Key points are not available for this paper at this time.
Objective: Fracture is a critical unfavorable prognostic factor in patients with rheumatoid arthritis(RA) and osteoporosis. At present, models involving clinical indices that accurately predict fracture are still uncommon. We addressed this gap by developing machine learning (ML)-based predictive models to individualize the risk of fracture in elderly patients with RA and osteoporosis and to identify a high-risk group for fracture. Methods: 487 patients diagnosed with RA and osteoporosis at the Central Hospital of Enshi Tujia and Miao Autonomous Prefecture were randomly divided into a training cohort (used for building the model) and a validation cohort (used for validating the model). Five ML-assisted models were developed from candidate clinical features using two-step estimation methods. The receiver operating characteristic curve (ROC), decision curve analysis (DCA), and clinical impact curve (CIC) were performed to evaluate the robustness and clinical practicability of each model. Results: A total of twenty-two candidate variables were included, and the prediction model was established by an ML-based algorithm. The areas under the ROC curve (AUCs) of the random forest classifier (RFC) model, artificial neural network (ANN), support vector machine (SVM), eXtreme gradient boosting (XGBoost), decision tree (DT), probability of major osteoporotic fractures (PMOF), and probability of hip fracture (PHF) ranged from 0.695 to 0.878. Among them, RFC obtained the optimal prediction efficiency via adding serum selenium and clinical indices, that is, glucocorticoid, and erythrocyte sedimentation rate (ESR). Conclusion: Based on the classic clinical parameters, the fracture risk of RA patients with osteoporosis can be accurately predicted. In particular, RFC prediction model shows good discrimination ability in identifying high-risk patients with fracture. Keywords: rheumatoid arthritis, osteoporosis, fracture, machine learning algorithm, risk factor, predictive model
Chen et al. (Sat,) studied this question.