A risk prediction model for hypertension and related diseases using echocardiographic data and machine learning algorithms

WANG Song; XU Yong; WAN Mujun

doi:10.3969/j.issn.1002-7386.2025.04.006

Abstract：Objective To construct a complication risk prediction model in hypertension patients based on echocardiographic parameters via machine learning,and to validate its performance.Methods A cross-sectional study was used to collect the clinical data of 415 patients with hypertension admitted from January 2021 to March 2024.The research objects were divided into the training set(n=311)and validation set(n=104)according to the ratio of 3∶1 by stratified random sampling.In the training set,311 patients were divided into the non-complication group(n=239)and complication group(n=72)according to the presence or absence of complications.Lasso regression method was used to screen the risk factors of complications in patients with hypertension.Machine learning algorithms were applied to construct six prediction models in the training set,including the logistic regression(LR),support vector machine(SVM),random forest(RF),extreme gradient boosting(Xgboost),naive bayes(NB)and light gradient boosting machine(lightGBM).The training set and validation set were used to validate the model.The area under the curve(AUC)of the receiver operating characteristic(ROC)curve and the calibration curve were used to evaluate the discrimination and accuracy of the model,respectively.The decision curve was used to evaluate the clinical value of the model.The optimal model was selected.Results In the training set,compared with the non-complication group,patients in the complication group were significantly older and had a higher body mass index(BMI),higher proportions of diabetes and uncontrolled blood pressure,lower number of weekly exercise,longer course of hypertension,higher systolic blood pressure(SBP),low-density lipoprotein cholesterol(LDL-C),triglyceride(TG),total cholesterol(TC),lower high-density lipoprotein cholesterol(HDL-C),greater left ventricular end-diastolic diameter(LVEDd),interventricular septal thickness(IVST),left ventricular posterior wall thickness(LVPWT),and peak E and peak A of the anterior mitral flow spectrum,and lower left ventricular ejection fraction(LVEF)(all P＜0.05).Through the dimension reduction analysis of the Lasso regression model,15 core predictors were further screened from the 17 variables for the construction of the prediction model of complications in patients with hypertension.Six machine learning models,including LR,SVM,RF,Xgboost,NB and lightGBM,were successfully constructed.The ROC curve results showed that the model constructed by Xgboost had good discrimination ability.The calibration curve showed that the model constructed by Xgboost had good prediction performance.The decision curve showed that the model constructed by Xgboost had good clinical benefit.Therefore,the prediction model constructed by Xgboost had the best performance.Conclusion Patients with hypertension have a higher risk of complications.Hypertension duration≥5 years,LVEDd thickening,history of diabetes,IVST thickening,high SBP,LVPWT thickening,older age,larger BMI,increased TC,uncontrolled blood pressure,increased TG,increased LDL-C,decreased HDL-C,increased E peak of anterior mitral flow spectrum,and decreased LVEF are risk factors for complications in patients with hypertension.Among the prediction models based on echocardiographic parameters constructed by machine learning algorithms,the Xgboost model has the strongest performance in predicting the overall risk of complications.