Machine Learning-Based Predictive Modeling for Mental Illness Risk in Chinese Older Adults

OOBJECTIVE: To develop a machine learning model for predicting mental illness risk in Chinese older adults using longitudinal data, facilitating early identification and intervention for high-risk individuals. METHODS: Data from 8,942 participants in the 2020 China Health and Retirement Longitudinal Study (CHARLS) were used to develop and validate predictive models. Four machine learning algorithms—logistic regression, naive Bayes, support vector machine, and random forest—were applied. Model parameters were optimized using the training set, and performance was evaluated on the testing set. RESULTS: LASSO regression selected 16 key features from 41 initial variables. The random forest (RF) model achieved the best performance on the test set, with specificity of 0.720, sensitivity of 0.543, accuracy of 0.632, and AUC of 0.680. CONCLUSION: Machine learning models, particularly RF, can effectively assess mental illness risk and offer interpretable insights into key predictive features.