Prediction models for suicide reattempts by lasso regression through machine learning models: Single versus multiple...

Area of Science:

• Psychiatry and Mental Health
• Computational Psychiatry
• Epidemiology

Background:

• Previous suicide attempts are the strongest predictor of future attempts.
• Identifying specific risk factors for suicide reattempts is crucial for effective prevention strategies.
• Machine learning (ML) offers advanced methods for detecting complex patterns in patient data.

Purpose of the Study:

• To develop and evaluate ML classification algorithms for distinguishing between single suicide attempters (SSA) and multiple suicide attempters (MSA).
• To explore sociodemographic and clinical variables that predict subsequent suicide attempts.
• To identify key risk factors for suicide reattempts in a Spanish national cohort.

Main Methods:

• A Spanish multicentre national cohort (SURVIVE) of 1443 adult patients was analyzed.
• Two logistic regression models using Least Absolute Shrinkage and Selection Operator (LASSO) with 10-fold cross-validation were developed.
• Models incorporated sociodemographic and clinical variables categorized into specific (Model I) or broad (Model II) groups.

Main Results:

• Both models successfully distinguished between SSA and MSA, outperforming chance classification (Model I: AUC=0.696, BAC=0.644; Model II: AUC=0.678, BAC=0.621).
• Key predictors for reattempts included bipolar II disorder, binge-eating disorder, and schizophrenia (Model I); and eating disorders, African birthplace, affective disorders, employment status, schizophrenia-spectrum disorders, and substance use disorders (Model II).
• Affective disorders, eating disorders, and schizophrenia-spectrum disorders were consistently important predictors of reattempts.

Conclusions:

• ML models can effectively identify patients at higher risk for suicide reattempts.
• Specific psychiatric diagnoses (affective, eating, schizophrenia-spectrum disorders) are significant predictors of suicide reattempts.
• These findings can inform tailored prevention strategies and interventions for individuals with a history of suicide attempts.