A0404
Title: Risk profiles for severe mental health difficulty: Classification and regression tree analysis
Authors: Yoshitake Takebayashi - Institute of Statistical Mathematics (Japan) [presenting]
Takafumi Kubota - Tama University (Japan)
Tsubaki Hiroe - Institute of Statistical Mathematics (Japan)
Abstract: Severe mental health difficulty is a leading cause of suicide. Its development has been associated with several risk factors. Comprehensive approaches must be used to examine the degree to which these factors co-act and interact to develop severe mental health difficulty risk profiles. The literature review reveals no report of a study exploring interactions among severe mental health condition factors in subsets of people with high suicidal risk (mental disorders, unemployment, and caregivers of relatives) in Japan. We aim to use a classification and regression tree (CART) approach to establish risk profiles and examine their performance for diagnostic accuracy. Data were obtained from the National Comprehensive Survey of Living Conditions. Outcome measures (K6) were categorized into low, moderate, and high, applying the recommended cut-off values. Socio-demographic status, financial status, and subjective stress were included as predictors in the CART model. CART analysis results indicate that subjective stress in daily life is the strongest predictor for severe mental health difficulties in the high-suicide-risk group. Additionally, results show that all high-suicide-risk group divided into several sub-groups that reflect interactions among predictors.
