When hypertension (high blood pressure), is diagnosed, the structure and function of arteries usually have changed. Therefore, if people at high risk of hypertension could be identified at early stage, intervention measures such as active exercise, reasonable diet can be introduced to reduce the incidence of hypertension.
Current complex hypertension risk assessment models have limited predictive ability and poor interpretability, which hindered its application.
Recently, researchers from Hefei Institutes of Physical Science (HFIPS), Chinese Academy of Sciences introduced new method to assess the risk of hypertension.
This method, which is called “hypertension risk assessment method based on simple risk factors", is low-cost and easy to operate. It can identify people at high risk of hypertension in the early stage with easily available lifestyle information and anthropometric information.
To establish the hypertension risk assessment model, researchers adopted the method of univariate logical regression analysis and optimized random forest.
In this research, the risk factors of hypertension were selected by univariate logical regression analysis. Following that the hyperparameters of random forest classifier were optimized by grid search. In this way, a hypertension risk prediction model was finally constructed. Aside from that, the importance of prediction variables would be calculated by the contribution of Area Under Curve (AUC) to the model to enhance the interpretability.
Comparing with traditional method with AUC 0.77-0.87, the model achieved AUC 0.92 according to clinical trial.
The study also found the top 6 risk factors of hypertension: Body Mass Index, age, family history, waistline, smoking and drinking.
This model provided an easy way for "early screening and early intervention" to prevention and control chronic diseases.
The related findings were published on Frontiers in Public Health.
“Hypertension risk assessment method based on simple risk factors" is low-cost and easy to operate.(Image by ZHAO Huanhuan)