Development and validation of a multimodal interpretable machine learning model for the identification of osteoporosis in patients with type 2 diabetes mellitus: a multicenter retrospective study - Report - MDSpire
Advertisement
Development and validation of a multimodal interpretable machine learning model for the identification of osteoporosis in patients with type 2 diabetes mellitus: a multicenter retrospective study
Clinical Report: Multimodal Machine Learning Model for Osteoporosis in T2DM
Overview
This study developed a multimodal machine learning model to detect osteoporosis in patients with type 2 diabetes mellitus (T2DM), achieving high accuracy in identifying osteoporosis risk. The model integrates various clinical data.
Background
Osteoporosis is a significant complication in T2DM patients, increasing fracture risk and mortality. Current screening methods are often inadequate. This study utilizes machine learning to enhance osteoporosis detection in this high-risk population.
Data Highlights
Cohort
Sample Size
AUC
Training Set
852
0.877 (95% CI: 0.830–0.923)
External Validation Set
126
0.911 (95% CI: 0.879–0.943)
Key Findings
Seven predictors identified: age, hemoglobin, neutrophil count, uric acid, lymphocyte-to-HDL ratio, skeletal muscle index at L3, and metabolic score for visceral fat.
The eXtreme Gradient Boosting (XGBoost) model outperformed other algorithms in predicting osteoporosis risk.
SHAP analysis indicated age, hemoglobin, and uric acid as the most significant contributors to the model.
Calibration and decision curve analyses confirmed the model's clinical utility.
The prevalence of osteoporosis among T2DM patients in China is estimated at 35%, projected to rise to 40.8% by 2030.
Clinical Implications
Further prospective validation is necessary to establish the clinical utility of the developed model.
Conclusion
This study presents a robust machine learning model for osteoporosis risk prediction in T2DM.
Higher annual oral corticosteroid exposure was associated with greater odds of systemic adverse events, with avascular bone necrosis and pneumonia showing dose-dependent associations with cumulative dose and osteoporosis associated with longer annual exposure duration.
