Predicting Next‐Day Passive Suicidal Ideation in At‐Risk Youth
Shane Kentopp, Luke Francisco, Megan Chen, Ambuj Tewari, Ewa CzyzABSTRACT
Introduction
Passive suicidal ideation (SI) is a well‐established risk factor for suicidal behavior but has received less attention than active SI. Although recent work has leveraged intensive longitudinal data and machine learning (ML) to forecast short‐term risk for active SI, passive SI remains understudied as a prediction target.
Methods
Seventy‐eight psychiatrically hospitalized youth (ages 13–17 years) completed baseline assessments and daily ratings of risk and protective factors for 28 days post‐discharge. Multiple ML models were trained to predict the presence of next‐day passive SI. Models with and without baseline variables were compared to assess the relative predictive value of time‐varying versus baseline features.
Results
ML models predicted next‐day passive SI with high accuracy (AUC = 0.90). The strongest predictors were within‐person 7‐day moving averages of passive SI duration and frequency. Including baseline variables had negligible performance impact, even during initial days post‐discharge.
Conclusions
Short‐term passive SI remains an underutilized but important target for suicide prevention. Forecasting next‐day passive SI using ML is feasible and highly accurate. Within‐person, time‐varying features outperformed baseline factors, even in early days post‐discharge. Additional research on SI facets, such as duration, is needed. Integrating passive SI into personalized intervention frameworks may enhance the precision of suicide prevention efforts.