1. Background
Suicide-related thoughts in adolescence represent a serious mental health problem, with consequences extending beyond the individual to families, schools, and the wider community. Across many countries, suicide remains among the leading causes of death in young people, and suicide-related cognitions are widely regarded as clinically important indicators of vulnerability (1, 2). In Taiwan, suicide-related problems among school-aged adolescents have also received increasing attention, and population-based studies suggest that such experiences are not rare (3, 4). Rather than arising from a single cause, adolescent suicidality is generally understood to result from multiple interacting influences. Previous research has linked a range of behavioral and relational experiences to suicide-related outcomes, including low physical activity, prolonged sedentary behavior, unhealthy eating patterns, limited parental support, family conflict, and poor peer connectedness (5-24). However, many of these factors have been studied in isolation. Fewer studies have evaluated behavioral and interpersonal domains concurrently within the same nationally representative sample of adolescents, particularly in Taiwan. In addition, earlier findings suggest that boys and girls may not show identical patterns of association across these domains (9, 10, 24). In suicidality research, suicidal ideation and suicide planning are closely related but distinct constructs. Suicidal ideation generally refers to thoughts about death or self-harm, whereas suicide planning may reflect a more specific cognitive progression toward potential suicidal action. In population-based adolescent research, examining these experiences together may help capture a broader spectrum of suicide-related cognitive vulnerability when the aim is to identify associated factors at the group level. At the same time, these constructs should not be considered interchangeable, as they may differ in correlates and clinical implications.
2. Objectives
Against this background, this nationwide study investigated the associations of lifestyle behaviors, dietary factors, and family and peer relationship factors with the composite outcome of suicidal ideation and/or suicide planning among Taiwanese adolescents. We also examined these associations separately in boys and girls to explore potential sex-related differences in the observed patterns. Given the cross-sectional design, the purpose of this study was to identify associations rather than to establish causality. The findings may therefore inform future research on adolescent mental health and prevention-oriented initiatives relevant to Sustainable Development Goal 3.
3. Methods
3.1. Study Design and Data Source
This study used a secondary analytical approach based on data from the 2012, 2014, and 2016 waves of the Taiwan Youth Health Survey (TYHS). The TYHS is a nationwide, school-based survey implemented by the Ministry of Health and Welfare in Taiwan to monitor health-risk behaviors and psychosocial conditions among junior high school students. For the present study, the three survey waves were combined and analyzed using a cross-sectional design.
3.2. Sampling and Data Collection
The TYHS applied a stratified, two-stage cluster sampling design to obtain a nationally representative sample of junior high school students. In the first stage, schools were stratified by urban-rural population density and selected with probability proportional to size. In the second stage, classes within selected schools were sampled, and all students in those classes were invited to participate. Data were collected in classrooms using standardized administration procedures by trained personnel. The survey questionnaire was developed for adolescent health surveillance in Taiwan and incorporated items adapted from both domestic and international school-health survey instruments, including the US Youth Risk Behavior Survey and the WHO Global School-based Student Health Survey. The TYHS uses standardized survey items for adolescent health monitoring; however, because the present study was a secondary analysis of existing survey data, formal psychometric evaluation of each item within this analytic sample was not available.
3.3. Participants
A total of 18,330 adolescents were available across the three TYHS waves included in this study. Participants were removed from the analysis if data were missing for the primary outcome or key explanatory variables required for model construction. Therefore, a complete-case strategy was used. After excluding 249 records with incomplete information, the final analytic sample comprised 18,081 adolescents. To enhance transparency regarding potential bias related to case exclusion, the characteristics of included and excluded participants were compared. Because complete-case analysis may introduce selection bias if missingness is not random, this issue was considered in interpreting the findings.
3.4. Conceptual Framework and Study Variables
The analysis was guided by an integrated framework in which adolescent suicide-related cognition was considered in relation to several domains of influence. Explanatory variables were grouped into three categories: Demographic characteristics, lifestyle and dietary behaviors, and family-peer relationship factors. All variables analyzed in this study were derived from self-reported responses in the TYHS questionnaire.
3.5. Outcome Variable
The primary outcome was a composite of suicidal ideation and/or suicide planning during the past 12 months. In the original TYHS survey, these two experiences were assessed using separate questions. For the present analysis, they were combined into a single binary indicator to reflect broader suicide-related cognitive vulnerability and to support stable multivariable estimation at the population level. Adolescents who reported either suicidal ideation or suicide planning within the previous 12 months were classified into the ever group, whereas those who reported neither were classified into the never group. Because suicidal ideation and suicide planning are conceptually related but not identical, the combined outcome should be interpreted as a broad indicator rather than implying equivalence between the two constructs.
3.6. Demographic Characteristics
Demographic variables included survey year (2012, 2014, or 2016), urbanization level (urban area, city, town, or rural area), grade level (7th, 8th, or 9th grade), age (≤ 12, 13, 14, 15, or ≥ 16 years), sex, and body image satisfaction. These variables were selected to describe the study population and to account for baseline differences among adolescents.
3.7. Lifestyle Behaviors
Lifestyle variables comprised physical activity, weekday sedentary time, and non-school screen time on school days. Physical activity was defined as the number of days in the previous 7 days on which the participant engaged in at least 60 minutes of activity and was categorized as 0 days, 1 - 3 days, or 4 - 7 days. Weekday sedentary time was grouped into < 1 hour/day, 1 - 4 hours/day, 5 - 6 hours/day, and ≥ 7 hours/day. Non-school screen time on school days was categorized as none, < 2 hours/day, or ≥ 3 hours/day.
3.8. Dietary Behaviors
Dietary variables included breakfast consumption, vegetable intake, fruit intake, high-fat food intake, fast-food intake, carbonated soft drink intake, and sugary non-carbonated beverage intake. Breakfast frequency during the previous 30 days was categorized as rarely/never, sometimes/often, or always. Fruit and vegetable intake were each classified as < 1 time/day, 1 - 2 times/day, or ≥ 3 times/day. Fast-food intake during the previous 7 days was categorized as 0 days, 1 day, or ≥ 2 days. Other dietary variables were analyzed according to the original TYHS response categories shown in the tables.
3.9. Family-Peer Relationship Factors
Family-peer variables included parental understanding, parental monitoring, parental intrusion of belongings, and peer support at school, all referring to the previous 30 days. Parental understanding captured whether parents understood the adolescent’s worries and problems. Parental monitoring referred to parental knowledge of the adolescent’s leisure-time activities. Parental intrusion of belongings reflected checking belongings without permission. Peer support at school referred to whether classmates were perceived as kind and helpful. Each variable was recorded on a five-level scale: Never, rarely, sometimes, often, and always.
3.10. Statistical Analysis
Descriptive statistics were used to summarize the characteristics of the study sample. Categorical data are presented as numbers and percentages, and continuous variables are presented as means with standard deviations. Differences between included and excluded participants were examined using chi-square tests for categorical variables and independent-samples t tests for continuous variables, as appropriate. For the main analysis, adolescents in the ever and never groups were initially compared using chi-square tests. Each explanatory variable was then examined separately in univariate logistic regression models to estimate crude odds ratios (ORs) and 95% confidence intervals (CIs). To evaluate the combined associations of multiple domains with the composite outcome of suicidal ideation and/or suicide planning, hierarchical multivariable logistic regression models were constructed. Demographic variables were entered in Model 1, lifestyle and dietary variables were added in Model 2, and family-peer relationship variables were further included in Model 3. This sequential approach assessed whether associations of interest remained after progressive adjustment for additional domains of potential influence. Variables were selected based on prior literature, conceptual relevance, and availability in the TYHS dataset. Because several behavioral and family-peer variables may be interrelated, model estimates were interpreted with attention to possible overlap among exposures. Additional analyses were performed separately for boys and girls to explore sex-stratified patterns. Because these subgroup models were not accompanied by formal interaction testing, differences between male and female estimates were interpreted cautiously. All analyses were performed using SAS Enterprise Guide (SAS Institute Inc., Cary, NC, USA), and a two-sided P < 0.05 was considered statistically significant.
3.11. Ethical Considerations
Ethical approval for this study was granted by the Institutional Review Board of Chung Shan Medical University Hospital, Taichung, Taiwan (IRB No. CS2 - 22081). The TYHS data used in this study were de-identified before release, and no personally identifiable information was available to the investigators. Accordingly, the requirement for informed consent for this secondary analysis was waived by the Institutional Review Board. All study procedures were conducted in accordance with the principles of the Declaration of Helsinki.
4. Results
4.1. Characteristics of Adolescents According to Suicidal Ideation and/or Suicide Planning
The characteristics of the study population are presented in Table S1 in the Supplementary File. Among the 18,081 adolescents included in the analysis, 4,206 reported suicidal ideation and/or suicide planning during the past 12 months. Compared with adolescents in the never group, those in the ever group included a higher proportion of girls and differed by survey year, urbanization level, and grade, whereas the age distribution was comparable between the groups. Adolescents in the ever group also reported less favorable behavioral profiles, including lower physical activity, longer sedentary time, greater non-school screen exposure, less regular breakfast consumption, lower fruit intake, and more frequent consumption of high-fat foods, fast food, and sugar-sweetened beverages. Differences were also observed in the family-peer domain, with the ever group reporting lower parental understanding, lower parental monitoring, more frequent parental intrusion of belongings, and lower peer support at school. Most between-group comparisons were statistically significant; although the difference in grade was smaller in magnitude, it remained significant (P = 0.0253).
4.2. Hierarchical Multivariable Logistic Regression Findings
The results of the hierarchical multivariable logistic regression models are presented in Table S2 in the Supplementary File. In the fully adjusted model, girls had higher odds of suicidal ideation and/or suicide planning than boys. Lower odds were observed among adolescents surveyed in 2014, those living in rural areas, and those in ninth grade. Several behavioral and dietary variables also remained associated with the outcome after adjustment. Engaging in physical activity on 1 to 3 days per week, consuming breakfast regularly, eating fruit 1 to 2 times per day, and reporting high-fat food intake of less than once per day were each associated with lower odds. In contrast, fast-food intake on 2 or more days per week was associated with higher odds of the outcome. Family-peer variables showed substantial associations in the adjusted model. Higher levels of parental understanding, parental monitoring, and peer support at school were each associated with lower odds of suicidal ideation and/or suicide planning. By comparison, parental intrusion of belongings showed a positive gradient across response levels, with higher frequencies corresponding to higher odds of the outcome.
4.3. Sex-Stratified Multivariable Findings
Sex-stratified results are shown in Table S3 in the Supplementary File. Among boys, lower odds of suicidal ideation and/or suicide planning were observed for rural residence (OR, 0.823; 95% CI, 0.710 - 0.953), physical activity on 1 to 3 days per week (OR, 0.713; 95% CI, 0.595 - 0.855) or 4 to 7 days per week (OR, 0.769; 95% CI, 0.643 - 0.920), breakfast consumption reported as sometimes/often (OR, 0.728; 95% CI, 0.582 - 0.909) or always (OR, 0.564; 95% CI, 0.452 - 0.703), fruit intake of 3 or more times per day (OR, 0.806; 95% CI, 0.650 - 0.998), and high-fat food intake of less than once per day (OR, 0.809; 95% CI, 0.679 - 0.963). Higher odds were observed among boys who consumed fast food on 2 or more days per week (OR, 1.225; 95% CI, 1.019 - 1.472), as well as among those reporting parental intrusion of belongings. Among girls, lower odds were observed for survey year 2014 compared with 2012 (OR, 0.728; 95% CI, 0.641 - 0.828), non-school screen time of less than 2 hours per day (OR, 0.784; 95% CI, 0.679 - 0.905), breakfast consumption reported as sometimes/often (OR, 0.707; 95% CI, 0.559 - 0.895) or always (OR, 0.512; 95% CI, 0.405 - 0.648), and fruit intake of 1 to 2 times per day (OR, 0.820; 95% CI, 0.732 - 0.920). Higher odds were observed for residence in cities (OR, 1.218; 95% CI, 1.041 - 1.425), fast-food intake on 2 or more days per week (OR, 1.398; 95% CI, 1.138 - 1.718), and greater parental intrusion of belongings. Across both strata, stronger parental understanding, better parental monitoring, and greater peer support at school were associated with lower odds of the outcome. Parental intrusion of belongings remained positively associated with suicidal ideation and/or suicide planning in a graded manner in both boys and girls. Although the magnitude of some estimates differed between the two groups, these subgroup patterns should be interpreted cautiously.
5. Discussion
Using data from a nationwide adolescent survey in Taiwan, this study examined the associations of demographic characteristics, lifestyle behaviors, dietary patterns, and family-peer relationship factors with the composite outcome of suicidal ideation and/or suicide planning. Adolescents in the ever group generally exhibited less favorable profiles in both daily behaviors and interpersonal experiences. Even after sequential adjustment across multiple domains, several factors remained associated with the outcome, suggesting that adolescent suicide-related vulnerability may reflect a combination of behavioral and relational influences rather than any single factor alone (2, 5-11). In the behavioral domain, lower physical activity, longer sedentary time, and greater non-school screen exposure were associated with higher odds of suicidal ideation and/or suicide planning. These patterns align with earlier studies reporting poorer mental health among adolescents with less healthy movement-related behaviors (12-15). Dietary variables showed a similar pattern. More regular breakfast consumption was associated with lower odds of the outcome, whereas frequent fast-food intake was associated with higher odds. Collectively, these findings suggest that everyday health behaviors may be associated with suicide-related vulnerability in adolescence (16-22).
Associations involving family and peer contexts were particularly notable. Adolescents who perceived greater parental understanding and stronger parental monitoring tended to have lower odds of reporting suicidal ideation and/or suicide planning. In contrast, parental intrusion into belongings was associated with progressively higher odds across response levels. Peer support at school was also inversely associated with the outcome. These findings are consistent with the broader literature indicating that supportive interpersonal environments may be linked to lower emotional risk, whereas intrusive or strained relational contexts may be linked to poorer mental health outcomes (5-10, 23, 24). The sex-stratified analyses suggested that the pattern of associations was not identical in boys and girls. In boys, physical activity appeared to show a clearer inverse association, whereas in girls the association with lower non-school screen time was more apparent. Breakfast consumption was inversely associated with the outcome in both groups, and parental intrusion into belongings appeared more strongly associated among girls. These subgroup findings should be interpreted cautiously, however, because stratified models alone do not establish formal interaction effects. Rather than indicating definitive sex-specific effects, these results suggest that correlates of adolescent suicide-related vulnerability may be patterned somewhat differently across sex strata. From a public health perspective, the results highlight the potential relevance of both school and family environments when considering suicide-related vulnerability in adolescents. Although the findings do not establish causation, they may help identify behavioral and interpersonal correlates that warrant closer attention in future longitudinal or intervention-based research. They may also inform more integrated and context-sensitive inquiry into adolescent mental health promotion under Sustainable Development Goal 3.
Several limitations should be considered when interpreting these findings. First, because the data were cross-sectional, temporal direction and causality cannot be determined. Second, all measures were derived from self-report and may therefore be influenced by recall inaccuracy or social desirability. Third, suicidal ideation and suicide planning were combined into a single outcome for this analysis. Although this decision allowed the study to capture a broader spectrum of suicide-related cognitive vulnerability, the two constructs are not clinically identical and should not be interpreted as interchangeable. Fourth, the complete-case approach may have introduced selection bias if participants with missing information differed systematically from those included in the final analysis. Finally, although the hierarchical models accounted for multiple related domains, residual confounding and potential overlap among correlated exposures cannot be excluded. Despite these limitations, the large, nationally representative sample and the simultaneous consideration of behavioral and interpersonal domains add value to the current evidence base (3, 4, 25).
5.1. Conclusions
Suicidal ideation and/or suicide planning among adolescents were associated with multiple lifestyle, dietary, and family-peer factors. More regular breakfast consumption, stronger parental understanding, better parental monitoring, and greater peer support at school were linked to lower odds of the outcome, whereas frequent fast-food intake and parental intrusion into belongings were linked to higher odds. The stratified analyses also suggested that some associations may differ in magnitude between boys and girls, although these subgroup patterns require cautious interpretation. Overall, the findings highlight the importance of considering both daily behaviors and interpersonal contexts in adolescent suicide-related research and may inform future longitudinal and prevention-oriented research.
