Background: The prevalence of suicidal behavior in children and adolescents is a critical public health issue, with rates in the coffee region exceeding the national average. This study analyzed responses to the Patient Health Questionnaire (PHQ-9) in a sample of schoolchildren in Pereira, with the aim of enhance risk stratification accuracy tools vía statistic and multivariate techniques. Methods: The magnitude of the effect of depressive symptoms on suicidal ideation was estimated via the nonparametric Cliff's delta statistic, and a hierarchical clustering algorithm (Ward's method) was implemented to detect latent patterns. Results: The results indicated that Item 2 (depressed mood) and Item 6 (negative self-perception) had the greatest discriminatory power, with correlations of \((r=0.62)\) and \((r=0.58)\), respectively, regarding whether the person directly expresses thoughts of death, surpassing those of the other somatic markers. Cluster analysis allowed us to isolate a high-risk profile without previous labels. Conclusions: Results highlight items 2 (feeling down) and 6 (feeling bad about oneself) as having greater discriminatory power for identifying subjects at higher suicidal ideation risk. Clinically, these complaints are often prominent in children and adolescents, who tend to verbalize low self-esteem or loss of motivation rather than sadness, consistent with DSM diagnostic criteria for depression.