A predictive model for depression in elderly people with arthritis based on the TRIPOD guidelines

Abstract

Objective

Through machine learning algorithms, a prediction model for depression in arthritis patients was established to provide a basis for related interventions.

Methods

Data from 4240 patients with arthritis were collected from the National Health and Nutrition Examination Survey database and divided into a training set (70 %) and a test set (30 %). LASSO Regression was employed for feature variable selection, and predictive models were constructed using five machine learning algorithms: Random Forest (RF), Logistic Regression (LR), Extreme Gradient Boosting (XGBoost), Gaussian Naive Bayes (GNB), and Gradient Boosting Decision Tree (GBDT). Model performance was evaluated through various metrics, including the Area Under the Receiver Operating Characteristic Curve (AUC), accuracy, sensitivity, cutoff, recall, Kappa, Positive Predictive Value(PPV) and Negative Predictive Value(NPV). Additionally, Shapley Additive Explanations (SHAP) analysis was conducted for personalized risk assessment.

Result

The predictive performance of the random forest model was the highest,with an area under curve (AUC) of 0.811 (SD:0.000) for the training set and 0.780 (SD:0.001) for the test set. The model identified eight significant variables associated with the occurrence of arthritis depression, including health status, difficulty standing for extended periods, difficulty getting in and out of bed, difficulty sitting for prolonged durations, education, gender, difficulty managing finances, and race. DCA demonstrated that the nomogram was clinically beneficial.

Conclusion

The predictive model developed for identifying depression in patients with arthritis exhibits high predictive ability, and good clinical applicability. This model can aid healthcare providers in the early detection of depression, thereby enabling timely interventions that can enhance patient prognosis and promote healthy aging. Future research should incorporate real-time biomarker monitoring to refine dynamic risk assessment.