Urbanization and the Network Role of Loneliness in Adolescent Suicide Risk: A Four-Stage Analysis Across 87 Countries

Background The loneliness–suicide association has been confirmed across countries, but what macro-level conditions account for cross-national variation in its strength remains unknown. Methods Four-stage analysis of 439,026 adolescents (13–17 years; 52.1% female) from 87 countries in five non-European WHO regions using GSHS data. Stage 1: psychometric networks (EBICglasso) with regional centrality comparisons. Stage 2: cross-regional XGBoost transfer with SHAP feature importance. Stage 3: ecological correlations between 11 macro-indicators and SHAP profiles ( N  = 42 countries), with GLMM and specification curve analysis. Stage 4: urbanization–edge weight associations across 38 country-specific networks. Results Suicidal ideation ranked in the top two by Strength centrality across all WHO regions (Stage 1). Prediction models failed to transfer across regions (mean AUC degradation = 17.3 percentage points), with loneliness as the most variable predictor (Stage 2). Urbanization showed the strongest association with loneliness SHAP importance ( r _s = .655, p _FDR = .0003, N  = 42); a GLMM confirmed this at the individual level (OR = 1.096, 95% CI [1.053, 1.140], p  = 6.2 x 10^-6), with 94.4% of 432 specification curve variants yielding p < .05 (Stage 3). Urbanization was positively correlated with the loneliness–suicidal ideation edge weight ( r _s = + .639, p _FDR < .001) and negatively with the friendship–loneliness edge weight ( r _s = − .799, p _FDR < .0001; Stage 4). Conclusions The network role of loneliness in adolescent suicide risk varies with national urbanization level: in higher-urbanization contexts, loneliness’s risk connections strengthen while protective connections weaken. If confirmed longitudinally, these patterns would support development-stage-adapted prevention over uniform global templates.