Psychological Versus Somatic Correlates of Adolescent Suicide Attempts: A 21-Year National Time-Series and Path-Analytic Study in South Korea (2005–2025)

Background: South Korea reports one of the highest adolescent suicide rates among OECD countries. Both psychological factors (perceived stress, depression) and somatic/behavioral factors (atopic disease, obesity, diet, physical activity) have been proposed as correlates of adolescent suicidality, but the extent to which their co-occurring long-term national trends reflect genuine, independent associations—rather than a shared secular (time) trend—has rarely been tested at the population level. Methods: We compiled national annual prevalence estimates (2005–2025) from the Korea Youth Risk Behavior Survey (KYRBS), disseminated via the Korean Statistical Information Service (KOSIS), for eleven indicators. We computed (1) raw Pearson correlations, (2) year-detrended partial correlations, (3) multiple and hierarchical regression, and (4) an exploratory observed-variable path model on the year-adjusted series. Results: In raw correlations, nearly all indicators were significantly associated with the suicide attempt rate (|r| = 0.30–0.92). After removing the shared time trend, only perceived stress (partial r = 0.82, p < 0.001) and depressive mood (partial r = 0.81, p < 0.001) remained strongly and independently associated with suicide attempts. Detrended residuals were approximately normal (Shapiro–Wilk p > 0.26), and the two psychological associations were confirmed by Spearman rank-based partial correlations and permutation tests (both p ≤ 0.0002). A joint regression (suicide ~ stress + depression, year-adjusted) explained 93.5% of variance (adjusted R2 = 0.923), and stress and depression added significant explanatory power beyond calendar year (ΔR2 = 0.229, p < 0.001). Conclusions: Among the indicators examined, perceived stress and depressive mood show a robust population-level association with adolescent suicide attempts that is not attributable to a shared secular trend, whereas most somatic and behavioral correlates do not. The path model is exploratory and descriptive only and does not test causal mechanisms. Findings are ecological (population-level) and cannot establish—or exclude—individual-level causation.