Development of risk prediction models for hypertension comorbidity in community-dwelling patients with type 2 diabetes mellitus based on machine learning
10.3760/cma.j.cn114798-20250120-00052
- VernacularTitle:基于机器学习算法构建社区2型糖尿病患者合并高血压的预测模型
- Author:
Wentao LI
1
;
Shuai JIN
;
Wenjuan GAO
;
Xinying LIU
;
Hao WU
Author Information
1. 首都医科大学全科医学与继续教育学院，北京 100069
- Publication Type:Journal Article
- Keywords:
Diabetes mellitus,type 2;
Hypertension;
Community health center;
Prediction model;
Machine learning
- From:
Chinese Journal of General Practitioners
2025;24(5):561-570
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To develop and validate risk prediction models for hypertension comorbidity in community-dwelling patients with type 2 diabetes mellitus(T2DM).Methods:The health records of 2 979 T2DM patients from two community health service centers in Fengtai District of Beijing from January 2023 to January 2024 were collected,including 2 591 cases from Fangzhuang Center(model development group) and 388 cases from Youanmen Center(external validation group). Patients in model development group were randomly assigned in a training set( n=1 813) and an internal validation set(778 cases) at a ratio of 7∶3. The risk factors associated with hypertention comorbidity in T2DM patients were identified with LASSO regression analysis,based on which risk prediction models was developed using six machine learning algorithms,including logistic regression(LR),classification and regression tree(CART),random forest(RF),extreme gradient boosting(XGB),support vector machine(SVM) and artificial neural network(ANN). The internal and external validations of the prediction models were conducted. Results:Among 2 979 patients with T2DM,2 158(72.44%) had concurrent hypertension,with 1 572 in the development set,280 in the internal validation set,306 in the external validation set. The LASSO regression identified 14 risk factors: age,educational level,occupation,medical insurance type,alcohol consumption,exercise frequency,BMI,SBP,TG/HDL-C,METS-IR,FBG,eGFR,duration of T2DM,and dyslipidemia. The nomogram model based on 14 predictive factors was constructed with XGB algorithm showed the best performance in predicting risk of hypertention for T2DM patients,showing the highest area under the curve(AUC) of 0.694(95% CI: 0.524-0.810) and effective calibration(Brier Score=0.121). Decision curve analysis confirmed the clinical utility of the predictive model. Conclusion:The risk prediction models based on machine learning algorithms have been developed in the study,which show good prediction perfomance for hiypertention comorbidity in community-dwelling T2DM patients.
