Open AccessPrognostic algorithms for the progression of chronic heart failure depending on the clinical phenotype
Author(s) -
Е. А. Полунина,
Л. П. Воронина,
Е. А. Попов,
И. С. Белякова,
О. С. Полунина,
Д. С. Тарасочкина
Publication year - 2019
Publication title -
kardiovaskulârnaâ terapiâ i profilaktika
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.158
H-Index - 16
eISSN - 2619-0125
pISSN - 1728-8800
DOI - 10.15829/1728-8800-2019-3-41-47
Subject(s) - medicine , phenotype , logistic regression , heart failure , ejection fraction , clinical phenotype , algorithm , cluster (spacecraft) , cardiology , mathematics , computer science , gene , biology , biochemistry , programming language
Aim. To develop a mathematical equation (algorithm) to predict the development of chronic heart failure (CHF) for three years, depending on the clinical phenotype. Material and methods. Three hundred forty five patients with CHF with a different left ventricular ejection fraction (preserved, mean, low) were examined. The control group included somatically healthy individuals (n=60). In all patients, 48 parameters that most widely characterize the pathogenesis of CHF (gender-anamnestic, clinical, instrumental, biochemical) were analyzed. To isolate phenotypes, dispersive and cluster analysis was used: the hierarchical classification method and the k-means method. In the development of algorithms we used binary logistic regression method. We used ROC curve to assess the quality of the obtained algorithms. Results. We identified four phenotypes in patients with CHF: fibro-rigid, fibro-inflammatory, inflammatory-destructive, dilated-maladaptive. For the first three phenotypes, a mathematical logistic regression method was used to develop mathematical models for predicting the progression of CHF for three years, with the release of predictors for each phenotype. Belonging to the dilatedmaladaptive phenotype according to the results of the analysis is already an indicator of an unfavorable prognosis in patients with CHF. Conclusion. The developed algorithms based on the selected phenotypes have high diagnostic sensitivity and specificity and can be recommended for use in clinical practice.