The role of psychological factors in predicting self-rated health: implications from machine learning models.

Abstract

Self-rated health (SRH) is a significant predictor of future health outcomes. Despite the contribution of psychological factors in individuals' subjective health assessments, prior studies of machine learning-based prediction models primarily focused on health-related factors of SRH. Using the Midlife in the United States (MIDUS 2), the current study employed machine learning techniques to predict SRH based on a broad array of biological, psychological, and sociodemographic factors. Our analysis, involving logistic regression, LASSO regression, random forest, and XGBoost models, revealed robust predictive performance (AUPRC > 0.90) across all models. Emotion-related variables consistently emerged as vital predictors alongside health-related factors. The models highlighted the significance of psychological well-being, personality traits, and emotional states in determining individuals' subjective health ratings. Incorporating psychological factors into SRH prediction models offers a multifaceted perspective, enhancing our understanding of the complexities behind self-assessed health. This study underscores the necessity of considering emotional well-being alongside physical conditions in assessing and improving individuals' subjective health perceptions. Such insights hold promise for targeted interventions aimed at enhancing both physical health and emotional well-being to ameliorate subjective health assessments and potentially long-term health outcomes.