Health and Family Factors Predicting Suicidal Ideation Among Middle-Aged Korean Adults: An Explainable Machine Learning Approach
- Author:
Hyeon-gyeong JO
1
;
Hae-Young KIM
;
Ki-Bong CHOI
;
Young-Sun KIM
;
Young-Bin SEO
;
HoJung AHN
;
Sunmi SONG
;
Junesun KIM
Author Information
- Publication Type:Original Article
- From:Psychiatry Investigation 2026;23(1):164-171
- CountryRepublic of Korea
- Language:English
-
Abstract:
Objective:Research specifically targeting suicidal ideation (SI) in middle-aged populations remains limited. This study aimed to predict future and concurrent SI in middle-aged Korean adults by applying four machine learning (ML) models to a nationally representative longitudinal dataset.
Methods:We analyzed data from 8,992 individuals aged 40–64 years who participated in the Korea Welfare Panel Study from the 7th (2011) to the 18th (2022) waves. Four ML algorithms were employed to develop the predictive models. The SHapley Additive exPlanations method was applied to enhance explainability.
Results:Approximately half of the participants’ mean age was 49.3±8.2 years (range, 40–64 years) and 52.2% were male. The average annual SI rate between 2011 and 2022 was 2.8%±1.2%. Predictive performance for future SI was satisfactory, with area under the receiver operating characteristic curve (AUC) values of up to 0.806 (logistic regression, LR). Predictions for concurrent SI demonstrated AUC values of up to 0.907 (LR). Key predictors of future SI included subjective health status, satisfaction with family and spousal relationships, housing environment, and educational attainment. Concurrent SI was strongly associated with immediate stressors such as family violence and income dissatisfaction.
Conclusion:The ML models demonstrated good-to-excellent predictive performance for SI. These findings emphasize the importance of health, family, and socioeconomic factors, alongside mental health indicators in the prevention of SI among middle-aged adults. Building on these findings, tailored intervention strategies that comprehensively address multidimensional risk factors are essential for effective SI prevention.
