Construction and validation of a risk prediction model for osteoporosis in COPD patients based on explainable machine learning.

To develop a machine learning (ML)-based predictive model for identifying high-risk populations of osteoporosis among patients with chronic obstructive pulmonary disease (COPD), thereby facilitating early detection and personalized management. Using data from patients diagnosed with COPD in the MIMIC-IV database, we divided the dataset into training and validation sets at a 7:3 ratio. LASSO regression and logistic regression were applied to screen 35 variables, and 6 ML algorithms were employed to construct predictive models with internal validation. Model performance was evaluated using multiple metrics, followed by SHAP analysis for interpretability. All 6 ML models achieved high AUC values in both the training and test sets, as measured by the area under the receiver operating characteristic curve, with the XGBoost model demonstrating the highest overall performance. Feature importance analysis revealed that age, sex, and prothrombin time were the top 3 factors influencing osteoporosis risk. This study developed an interpretable ML-based risk prediction model for osteoporosis in COPD patients. The model provides clinicians with a novel tool for individualized osteoporosis risk assessment and early intervention, supporting personalized patient management and improved prognosis. However, this model currently relies on internal validation only and requires external testing in independent cohorts prior to clinical application.