Abstract: Objective To establish a predictive model of depression risk for elderly people living alone, and to provide scientific basis for early prevention and control of depression. Methods Empirical analysis was conducted using the fifth round of China Health and Retirement Longitudinal Study (CHARLS) survey project, and the data set was randomly divided into a prediction set and a validation set with a 7∶3 ratio. Multivariate logistic regression was used to construct a predicting model of depression risk for elderly people living alone. The receiver operating characteristic(ROC) curve and the area under the curve(AUC) was used to determine the model’s discrimination, and the Hosmer-Lemeshow goodness of fit test, calibration curve graph, clinical decision curve, and clinical impact curve were used to evaluate the accuracy and clinical benefits of the model. Results The results of multiple logistic regression analysis showed that self-rated health, life satisfaction, pain and discomfort, and physical activity were the main influencing factors for depression in elderly people living alone (P<0.05). The AUC of the training set and validation set risk models in predicting depression were 0.754 (95% CI: 0.716-0.792) and 0.813 (95% CI: 0.763-0.861), respectively. The P-value of the Hosmer-Lemeshow test for the training set was 0.591, and for the validation set was 0.779(P>0.05) with good goodness of fit. The clinical decision curve and clinical impact curve suggested that the prediction model had good clinical net benefits. Conclusions Decreased self-rated health, decreased life satisfaction, physical pain and discomfort, and inability to engage in physical activity can increase the risk of depression in elderly people living alone, which provides a basis for early screening of depression in elderly people living alone.

Key words: living alone, aged, depression, risk prediction model, nomogram