Revisiting satisfaction with life: A hybrid study based on linear models and machine learning

Based on the Health Action Process Approach (HAPA) theory, this study examines how Health Literacy is associated with Satisfaction with Life among Chinese university students via Exercise Adherence and Rest Intolerance, from a combined perspective of linear models and machine learning. It aims to identify factors associated with university students' Satisfaction with Life within the context of "involution" (neijuan).

Data were collected via electronic questionnaires from 21,095 students across 104 Chinese universities using scales for Health Literacy (HLS-SF9), Exercise Adherence (EAS), Rest Intolerance (RIS-8), and Satisfaction with Life (SWLS). Analyses included correlation, regression, mediation tests, and machine learning algorithms (Random Forest, KNN, XGBoost) with SHAP and PDP for interpretability.

Health Literacy (r = 0.371) and Exercise Adherence (r = 0.394) positively correlated with Life Satisfaction (p < 0.001), while Health Literacy negatively correlated with Rest Intolerance (r = -0.081). The total effect of Health Literacy on Life Satisfaction was 0.503 (95% CI: 0.485, 0.520), with a direct effect of 0.334 (95% CI: 0.316, 0.352). Exercise Adherence mediated the relationship (effect = 0.172). XGBoost performed best (R² = 0.288); SHAP identified Exercise Adherence (32.6%), Health Literacy (25.4%), and Rest Intolerance (24.9%) as key predictors. PDP revealed nonlinear relationships for all three variables.

Health Literacy is directly and indirectly associated with Satisfaction with Life. The positive predictive role of Rest Intolerance suggests performative pressures within involution, highlighting the need for universities to guide students toward a balanced view of health and rest, to prevent health behaviours from becoming tools for competition.