Towards Identifying of Effective Personalized Antihypertensive Treatment Rules from Electronic Health Records Data Using Classification Methods: Initial Model

Traditional clinical diagnosis and management are regulated by standards, patient management protocols with specific nosology and clinical guidelines that are limited, and their use in practice is confronted with the gap between “efficacy” and “effectiveness”. Data stored patients’ electronic health records (EHRs) provide previously unknown predictors that have affected the disease outcome, and allow to develop personalized treatment guidelines with the application of statistical methods and powerful machine learning techniques. This study aims to predict treatment effect of monotherapy with five main classes of antihypertensive drugs based on individual patients’ profiles for a single decision time point. We transform the estimation of effective personalized antihypertensive treatment rules into a classification problem, and propose the method to adapt the CART algorithm for building a decision tree for effective personalized approach to choose monotherapy in hypertension.