Predicting Depression Risk in Physically Inactive Older Adults Using Dietary Antioxidants and Machine Learning: A...

Physically inactive older adults represent a high-risk group for depression. However, whether dietary antioxidant intake profiles can help stratify depression risk within this population has not been well established. This study aims to evaluate the predictive ability of dietary antioxidant intake for depression risk in physically inactive adults aged 60 and older using machine learning methods.

This study utilized data from the 2007-2010 and 2017-2018 cycles of the National Health and Nutrition Examination Survey (NHANES), including 2,496 physically inactive adults aged 60 years and older. A total of 44 dietary antioxidants and two composite indices: The Composite Dietary Antioxidant Index (CDAI) and the Oxidative Balance Score (OBS) were assessed. Feature selection was performed using the random forest algorithm, followed by the development of six machine learning models: Random forest, XGBoost, k-nearest neighbors, support vector machine, decision tree, and naïve Bayes. Model performance was evaluated using multiple metrics, including area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, F-beta score, and area under the precision-recall curve (PR AUC). Ten-fold cross-validation and bootstrap resampling were employed to validate model robustness. Additionally, Shapley Additive Explanations (SHAP) analysis was conducted to facilitate individualized risk interpretation.

The random forest model demonstrated the best performance, with an accuracy of 94.9%, an ROC AUC of 0.943, and a sensitivity of 99.96%. SHAP analysis identified vitamin E, luteolin, total flavonoids, copper, magnesium, and iron as the most influential predictors, all of which showed a nonlinear inverse association with depression risk. Multivariable interaction analysis revealed synergistic protective effects between vitamin E and copper, as well as between luteolin and total flavonoids. In addition, an online risk prediction tool was developed based on the model, allowing for real-time, personalized depression risk assessment upon input of key dietary antioxidant intake data.

Dietary antioxidant intake demonstrated significant value in predicting depression risk among physically inactive older adults. Key nutrients and their interactions identified through machine learning and SHAP analysis provide new evidence and practical tools for targeted nutritional interventions and early screening of depression.