Two approaches to modeling the risk of progressive atherosclerosis

Abstract

Progressive or accelerated atherosclerosis is accompanied by unfavorable clinical outcomes. Studying and understanding this process and creating a personalized method for assessing the risk and prognosis of this disease are necessary to optimize approaches to treatment and prevention.Aim: To compare two approaches to the creation of prognostic risk model of progressive atherosclerosis: non-linear regression model of logistic type and free cross-platform visual programming system Orange method.Material and Methods. The retrospective cohort study included 202 patients with confirmed coronary heart disease: 147 men and 55 women. The mean age of the patients was 53.3 ± 7.16 years. Group 1 included patients with myocardial infarction or unstable stenocardia, emergency arterial stenting, stroke, peripheral arterial thrombosis, critical ischemia and lower extremity amputation within 2 years before inclusion in the study. Patients in the comparison group did not have these events. Predictive models of the influence of different studied parameters on the probability of rapid progression of atherosclerosis were built using factor and correlation analysis and free cross-platform Orange visual programming system.Results. The authors’ suggested approaches to the evaluation of the risk of progressive atherosclerosis have a good prognostic accuracy (sensitivity 94.1, specificity 97.0 and accuracy 95.5 coefficients, respectively) for the regression model and 0,950 (95,0%) for the machine learning model. However, the construction of the regression model is a more complex procedure compared to the second approach, where the choice of informative indicators for the prediction model is made by Orange. Nevertheless, the above two approaches can successfully complement each other, allowing to build more accurate predictive risk models.Conclusion. The proposed authors’ approaches to assessing the risk of progressive atherosclerosis have a good prognostic accuracy.