Association between traumatic events with suicidality among adolescents: A large-scale cross-sectional study of 260,423 participants

,

Suicide stands as a prominent concern among the health-related outcomes associated with traumatic events, serving as a warning signal of early intervention failure and a breakdown of mental wellbeing (Boduszek et al., 2021;Devries et al., 2022;Guo et al., 2018).Traumatic events, particularly those associated with abuse and violence, known as interpersonal violence (IPV; the intentional use of physical force or power against other persons) (Devries et al., 2014;Dworkin et al., 2022;Liu et al., 2018;Maniglio, 2011), have been linked to a range of suicidality outcomes (Angelakis et al., 2020;Paul et al., 2016;Zatti et al., 2017).For instance, a recent meta-analysis revealed the pooled prevalence of any suicidal ideation or attempts in individuals exposed to sexual assault was 27.25 %, compared to 9.37 % in those without such experiences (Dworkin et al., 2022).
However, research on early life traumatic events has primarily focused on the traditional definition of adverse childhood experiences (ACEs), which mainly centers around household dysfunction.There has been limited attention paid to adverse experiences outside the household, such as disasters, accidents, and community violence (Björkenstam et al., 2017;DeVille et al., 2020).Moreover, prior research has predominantly focused on long-term consequences of early life traumatic events in adulthood (Angelakis et al., 2019;Huffmaster et al., 2022;Liu et al., 2021;Tran et al., 2022), However, there is a need to redirect the focus towards prevention and minimizing the adverse impacts of traumatic events during early life to address the gap in our understanding of the early psychological associations of such events (Narayan et al., 2021).Additionally, most studies have concentrated solely on specific traumas or cumulative trauma numbers (Devries et al., 2022), while a comprehensive view of associations between traumatic patterns and suicidality within population-based samples remains lacking.Furthermore, limited research has explored the role of early life traumatic events in the development of suicidality, considering important stages such as suicidal ideation, plan, and attempt.
Given these considerations, our study is driven by the following research question: the extent to which early life traumatic events are associated with the development of suicidality in adolescents and how different patterns of trauma exposure, along with the dose-response relationship, play a role in this phenomenon.Our study aims to contribute novel insights and provide a holistic perspective on the early associations between traumatic events and suicidality in adolescents.Additionally, our study seeks to identify vulnerable adolescents and investigate modifiable demographic factors, informing targeted interventions and preventive strategies.

Study population
The data for this study were collected during the annual regional school-based mental health screening in September 2021 in Deyang, a city situated in the southwestern province of Sichuan, China.The study involved students from upper grades (grades 5-6) in all primary schools, middle schools, and high schools in Deyang.The survey took place in the school's computer room settings, and participants answered the questions through the 'Online School Student Mental Health Management System', which was specifically designed for this series of school-based mental health survey projects in Sichuan Province.The questionnaire began with a section for collecting basic demographic information (including student ID for response tracking) and was followed by a series of psychological measures.On average, each student took about 8 min to complete the questionnaire.Teachers received training from the research team to provide students with basic clarifications on questionnaire items, ensuring comprehension without influencing the students' independent responses.Participants and their guardians completed informed consent along with the electronic questionnaire, and the study received ethical approval from the institutional review board of West China Hospital, Sichuan University (2022University ( -1709)).
In this cross-sectional study, the questionnaire comprised compulsory questions, ensuring a complete dataset with no missing values.The exclusion criteria applied to individuals who were absent on the assessment day, those who faced challenges in understanding specific questionnaire items despite teachers' basic interpretation, or those who did not complete the entire questionnaire.The sample consisted of 260,423 participants (predominantly Chinese Han ethnicity), providing a representative sample of the general population of children and adolescents in China.

Assessment of traumatic events
Traumatic events were evaluated by the UCLA Posttraumatic Stress Disorder Reaction Index, which is specifically designed for children and adolescents population according to the DSM-IV criteria (Steinberg et al., 2004).This tool, previously employed and validated in Chinese adolescent populations (Li et al., 2019), was used to evaluate the presence or absence of 13 traumatic events in our study.According to the previous study (Darnell et al., 2019) and the characteristics of this sample, 13 traumatic events in the original data merged into 8 traumatic events to reduce double-counting: (1) natural disaster; (2) bad accident; (3) community violence; (4) (witnessing) domestic violence; (5) (domestic) physical abuse; (6) sexual abuse; (7) death/serious injury of a loved one (DSILO), and (8) painful medical treatment.Each traumatic event was categorized using a binary (yes or no) response, creating individual variables for each trauma type.Additionally, we calculated a cumulative number of trauma types, represented both continuously and in categories (0, 1, 2, 3, ≥4), to examine the dose-response relationship.Furthermore, we identified distinct patterns of trauma with latent class analysis (LCA), to provide deeper insight into the co-occurrence of trauma types and their collective relationships with the outcome.Ultimately, we aimed to use this multifaceted approach to provide a comprehensive picture of how trauma is experienced and how it affects the targeted population.

Assessment of suicidality
Suicidality was measured using the Suicide Behavior Questionnaire-Revised (SBQ-R), a brief screening tool for suicide ideation and behavior, which provides total scores ranging from 3 to 18, with higher scores indicating increased levels of suicidality.In the nonclinical population, a total score of 7 is commonly used as the cut-off point (Osman et al., 2001).The Chinese version of the instrument previously showed satisfactory reliability, validity, and factor structures within the adolescent populations (Huen et al., 2022;Xu et al., 2018), and the tool showed good internal consistency reliability (Cronbach's alpha = 0.81) in our study.The questionnaire yielded three outcome variables: SBQ-R score as a continuous variable, low/high suicidality group (defined as SBQ-R score < 7/≥ 7) as a binary variable, and three types of suicidality (with suicidal ideation or not, with suicidal plan or not, with suicidal attempt or not) as binary variables based on the first question.

Covariates
Demographic information associated with both traumatic events and suicidality was collected as covariates: gender, age, residence (living in urban or rural areas), only-child status (yes or no), left-behind child status (yes or no), family structure (parents staying together, parental separation or parental death), parental education (represented by higher degree from either of the parents: primary school and below, middle school, high school, undergraduate and above, or unknown), family income (considering the local living standards < ¥3000/month, ¥3000-6000/month, ¥6000-10,000/month, ¥10,000-20,000/month, > ¥20,000/month).To prevent residual confounding (Groenwold et al., 2013), age was transformed from a continuous variable into age groups (≤ 12, 13-15, ≥ 15) according to the Chinese school system and adjusted as a categorical variable in all models.

Statistical analysis
Descriptive characteristics were compared between groups using ANOVAs for continuous data (described with mean and standard deviation [SD]) and Chi-square tests for categorical data.
LCA was performed to extract the latent traumatic pattern for each participant based on their response to trauma events with the poLCA package in R (Linzer and Lewis, 2011;Zhang et al., 2018).Adjusted X. Shi et al.Akaike information criteria (aAIC), adjusted Information Criteria BIC (aBIC), likelihood ratio/deviance statistic, and entropy were used to determine the best number of classes, with lower information criteria (IC) and likelihood ratio/deviance statistic and higher entropy indicating better model.Since more classes often mean a decrease in IC, favoring the more complex model, an elbow plot was used to identify a point of inflection (Sinha et al., 2021).Posterior probability was calculated to represent the average probability of individuals being assigned to each class, and therefore it helped to identify and give a name to each traumatic pattern (latent class).As Sichuan is a region that is susceptible to natural disasters, and the frequency and severity of these disasters can vary over time, a sensitivity analysis was performed.A new model was developed, where natural disasters were excluded from the original LCA model, to assess the consistency of the extracted patterns between the two models.
For continuous and binary outcomes, linear regression models assuming normally distributed errors (β with 95 %CI) and log-linear Poisson regression models with robust error variance (prevalence ratios [PR] with 95 %CI) were used to examine the associations between traumatic events and suicidality, respectively.In all analyses, model 1 was adjusted for age group and gender, model 2 was additionally adjusted for residence, only child status, left-behind child status, family structure, parental education, and family income.When analyzing associations between each traumatic event and suicidality, all traumatic events simultaneously entered into model 3 and meanwhile adjusted for all the covariates mentioned in model 1 and model 2, considering the cooccurrence feature of traumatic events (Daníelsdóttir et al., 2022).
All statistical analyses were performed using R (version 4.1.1).All P values were 2-tailed and a P value of less than 0.05 was considered statistically significant.First, we examined the associations between each traumatic event and suicidality, with models further stratified by demographic statistics, to better compare associations among demographic subgroups.Secondly, associations between the number of trauma types and suicidality were assessed and stratified as above.Thirdly, associations between traumatic patterns and suicidality were analyzed.

Descriptive characteristics
A total of 260,423 participants (48.4 % female) were included in the study, with a mean age of 13.0 years (SD: 2.69).Ethnic distribution included 96.65 % Han, 1.63 % Yi, 0.71 % Tibetan, and 1.72 % other ethnicities.The majority of participants (57.8 %) reported traumatic exposure.The average SBQ-R score was 4.68 (SD: 2.81).Among the participants, 18.7 % were classified into the high suicidality group, with 29.2 % reporting suicidal ideation, 10.2 % reporting a suicidal plan, and 2.5 % reporting a suicidal attempt.Participants with higher suicidality were more likely to be female, older, living in urban areas, have siblings, be left-behind children, have experienced parental separation or death, have relatively less-educated parents, and have lower family income (Table 1).

Associations between each traumatic event and suicidality
Among all the traumatic events examined in the study, community violence, domestic violence, physical abuse, and sexual abuse were most strongly associated with elevated suicidality, with physical abuse showing the strongest association (Fig. 1).Compared to participants unexposed to physical abuse, the SBQ-R score of participants exposed to it increased by 2.51 (95 %CI: 2.48-2.54;model 2), and they were more likely categorized into the high suicidality group (PR: 2.97, 95 %CI: 2.93-3.02;model 2; eTable 1 in Supplementary Materials).Similarly, ) in all participants.Notably, we observed significant associations between physical abuse and specific stages of suicidal behavior within subgroups.In individuals with suicidal ideation, physical abuse was strongly associated with an increased likelihood of developing a suicidal plan (PR: 1.7, 95 %CI: 1.67-1.73).Similarly, among those with a suicidal plan, physical abuse was strongly associated with an elevated risk of attempting suicide (PR: 1.31, 95 %CI: 1.26-1.37;eTables 2-7 in Supplementary Materials).Associations of traumatic events and suicidality were more obvious in females, participants aged 13-15, and those with parents staying together.Participants in urban areas were more likely to have elevated suicidality than those in rural areas when exposed to bad accidents, domestic violence, and community violence.Participants as the only child in their family were more likely to have elevated suicidality than those having siblings when exposed to a bad accident and community violence.Participants as left-behind children were more sensitive when exposed to natural disasters, community violence, sexual abuse, and DSILO in terms of suicidality.When exposed to most traumatic events, the associations were more evident in those with the highest and lowest socioeconomic status considering the level of parental education and family income, forming a U-shaped curve (eFig. 1 in Supplementary Materials).

Associations between traumatic patterns and suicidality
As aAIC, and aBIC saw a point of inflection and there was a relatively greater entropy, the four-class model was the optimal one (eFigure 4 in Supplementary Materials), and patterns were extracted as follows: (1) low traumas pattern: relatively low probabilities of all traumatic events; (2) high DSILO pattern: high probabilities of DSILO; (3) high physical abuse pattern: high probabilities of physical abuse; (4) multiple traumas pattern: high probabilities of all traumatic events (eFigure 5 in

Supplementary Materials
).The categorization of latent patterns before and after excluding natural disasters from the traumatic events list was approximately similar (eFigure 6-8 in Supplementary Materials), therefore further analysis was conducted with the patterns derived from all 8 traumatic events.Compared to individuals from the low traumas pattern, those from the other three patterns showed higher suicidality, especially for participants from the multiple traumas pattern (β: 3.68,PR: 5.12,.The high physical abuse pattern was more associated with high suicidality than the high DSILO pattern (β: 2.59 [95 % CI, 2.56-2.62]vs 1.08 [95 %CI,; PR: 4.05 [95 %CI,) (Table 3 and eTable 9 in Supplementary Materials).Associations after excluding natural disasters from the traumatic events list were shown in eTable 10-11 in Supplementary materials.Furthermore, associations between traumatic patterns and suicidality in each demographic group were demonstrated in Fig. 2. In all traumatic patterns, the percentage of females with high suicidality was higher than that of males.Except for the low traumas pattern, participants aged 13-15 had a higher percentage of participants with high suicidality than other age groups.In the low traumas pattern, the highest proportion of individuals with high suicidality was in the eldest age group (≥16 years).

Discussion
In this large sample research of Chinese adolescents, we used a comprehensive measure of traumatic events and suicidality, and it revealed that traumatic exposure showed early associations with higher suicidality in a dose-dependent manner.Our main findings included: (1) Traumatic events related to interpersonal violence, particularly physical abuse, were most strongly associated with suicidality; (2) The incremental suicidality was more obvious as the number of traumatic event types increased; (3) We identified four distinct traumatic patterns within the study population, and individuals falling into the multiple traumas pattern demonstrated the highest likelihood of experiencing elevated suicidality; (4) Traumatic exposure showed stronger associations with suicidality in females, participants aged 13-15, urban residents, only children, and left-behind children, compared to weaker associations in participants with intact parental relationships and middle socioeconomic status.

Associations between traumatic events and suicidality
Previous studies have investigated the links between abuse, violence, and suicidality separately (Castellví et al., 2017;Mercy et al., 2017), we innovatively integrated them into a unified construct as IPV, to highlight its crucial role in adolescent suicidality development comprehensively.Its repetitive and persistent nature stemming from its origins in family and community environments makes it stand out from other traumatic events, leading to a decrease in the fear of death and a greater likelihood of suicide as a perceived escape route (Chu et al., 2017).As a common IPV, physical abuse showed the strongest association with higher suicidality, likely attributed not only to its strong correlation with a reduced fear of death but also to an increased tolerance for physical pain, both critical facets of acquired capability for suicide in the interpersonal-psychological theory of suicide (Stewart et al., 2017).Additional contributing factors may include the mediation of clinically elevated attention and depressive/anxious problems (Paul and Ortin, 2019) and exacerbated stigmatization and internal attribution led by physical abuse (Feiring et al., 2002).Our study findings reveal a strong association between IPV with suicidal ideation, plan, and attempt, which suggests that physical abuse may contribute to an increased risk of transitioning from having thoughts of self-harm to formulating specific plans to carry out these intentions.To cope with it, physical activity is a practical approach to mitigate IPV's impact on suicidality, by aiding physical recovery and equipping victims with coping mechanisms to reduce psychological repercussions (Donofry et al., 2021;Jiang et al., 2022).
Previous literature has established a graded association between traumatic exposure and adolescents' mental health (Daníelsdóttir et al., 2022;Hughes et al., 2019).Our study extends this understanding by revealing that the impact of trauma on suicidality is not simply additive but multiplicative.The slope of the suicide risk score increased as the number of traumatic event types increased, indicating a wider range of traumatic exposure is connected with considerably devastating impacts.This observation was further supported by our finding that the multiple traumas pattern was linked with the highest levels of suicidality.One possible explanation is this relationship between exposure to multiple traumatic events and suicidality could be mediated by posttraumatic hyperarousal (Briere et al., 2015), where the accumulation of traumatic experiences intensifies post-trauma symptoms, leading to a heightened risk of suicidality (Panagioti et al., 2015).Another explanation is that exposure to multiple traumas may lead to a sense of helplessness and a perceived lack of control, factors often associated with depression and suicidal thoughts (Hammack et al., 2012).The cumulative impact of trauma can also erode an individual's coping mechanisms and resilience, making it harder to deal with subsequent stressors effectively (Hamby et al., 2021).
In addition to identifying low and multiple trauma patterns, similar to previous studies (Li et al., 2021;Xiao et al., 2022), our research introduces two novel patterns: the high physical abuse pattern and the high DSILO pattern.The high physical abuse pattern featured high exposure to IPV (especially physical abuse), while the high DSILO pattern featured high exposure to DISLO and natural disasters (probably because participants were recruited from Sichuan, an area where natural disasters are more frequent).This four-pattern model was robust even after excluding natural disasters from the traumatic events list, meaning that the model could also be applicable in regions without the profound impacts of frequent natural disasters.

Role of adolescents' demographic characteristics in the association between traumatic events and suicidality
To our knowledge, this is the first study to include various demographic information to conduct a stratified analysis when examining the association between traumatic events and suicidality in adolescence, emphasizing the vital role of these factors in the association.Note: Low traumas pattern refers to reporting relatively low probabilities of all traumatic events; high death/serious injury of a loved one (DSILO) pattern refers to reporting high probabilities of DSILO; high physical abuse pattern refers to reporting high probabilities of physical abuse; multiple traumas pattern refers to having relatively high probabilities of all traumatic events.a Model 1 was adjusted for gender and age groups.
b Model 2 was additionally adjusted for residence, only child status, leftbehind child status, family structure, parental education, and family income.
X. Shi et al.Across all demographic stratification conducted with different measures of exposures and outcomes, the female gender and individuals aged 13-15 are more susceptible to suicidality than their male counterparts in terms of traumatic exposure.Previous studies explained it by suggesting females are more empathetic, sensitive, and likely to cope with stress using internalizing strategies (Jones et al., 2022;Kwok et al., 2015).In China, adolescents aged 13 face a challenging transition from non-residential to residential schooling, coupled with an increased workload.Meanwhile, parental inflexibility in adjusting their parenting style can hinder their children from developing the necessary skills to effectively cope with adversities (Feder et al., 2019).A worldwide meta-analysis of 192 epidemiological studies demonstrated that the Four traumatic patterns were identified: Low traumas, high DSILO (Death/serious injury of a loved one), high physical abuse, and multiple traumas.For all traumatic patterns, a higher percentage of females reported high suicidality compared to males.Participants aged 13-15 had a higher percentage of high suicidality compared to other age groups, except for the low traumas pattern where the highest proportion of individuals with high suicidality was in the eldest age group (≥16 years).peak onset age for any mental disorder and stress-related disorders were 14.5 and 15.5 years respectively (Solmi et al., 2022).Our research further pointed out the transition from early to middle adolescence (13-15 years) was also a highly vulnerable period for falling victim to trauma-related suicide.On the contrary, adolescents aged ≥16 were more resilient than others investigated, which may be attributed to their gaining greater capacity in emotional regulation, active coping, and cognitive reappraisal as they grow older (Horn and Feder, 2018).Accordingly, the focus ought to be placed on preventive and early intervention strategies aimed at the primary and secondary school stage.To achieve this, school-based programs (e.g., death education) should be established combined with family-involved efforts to leverage the protective effects of school connectedness and family communication (Angelakis and Gooding, 2022).
Furthermore, particular attention should be directed towards those with vulnerable socioeconomic circumstances, such as being the only child or left-behind child, experiencing parental separation or loss, and coming from both the most and least privileged families.Such individuals may be at higher risk due to the lack of protective family support during their developmental years, which can hinder their ability to recognize danger, which may result in reduced capacity to effectively manage the effects of traumatic exposure (Cassels et al., 2018;Dhondt et al., 2019).We regard demographic conditions as modifiable factors (Hughes et al., 2017), and person-centered strategies should be introduced in the civil welfare and healthcare system to help teenagers adapt to socioeconomic changes and mitigate the adverse impact of trauma exposure.Having illustrated the linkage between demographic characteristics and traumatic events, especially certain traumatic patterns, we believe it is beneficial to clinicians, psychiatrists, and clinical psychologists to better identify and help those needy youngsters.

Limitations
First, as we focus on short-term interactions rather than long-term effects, we are incapable of inferring causality due to the nature of cross-sectional data or concluding the specific role of traumatic events in the progression of suicidality.Second, there is a potential for selection bias given the study's approach as a regional school-based mental health screening.The sample might not fully capture the diversity of the broader student population in Deyang due to unintentional exclusions.Additionally, relying solely on data from this single site could impact the study's relevance beyond the Deyang region when applied to the broader context of China.Third, recall bias is a concern due to the retrospective nature of our self-report data, and the omission of the precise timing of traumatic exposure limits our control over timevarying confounders.To enhance data accuracy, future research should cross-verify self-reported information with alternative sources such as parent reports or psychiatric records.Fourth, while our data collection system includes built-in checks to ensure response integrity, we recognize that not implementing additional quality control measures during the analysis stage may lead to potential biases such as overlooking of subtle inconsistencies in response patterns.Fifth, while merging similar traumatic events helped reduce redundancy, it may have also reduced the detail in our data.Since our findings emphasize a strong link between interpersonal violence and suicidality, future studies should focus on untangling this complex relationship, which might involve a closer look at domestic and community violence, considering different roles (witness, perpetrator, victim), and a more indepth exploration of sexual abuse, including specific types of incidents.

Conclusions
In conclusion, our population-based study provides valuable insights into the association between traumatic events and suicidality among adolescents.We observed traumatic exposure and experiencing a greater number of traumatic event types are significantly linked with higher suicidality, with interpersonal violence, particularly physical abuse, standing out.Our findings underscore the necessity of targeted prevention and intervention strategies to mitigate the effects of traumatic events on adolescent mental health, especially for those in high-risk demographic subgroups.Clinicians must be mindful of the heightened suicide risk among adolescents with a traumatic history, and traumatic patterns identified in our study can guide appropriate screening and management.Future research should concentrate on identifying effective measures to decrease the prevalence and impact of trauma in highrisk populations.

Fig. 1 .
Fig. 1.Associations between traumatic events and Suicide Behavior Questionnaire-Revised (SBQ-R) score estimated by linear regression models.a. Model 1 was adjusted for gender and age groups.b.Model 2 was additionally adjusted for residence, only child status, left-behind child status, family structure, parental education, and family income.c.Model 3 included all covariates mentioned above with mutual adjustment of all traumatic events.

Fig. 2 .
Fig. 2. Associations among traumatic patterns, Suicide Behavior Questionnaire-Revised (SBQ-R) score, and demographic characteristics.Four traumatic patterns were identified: Low traumas, high DSILO (Death/serious injury of a loved one), high physical abuse, and multiple traumas.For all traumatic patterns, a higher percentage of females reported high suicidality compared to males.Participants aged 13-15 had a higher percentage of high suicidality compared to other age groups, except for the low traumas pattern where the highest proportion of individuals with high suicidality was in the eldest age group (≥16 years).

Table 2
Associations between number of traumatic event types and Suicide Behavior Questionnaire-Revised (SBQ-R) score estimated by linear regression models.
a Model 1 was adjusted for gender and age groups.bModel 2 was additionally adjusted for residence, only child status, leftbehind child status, family structure, parental education, and family income.X.Shi et al.

Table 3
Associations between traumatic patterns and Suicide Behavior Questionnaire-Revised (SBQ-R) score estimated by linear regression models.