Stratification algorithm for persons being at cardiovascular progression risk from among patients with essential arterial hypertension after 12-month follow-up, and its validation

Abstract. Introduction. The development and progression of essential arterial hypertension is caused by different pathogenetic factors/predictors, and the study thereof opens up the prospect of improving the diagnosis, treatment and prevention. Aim. To develop a personalized algorithm to stratify persons being at cardiovascular progression risk from among patients with stage I arterial hypertension after 12-month follow-up and to validate it. Material and Methods. A cohort of patients with stage I arterial hypertension and uncontrolled course (n=90) was selected for the study. After 12 months, patients were divided into two subgroups based on the cardiovascular risk dynamics: With increased cardiovascular risk and with constant cardiovascular risk. Klotho protein levels were analyzed in blood serum by enzyme immunoassay. The statistical analysis was carried out using the IBM SPSS Statistics 26.0 program (USA). Results and Discussion. The final model of classification tree includes the following predictors: Presence of concomitant pathology, body mass index, and Klotho protein levels. Based on the classification tree obtained, we have created an algorithm that allows the stratification of patients with stage I arterial hypertension and a high probability of cardiovascular risk progression after 12 months. To assess the quality of our proposed algorithm, it was validated based on a sample of 48 patients. Patients were randomly selected according to inclusion and non-inclusion criteria. When validating the algorithm, 43 out of 48 cases were correctly recognized, that is, the percentage of correct predictions was 89.6%. Conclusions. An algorithm was developed to stratify persons being at risk of cardiovascular progression among patients with stage I arterial hypertension after 12-month follow-up with the sensitivity of 93.3% and specificity of 100.0%. The correct forecast for the developed algorithm validation was 89.6%.