Delineating impulsivity-based pathways to suicide deaths: A cluster analysis

Abstract

Background. The significant heterogeneity among individuals who die by suicide complicates prevention, suggesting that a “one-size-fits-all” approach is insufficient. It is crucial to identify distinct subgroups for targeted strategies. This study aims to characterize suicide profiles based on trait impulsivity and related factors. Methods. Data from the FRieNDS project (Factores de Riesgo en Defunciones por Suicidio – Risk Factors in Suicide Deaths), a psychological autopsy study of408 suicide deaths, were used. After determining the optimal number ofclusters via stability analysis through agglomerative nesting, a final cluster analysis was performed on 391 valid suicide deaths (defined as cases with no missing data on the variables used for clustering) using k-means on a lower-dimensional representation of the data encoded by an autoencoder. Key clustering variables included sex, impulsivity (Barratt Impulsivity Scale-11), aggression, intent to die, previous history of suicide attempts, history of substance abuse, psychotic and affective disorders, and the presence of a depressive episode at the time of death. Results. We identified three clusters: (1) Impulsive-aggressive (29.8%), characterized by high rates of Cluster B disorders, substance abuse, more stressful events, and low lethal intent; (2) depressive prior attempters (24.5%), which comprised mostly women and showed greater behavioural changes before death; and (3) non-impulsive/aggressive (45.7%), a group with no clear psychopathological profile, less healthcare contact, and minimal communicated intent to die, despite having few prior attempts. Conclusions. Our study identified three suicide clusters with varying impulsivity levels, high- lighting the need for tailored interventions and community-level research for better suicide prevention strategies.