Development of a Suicide Prediction Model for the Elderly Using Health Screening Data

Area of Science:

• Gerontology
• Public Health
• Machine Learning in Healthcare

Background:

• Suicide is a significant global health concern, with a disproportionately high impact on the elderly population.
• Effective suicide risk assessment tools are crucial for targeted interventions in geriatric populations.
• Existing prediction models often rely on subjective reporting, limiting their utility in large-scale screening.

Purpose of the Study:

• To develop and validate a machine learning-based predictive model for suicide risk assessment in individuals aged over 65.
• To identify key demographic, clinical, and behavioral factors associated with suicide in the elderly.
• To evaluate the model's performance using objective healthcare utilization data.

Main Methods:

• Utilized a large-scale big data health screening cohort from the National Health Insurance Sharing Service.
• Applied machine learning techniques to a dataset of 48,047 subjects, including individuals who died by suicide.
• Analyzed demographic factors, medical history (depression), medication prescriptions (benzodiazepines), and physiological measurements (BMI, cholesterol).

Main Results:

• The suicide group comprised older individuals, with a significantly higher proportion of men.
• A history of depression and higher medication use, particularly benzodiazepines, were strongly associated with suicide.
• Lower body mass index, waist circumference, total cholesterol, and low-density lipoprotein levels were observed in the suicide group.

Conclusions:

• A machine learning model effectively predicts suicide risk in the elderly population using objective healthcare utilization data.
• The model demonstrates potential for identifying at-risk individuals without relying on subjective self-reports.
• Findings highlight the importance of considering depression history, specific medication use, and certain physiological markers in geriatric suicide prevention.