Personalized One-Step-Ahead Prediction of Mental Health in Early-Stage Breast Cancer Survivors Using Big Five Personality Traits and Adverse Life-Event Indicators: A Hybrid Fixed-Effects and Machine-Learning Framework.

Psychological distress in early-stage breast cancer survivors is common and can worsen treatment adherence, disease progression, and survival. We present a two-stage hybrid framework for personalized, one-step-ahead prediction of mental-health deterioration, designed to support digital-health interventions. In Stage 1, a within-patient fixed-effects regression on longitudinal mental health status scores (Waves 0–2) captures each patient’s baseline mental-health trajectory and sensitivity to adverse life events. In Stage 2, these personalized estimates are combined with with the rest factors in machine-learning classifiers (SVM, Random Forest, Gradient Boosting) to predict the mental health status in the last unobserved Wave. Across four European cohorts (N = 1,330), the framework achieved 90% accuracy and F₁ > 94% in cross-validation, with similar out-of-sample performance. Subgroup and country-level analyses confirmed robustness, while highlighting vulnerable groups where prediction is more challenging. By linking deterioration risk to actionable drivers such as social support and financial strain, the framework provides interpretable, clinically relevant insights to enable timely, personalized psychosocial interventions in pervasive healthcare settings.