Constructing prediction models and analyzing factors in suicidal ideation using machine learning, focusing on the...

Area of Science:

• Gerontology
• Public Health
• Computational Psychiatry

Background:

• Suicide among older adults is a critical public health concern in South Korea.
• Understanding suicidal ideation is crucial for timely intervention in the elderly population.
• Predictive modeling can identify at-risk individuals for targeted support.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting suicidal ideation in older South Koreans.
• To identify key socioeconomic, behavioral, and health-related factors associated with suicidal thoughts.
• To compare the performance of various machine learning algorithms and logistic regression.

Main Methods:

• Six machine learning algorithms and logistic regression were utilized.
• A hierarchical modeling approach was employed, incorporating socioeconomic, behavioral, physical health, and mental health factors across three models.
• Model fit was compared between machine learning and logistic regression analyses.

Main Results:

• The gradient boosting algorithm demonstrated superior performance in predicting suicidal ideation.
• Key predictors included household income quintile, subjective health status, oral health, exercise ability, anxiety, and depression.
• Economic and residential vulnerabilities significantly correlated with increased suicidal thoughts.

Conclusions:

• Machine learning models effectively predict suicidal ideation in older adults.
• Identifying socioeconomic and health vulnerabilities can guide targeted suicide prevention strategies.
• The hierarchical approach aids in pinpointing vulnerable populations for intervention.