 1. In total, 66 and 59 participants completed the PCL-5 and ICG questionnaires, respectively. Of the 66 participants who completed the PCL-5, 7.6% (n = 5) were classified into the probable PTSD group. Of the 59 participants who completed the ICG, 18.6% (n = 11) were classified into the probable complicated grief group. The mean FSSQ total score of the 66 participants included in the study was 55.2. The probable PTSD and CG classification at baseline by FSSQ trajectory over time is reported in Supplementary Tables 2 and 3. The changes in PTSD, CG and FSSQ over time by FSSQ trajectory group is provided in Supplementary Table 4. Table 1Sociodemographic characteristics and probable diagnosis classification of study population at baselineVariablesPTSD symptoms (PCL-5)Complicated grief (ICG)Lowera (n = 61)Elevateda (n = 5)Totalb (N = 66)P valueNon-probable CGa (n = 48)Probable CGa (n = 11)Totalb (N = 59)P valueSex0.3660.316Male29 (96.7)1 (3.3)30 (45.5)23 (88.5)3 (11.5)26 (44.1)Female32 (88.9)4 (11.1)36 (54.5)25 (75.8)8 (24.2)33 (55.9)Birth year0.1070.604199755 (94.8)3 (5.2)58 (87.9)43 (82.7)9 (17.3)52 (88.1)19986 (75.0)2 (25.0)8 (12.1)5 (71.4)2 (28.6)7 (11.9)Mental health service utilization0.3250.710Yes45 (90.0)5 (10.0)50 (75.8)34 (79.1)9 (20.9)43 (72.9)No16 (100.0)0 (0.0)16 (24.2)14 (87.5)2 (12.5)16 (27.1)Type of health insurance0.5970.006* National health insurance15 (88.2)2 (11.8)17 (25.8)9 (56.3)7 (43.8)16 (27.1)Others46 (93.9)3 (6.1)49 (74.2)39 (90.7)4 (9.3)43 (72.9)Co-residence with parents0.0830.310Yes56 (94.9)3 (5.1)59 (89.4)44 (83.0)9 (17.0)53 (89.8)No5 (71.4)2 (28.6)7 (10.6)4 (66.7)2 (33.3)6 (10.2)FSSQ (min = 31; max = 70)55.5 ± 11.252 ± 12.655.2 ± 11.30.52954.4 ± 11.451.5 ± 10.253.9 ± 11.10.399FSSQ trajectory over time0.9990.182Group 4 (highest baseline)13 (92.9)1 (7.1)14 (21.2)9 (100.0)0 (0.0)9 (15.3)Group 3 (higher baseline)16 (94.1)1 (5.9)17 (25.8)15 (88.2)2 (11.8)17 (28.8)Group 2 (moderate baseline)22 (91.7)2 (8.3)24 (36.4)15 (68.2)7 (31.8)22 (37.3)Group 1 (lower baseline)10 (90.9)1 (9.1)11 (16.7)9 (81.8)2 (18.2)11 (18.6)Values are presented as number (%) or mean ± standard deviation.Subgroup comparisons were performed using Fisher’s exact tests for categorical data and univariable linear regression for numerical data.PTSD = post-traumatic stress disorder, PCL-5 = Post-traumatic Stress Disorder Checklist for the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders, ICG = Inventory of Complicated Grief, FSSQ = Duke-UNC Functional Social Support Questionnaire.aRow percentage was presented; bColumn percentage was presented.*P < 0.05. Sociodemographic characteristics and probable diagnosis classification of study population at baseline Values are presented as number (%) or mean ± standard deviation. Subgroup comparisons were performed using Fisher’s exact tests for categorical data and univariable linear regression for numerical data. PTSD = post-traumatic stress disorder, PCL-5 = Post-traumatic Stress Disorder Checklist for the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders, ICG = Inventory of Complicated Grief, FSSQ = Duke-UNC Functional Social Support Questionnaire. aRow percentage was presented; bColumn percentage was presented. *P < 0.05. Social support trajectories Fig. 2 shows the social support trajectory groups. The three-group (BIC = −1,789.69) and four-group (BIC = −1,789.37) solution models showed a minimal BIC difference (ΔBIC = 0.32), indicating statistical equivalence threshold (ΔBIC < 2). Despite this, the four-group solution was selected due to its superior clinical interpretability: 1) Distinct trajectories aligned with prior disaster survivor studies. 2) Each subgroup represented meaningful patterns of social support dynamics. A four-group solution consisting of two linear functions and two quadratic functions was determined as the optimal model through GBTM. This model had the largest information criteria indices except for the models with five or more groups, which included at least one statistically non-significant trajectory group (Supplementary Tables 5 and 6). The average posterior probability was 0.89, 0.90, 0.86, and 0.91 for trajectory groups 1–4, respectively, which indicates that all four trajectory groups had good internal reliability. Trajectory group 1 was characterized by a low level of social support (n = 11 [16.6%]; FSSQ score at baseline = 37.4). Trajectory group 2 was characterized by a moderate level of perceived support and an increase therein over time (n = 24 [36.4%]; FSSQ score at baseline = 52.3). Trajectory group 3 was characterized by a high level of social support that persisted over time (n = 17 [25.8%]; FSSQ score at baseline = 61.8). Finally, trajectory group 4 was characterized by the highest level of social support (n = 14 [21.2%]; FSSQ score at baseline = 65.4). All four trajectories of social support generally increased over time, but there were clear differences in the level of support at baseline. Four-group solution for trajectories of social support. FSSQ = Duke-UNC Functional Social Support Questionnaire. Trajectories of social support and factors associated with post-traumatic stress and complicated grief Table 2 summarizes the results of two GEE models that investigated the relationship between social support trajectories and factors related to elevated PTSD symptoms and probable complicated grief. Trajectory group 2 (moderate social support at baseline) had a significantly higher odds of elevated PTSD symptoms (odds ratio [OR], 3.693; 95% confidence interval [CI], 1.023–13.334; P = 0.046) compared with trajectory group 4 (highest level of social support at baseline). While the primary analysis used the standard PCL-5 cutoff of 33, supplementary analyses with lower cutoffs demonstrated consistent patterns of elevated PTSD symptoms in trajectory group 2 (Supplementary Table 7). Trajectory group 1 (low level of social support at baseline; OR, 8.192; 95% CI, 1.664–40.333; P = 0.010 and trajectory group 2 (moderate level of social support at baseline; OR, 9.000; 95% CI, 2.063–39.268; P = 0.003) had a significantly higher risk for probable complicated grief compared to trajectory group 4 (highest level of social support at baseline). In terms of sociodemographic covariates, females had a significantly higher risk of elevated PTSD symptoms (OR, 5.664; 95% CI, 2.250–14.255; P < 0.001) and complicated grief (OR, 2.751; 95% CI, 1.060–7.141; P = 0.038). Participants born in 1998 had a significantly higher risk of probable complicated grief (OR, 4.807; 95% CI, 1.504–15.367; P = 0.008). Participants not cohabiting with their parents had a significantly higher risk of elevated PTSD symptoms (OR, 2.127; 95% CI, 1.080–4.189; P = 0.029). Table 2Relationship between trajectories of social support over time and post-traumatic stress or complicated grief symptomsVariablesElevated PTSD symptoms (PCL-5) (n = 66)Probable complicated grief (ICG) (n = 59)ORRobust SEP valueLLULORRobust SEP valueLLULFSSQ trajectory over timeGroup 4 (highest baseline)1.000Ref1.000RefGroup 3 (higher baseline)1.5751.2520.5670.3327.4792.6912.5310.2930.42617.005Group 2 (moderate baseline)3.6932.4190.046* 1.02313.3349.0006.7650.003* 2.06339.268Group 1 (lower baseline)3.5122.8930.1270.69917.6498.1926.6620.010* 1.66440.333SexMale1.000Ref1.000RefFemale5.6642.667< 0.001* 2.25014.2552.7511.3390.038* 1.0607.141Birth year19971.000Ref1.000Ref19982.8141.5030.0530.9888.0174.8072.8500.008* 1.50415.367Mental health service utilizationYes1.000Ref1.000RefNo0.5910.1950.1110.3091.1281.0630.5290.9020.4012.821Type of health insuranceNational health insurance1.000Ref1.000RefOthers0.9230.2710.7860.5191.6430.7800.2450.4300.4221.445Co-residence with parentsYes1.000Ref1.000RefNo2.1270.7350.029* 1.0804.1891.5830.6910.2920.6733.722PTSD = post-traumatic stress disorder, PCL-5 = Post-traumatic Stress Disorder Checklist for the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders, ICG = Inventory of Complicated Grief, OR = odds ratio, SE = standard error, LL = lower limit of 95% confidence interval, UL = upper limit of 95% confidence interval, FSSQ = Duke-UNC Functional Social Support Questionnaire, ref = reference group.*P < 0.05. Relationship between trajectories of social support over time and post-traumatic stress or complicated grief symptoms PTSD = post-traumatic stress disorder, PCL-5 = Post-traumatic Stress Disorder Checklist for the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders, ICG = Inventory of Complicated Grief, OR = odds ratio, SE = standard error, LL = lower limit of 95% confidence interval, UL = upper limit of 95% confidence interval, FSSQ = Duke-UNC Functional Social Support Questionnaire, ref = reference group. *P < 0.05. DISCUSSION PTSD symptoms and complicated grief of student survivors of the Sewol ferry accident were influenced by the perceived level of social support, indicating an important role of such support in the long-term psychiatric outcomes of disaster survivors. Female gender and younger age were associated with increased severity of probable PTSD and complicated grief symptoms, as in previous studies.34 Living with parents was related to decreased severity of posttraumatic symptoms35 but not complicated grief (Table 2). Although there are several shipwreck catastrophes in the twenty-first century with comparable casualties,36,37 the significance of Sewol Ferry tragedy is unique since it is the only maritime disaster of its magnitude in which a unique group of students from one single closely connected group of homogeneous classmates were sacrificed. Because our participants survived a life-threatening event but also lost many close classmates,38 they had clinical characteristics typical of both disaster survivors and bereaved people. The key results are discussed below in light of this. First, the level of social support in all trajectory groups improved over time, which may be explained by the beneficial effects of compensation schemes including a long-awaited court decision which admitted state culpability and ordered a financial compensation averaging approximately 530,000 USD per victim,39 and policies implemented in the wake of the disaster, which attracted national concern.40 For example, a volunteer group of 72 psychiatrists offered psychological first aid, psychoeducation, group therapy and anxiety reduction interventions as well as mental health screening.41 In other studies42,43 that have explored the longitudinal course of social support, consistent, increasing, and decreasing levels of support were all observed. In trajectory group 2 in this study, social support increased rapidly over time. Second, groups 1 and 2, which had the lowest levels of social support, had significantly more severe complicated grief symptoms, which accords with previous studies of bereaved youths.44 Quality of life, which is associated with the relationship with parents, also appeared to be related to complicated grief in previous studies2 of the same participants at the time of graduation (i.e., 2 years after the disaster). Third, group 2 participants, who experienced a moderate level of baseline social support, showed significantly more severe PTSD symptoms than group 4 did, while group 1 (low baseline support) showed a trend toward an increase in PTSD symptoms. These results accord with previous studies of the same students, in which low peers, parental and social support as well as lower autonomy and negative emotions were correlated with probable PTSD symptoms3 as well as complicated grief.4 Notably, in our analysis trajectory group 2 showed a peculiar longitudinal pattern in terms of their perceived social support level compared to other groups. Participants of group 2 experienced moderate level of initial social support, and during the follow-up period they report improvement in the support received, so that the self-reported level of support of trajectory groups 2 and 3 converged over time. However, the severity of psychopathology in terms of post-traumatic symptoms and complicated grief in group 2 were comparable to group 1, who experienced relatively lower support throughout. This implies that, in addition to low support itself, unmet need for support might also be a potential vulnerability factor. Although disaster-exposed individuals experience changes in their level of distress or grief during follow-up45,46 changes in perceived social support level in such victims are largely unavailable. Our result may indicate that future studies might need to measure the change in social support over time. While group 2 participants were at more risk for posttraumatic symptoms, participants of both groups 1 and 2 showed increased risk for complicated grief. Although in traumatically bereaved people posttraumatic symptoms and grief are closely related, both similarity and disparity exist between the two phenomena.47 Participants in group 1 may have had consistently low expectations for social support, thus their perceived lack of support, while associated with complicated grief, may have been less salient in driving posttraumatic symptoms. Alternatively, it may be the mismatch between high expectation for support and insufficient initial support could have constituted a stronger risk for posttraumatic symptoms than consistent low support did. Additionally, younger participants were more vulnerable to complicated grief. This may be understood in light of psychosocial and developmental factors affecting our adolescent participants.48 Although some meta-analyses have suggested that younger age at trauma exposure could be a risk for adverse outcomes, findings across populations have been inconsistent and the predictive value of age was variable depending on study context and methodology.49 In our study, participants were from the same school grade and the age difference was within one year. This might have limited the influence of age difference as a determinant of vulnerability. Those not living with parents were more likely to have PTSD symptoms. The absence of parental cohabitation may have limited immediate emotional support and access to a stable caregiving environment, thereby exacerbating vulnerability to posttraumatic psychopathology.50,51 Females were more vulnerable to both. Gender differences have also been consistently reported, with females showing greater susceptibility to trauma-related distress, possibly due to biological stress reactivity as well as social and cultural factors that shape emotional processing.52 Studies reporting that women might be less likely to use positive coping strategies, and more sensitive to threats and tended to interpret disasters more negatively.53,54 This is also in line with our previous reports regarding the same subjects.3,4,15 The findings of this study have some practical implications for mental health interventions and long-term support for disaster survivors. First, the results underscore the need for a proactive and sustained approach to providing social support, especially for adolescents who may be particularly vulnerable. Support systems should be designed to monitor and adapt to the changing needs of survivors over time, rather than ceasing after the initial acute phase. The finding that the subgroup with initially moderate support (group 2) remained vulnerable highlights that even when overall support appears to improve, there might be a potential mismatch between perceived support and its actual benefit, and individual’s specific emotional state and clinical trajectory requires attention. Our study also highlights that key demographic vulnerabilities including female gender and younger age as well as family-related factors might be considered when designing targeted, personalized interventions for different subgroups of survivors. This study had some limitations. The primary limitation of this study is the relatively small sample size, which was an inevitable constraint given that the study population comprised student survivors from the Sewol ferry disaster. The small sample size may affect the stability and generalizability of the results, particularly when employing complex statistical approaches such as GBTM. Specific methodological challenges include potential instability in trajectory group classification, reduced precision in effect size estimates, and limitations in the reliability of posterior probability calculations. Therefore, these findings should be regarded as exploratory, and future validation studies with larger samples are warranted to more conclusively determine the longitudinal effects of social support on complicated grief and post-traumatic stress symptoms among disaster survivors. Nevertheless, we made considerable efforts to enroll most eligible survivors and to maintain long-term follow-up. Notably, our enrollment rate (89.3%) exceeds typical in longitudinal disaster studies.55,56,57 This high enrollment rate enhances the representativeness of our findings within this unique survivor cohort. Second, because the participants received psychiatric care only after the incident, their mental health status before the disaster is unknown. Third, social support was measured based on subjective self-report measures. Fourth, data regarding the level of psychiatric service or national support for each individual is rather limited, thereby the association between FSSQ scores and the actual practical support could not be investigated. Finally, data were not available for the period between the accident and study enrollment, such that only limited interpretation is available for the possible impact of social support before the first assessment,58 and the students’ emotional reactions to the anniversary of the disaster were not assessed.59,60 This limitation happened because originally the Ministry of Education took care of the participants until they reached adulthood when they referred them to the ministry of Health and Welfare, and only at this time the researchers could access and recruit the participants. In conclusion, this study demonstrates that both the level and change of social support are critical in understanding post-trauma outcomes among adolescent disaster survivors. While consistently low support is a clear risk for complicated grief, discrepancies between expected and actual support may constitute a stronger vulnerability factor for PTSD. These results underscore the need for sustained, adaptive, and tailored interventions to support vulnerable subgroups in the aftermath of disasters. Footnotes Funding: The authors gratefully acknowledge the National Medical Center for providing the overall research environment and funding the publication fees for this study. Disclosure: The authors have no potential conflicts of interest to disclose. Author Contributions: Conceptualization: Chon MW, Noh JW, Kim KB, Lee SH, Kim EJ.Data curation: Chon MW.Formal analysis: Noh JW, Kim KB.Funding acquisition: Lee SH.Investigation: Chon MW, Kim KB.Methodology: Noh JW.Project administration: Lee SH, Kim EJ.Resources: Lee SH.Software: Noh JW, Kim KB.Supervision: Lee SH.Validation: Kim EJ.Visualization: Noh JW.Writing - original draft: Chon MW, Noh JW.Writing - review & editing: Kim KB, Lee SH, Kim EJ. Conceptualization: Chon MW, Noh JW, Kim KB, Lee SH, Kim EJ. Conceptualization: Chon MW, Noh JW, Kim KB, Lee SH, Kim EJ. Data curation: Chon MW. Data curation: Chon MW. Formal analysis: Noh JW, Kim KB. Formal analysis: Noh JW, Kim KB. Funding acquisition: Lee SH. Funding acquisition: Lee SH. Investigation: Chon MW, Kim KB. Investigation: Chon MW, Kim KB. Methodology: Noh JW. Methodology: Noh JW. Project administration: Lee SH, Kim EJ. Project administration: Lee SH, Kim EJ. Resources: Lee SH. Resources: Lee SH. Software: Noh JW, Kim KB. Software: Noh JW, Kim KB. Supervision: Lee SH. Supervision: Lee SH. Validation: Kim EJ. Validation: Kim EJ. Visualization: Noh JW. Visualization: Noh JW. Writing - original draft: Chon MW, Noh JW. Writing - original draft: Chon MW, Noh JW. Writing - review & editing: Kim KB, Lee SH, Kim EJ. Writing - review & editing: Kim KB, Lee SH, Kim EJ. SUPPLEMENTARY MATERIALS Supplementary Table 1 Sociodemographic characteristics and probable diagnosis classification of the study population at baseline among those followed up for six years Supplementary Table 2 Baseline characteristics and probable PTSD classification of study population by FSSQ trajectory over time Supplementary Table 3 Baseline characteristics and probable CG Classification of study population by FSSQ trajectory over time Supplementary Table 4 PTSD, CG, and FSSQ changed over the study period by FSSQ trajectory group Supplementary Table 5 Selection result for optimal number of trajectory group Supplementary Table 6 Selection result for optimal trajectory shapes Supplementary Table 7 Relationship between trajectories of social support over time and alternative PCL-5 cutoff scores References Associated Data Supplementary Materials Supplementary Table 1 Sociodemographic characteristics and probable diagnosis classification of the study population at baseline among those followed up for six years Supplementary Table 2 Baseline characteristics and probable PTSD classification of study population by FSSQ trajectory over time Supplementary Table 3 Baseline characteristics and probable CG Classification of study population by FSSQ trajectory over time Supplementary Table 4 PTSD, CG, and FSSQ changed over the study period by FSSQ trajectory group Supplementary Table 5 Selection result for optimal number of trajectory group Supplementary Table 6 Selection result for optimal trajectory shapes Supplementary Table 7 Relationship between trajectories of social support over time and alternative PCL-5 cutoff scores