| Objective To explore the risk factors for osteoporosis (OP) in patients with rheumatoid arthritis (RA) and to construct a predictive model. Methods Data of 2 311 cases from RA patients were included in this study. Among them, 302 cases (13.07%) were diagnosed with OP. The data were randomly divided into training set and validation set at ratio of 7:3. Forty-five variables were screened through LASSO and univariate and multivariate backward stepwise regression analyses. Eight variables, namely gender, age, BMI, family history of OP, history of GCs use, RBC, AST, and TG were ultimately incorporated into the construction of the predictive model. The discriminatory ability, predictive probability, and clinical utility were assessed for constructing the ROC, CCA, and DCA. Nomogram and web calculator were constructed, and its reasonableness was analyzed. The DT model was re-selected to assess its applicability. The tree structure diagram and confusion matrix were generated. Results Significant differences were observed in age, RBC, family history of OP, and history of GCs use between the OP group and the non-OP group of RA patients (P<0.05). The ROC, CCA, and DCA of the training set (AUC=0.789; HL=7.382) and the validation set (AUC=0.789; HL=8.078) demonstrated high degree of consistency. The sensitivity, specificity, and accuracy of the DT model exhibited robust performance. Conclusion This study is the first to evaluate the risk of OP among participants with RA within the NHANES database through the construction of a predictive model. The predictive model developed in this study possesses good convenience and applicability, enabling clinicians to differentiate between high-risk and low-risk populations and thereby enhancing the efficiency of disease management. |