<![CDATA[The comorbidity of hypertension and diabetes in the elderly is a significant health problem because it increases the risk of complications and the burden on medical services. In resource-constrained healthcare facilities, machine learning approaches can facilitate more efficient risk-based screening. This study aimed to evaluate the ability of machine learning models to predict diabetes, hypertension, and their comorbidities in elderly patients in Bangka, and to identify key clinical factors contributing to these risks. This study used medical record data from 279 elderly patients in the emergency department. Two models were compared: logistic regression and Random Forest, with predictor variables including age, sex, BMI, waist circumference, and blood pressure. The results showed that the Random Forest model had moderate discrimination ability for diabetes (AUC 0.67) and hypertension (AUC 0.64), but poor discrimination ability for comorbidities (AUC 0.56–0.57). Adiposity (BMI and waist circumference) and blood pressure were the main determinants of risk. Overall, this model has the potential to serve as an initial screening tool, though it still requires recalibration, the addition of clinical features, and external validation before operational implementation.]]>
Copyrights © 2026
