Prognostic Model of Prehypertension Risk Based on Molecular Markers.

ABSTRACT

The prognostic models assessing the risk of prehypertension in coming 1-2-year period for 30-60-year-old subjects were developed with the help of computer recognition technology using 6 recognition methods. These models are based on the content of molecular markers in blood serum and the risk factors for the development of prehypertension in men and women who had "optimal" BP for last 3 years and in patients with newly diagnosed prehypertension. The models were compared for their prediction power. The most effective model was obtained with gradient boosting method based on the content of molecular markers. It is characterized with a high predictive power (ROC AUC=0.76), specificity (96.4%), and overall accuracy (86.6%) accompanied with close relationship between prognosis and actual symptoms of prehypertension (p=0.001).