Machine Learning Reveals Social Risk Clusters Linked to Suicide Across the United States

A groundbreaking study utilizing machine learning has identified three distinct social and economic profiles associated with increased suicide risk across the U.S. Conducted by researchers at Weill Cornell Medical College and Columbia University Vagelos College of Physicians and Surgeons, the study analyzed data from 3,018 counties, examining 284 social determinants of health including poverty, housing quality, healthcare access, environmental exposures, and family stress. The researchers found significant variations in suicide rates among the identified social clusters, which differ geographically.

Published in Nature Mental Health, the research employs unsupervised machine learning — a method that discerns hidden patterns in large datasets without prior labeling or assumptions — allowing for a comprehensive understanding of community-level factors influencing suicide risks. This approach revealed three main clusters related to social determinants: "REMOTE," "COPE," and "DIVERSE."

The "REMOTE" cluster includes rural, often mountainous areas where residents tend to be elderly, living in aged or abandoned housing, with a high prevalence of firearms in suicide deaths, especially among men. "COPE" represents communities experiencing complex family dynamics, such as single parenthood or multigenerational households, often situated in the southern U.S., with higher suicide rates among middle-aged white individuals. The "DIVERSE" cluster comprises racially and ethnically diverse urban areas on the coasts, marked by income inequalities, housing costs, poor air quality, and limited healthcare access, with higher suicide rates among women, youth, and Black or Hispanic populations.

This research emphasizes that suicide prevention must extend beyond individual risk factors to address broader social and environmental conditions. tailored interventions could include enhancing mental health services and reducing social isolation in remote areas, addressing economic and substance use issues in communities facing social stress, and improving mental health support and environmental factors in diverse urban settings.

Furthermore, the study highlights that regional policies, such as Medicaid expansion, have already contributed to reductions in suicide rates. Moving forward, the researchers aim to integrate additional data, like electronic health records, to better understand and target these social clusters. These insights can lead to more effective, region-specific prevention strategies, ultimately aiding efforts to curb the rising trend of suicides across the nation.

Source: https://medicalxpress.com/news/2025-05-machine-uncovers-social-clusters-linked.html