Establishment of a TreeNet algorithm-based model for early prediction of essential hypertension / 预防医学

ABSTRACT

Objective@#To create a model for early prediction of essential hypertension (EH) based on the TreeNet algorithm, so as to provide a tool for early monitoring of EH. @*Methods@#The health examination data were collected from individuals receiving health examinations in Hangzhou Haiqin Health Examination Center or Shanghai Yibao Health Management Co., Ltd from 2014 to 2016, and a predictive model for EH was created based on the TreeNet algorithm. The effectiveness of the model for early prediction of EH was evaluated using root mean square error (RMSE), mean absolute deviation (MAD), coefficient of determination (R2) and receiver operating characteristic (ROC) curve. @*Results@#A total of 12 variables were included in the model, and the highest contributing variable was body mass index (BMI), followed by BMI difference, two-year BMI difference, two-year triglyceride (TG) difference, two-year total cholesterol (TC) difference, high-density lipoprotein cholesterol (HDL-C) in 2014, TG in 2014, low-density lipoprotein cholesterol (LDL-C) in 2014, body weight in 2015, fasting blood glucose in 2015, TG in 2015, urea nitrogen difference and platelet in 2015. The highest predictive accuracy was 100.00%, and the lowest was 56.89%. The risk of EH significantly increased among individuals with BMI in 2015 of >25 kg/m2, two-year BMI difference of >0.5 kg/m2, two-year TG difference ranging from 1.3 to 3.3 mmol/L, TC in 2015 of 2.0 to 2.4 mmol/L and HDL-C in 2014 of <0.52 mmol/L. The model presented RMSE of 0.082, MAD of 0.064, R2 of 0.811, area under the ROC curve of 0.788 (95%CI: 0.741-0.815), sensitivity of 69.05% and specificity of 66.21% for prediction of EH@*Conclusion@# The TreeNet algorithm-based model is effective for early monitoring of high-risk individuals for EH.