Machine learning-driven risk prediction and feature identification for major depressive disorder and its progression:...

Major depressive disorder (MDD) presents significant public health challenges due to its increasing prevalence and complex risk factors. This study systematically analyzed data from 2019 to 2023 to explore trends in MDD incidence, symptom progression, and multidimensional influencing factors.

The study aimed to identify key risk factors and mechanisms of MDD progression through multidimensional variable analysis and machine learning model evaluation, optimizing predictive models to provide a scientific basis for early screening, targeted interventions, and public health policy formulation.

Data from the U.S. National Health Interview Survey (NHIS) were analyzed using LASSO regression for feature selection and various machine learning models (e.g., AdaBoost, Gradient Boosting, Gaussian NB, Extra Trees) for risk prediction. SHAP value analysis was employed to interpret the models and identify key predictors across psychological, social, and biological domains.

Depression risk was closely associated with psychological health (e.g., PHQ82, ANXLEVEL), social support (e.g., DISCRIM1, PCNTADLT), and biological health (e.g., COGMEMDFF, DEPEV). AdaBoost, Gradient Boosting, Gaussian NB, and Extra Trees demonstrated superior performance in different tasks, with SHAP analysis highlighting the critical role of psychological health in risk prediction and the importance of social support and cognitive deficits in symptom progression.

This study provides a scientific foundation for early screening of MDD, personalized intervention strategies, and public health policy optimization. Future research should expand data coverage, optimize models, and incorporate causal inference methods to enhance the precision and applicability of depression risk prediction.