Comprehensive Summary
This study compared six machine learning (ML) models to predict osteoporosis (OP) risk in Chinese patients with type 2 diabetes (T2D). 775 T2D patients were enrolled, with 70% of the data used for training and 30% for testing. Findings indicated that gradient boosting machine (GBM) performed the highest, achieving a 93% precision rate and 0.96 AUROC. Key predictors included gender, age, BMI, heart rate, and alkaline phosphatase levels, with these variables showing the strongest discriminatory power. Researchers developed a web-based calculator that provides rapid risk estimates from simple input data.
Outcomes and Implications
Underdiagnosis of OP in T2D patients is often due to inadequate standard screening methods. The GBM has potential to fill this gap by relying solely on routinely collected data. The accompanying web-based calculator allows for more streamlined OP risk assessments through a user-friendly interface. Although the model has not been clinically implemented, it demonstrates strong potential in earlier identification and intervention of OP in T2D patients.